{ "meta": { "title": "Modern Gait Recognition: A Survey from Deep Learning to Foundation Models", "updated": "2026-10-14", "coverage": "January 2020-September 2026, with foundational and related references", "description": "The authors curate the literature, assess the evidence and take responsibility for the survey's interpretations. Language-model tools assisted literature organization, metadata extraction and wording of catalogue summaries, which were checked against their sources. Results remain attributable to the original studies.", "wordCloudMethod": "Document frequency in the three summary fields: a term contributes at most once per paper. Counts describe this collection, not citation impact or a complete publication census.", "counts": { "total": 281, "complete": 281 }, "githubOwner": "ChaoFan996", "githubRepo": "gait-survey", "statisticsDescription": "The smaller 2020–2025 modality archive defines membership. Specified venues only; corrected publication identity removes exact duplicates; withdrawn and wearable-only studies are excluded. Distinct conference/journal publications are separate." }, "papers": [ { "id": "xiong2026cod2-aaai-2026", "citationKey": "xiong2026cod2", "archiveKey": "xiong2026cod2", "method": "CoD2", "title": "Gait Recognition via Collaborating Discriminative and Generative Diffusion Models", "authors": "Haijun Xiong and Bin Feng and Bang Wang and Xinggang Wang and Wenyu Liu", "year": 2026, "venue": "AAAI", "motivation": "Generation and identity discrimination provide complementary supervision.", "idea": "Couple identity-conditioned generation with discriminative learning.", "techniques": "Discriminative-generative collaboration; identity conditioning; low-level detail modeling.", "inputs": "Gait sequences (exact representation not extracted)", "backbone": "Diffusion model + discriminative extractor", "loss": "", "datasets": "SUSTech1K; CCPG; GREW; Gait3D", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://ojs.aaai.org/index.php/AAAI/article/view/38096", "bibtex": "@inproceedings{xiong2026cod2,\n title = {Gait Recognition via Collaborating Discriminative and Generative Diffusion Models},\n author = {Haijun Xiong and Bin Feng and Bang Wang and Xinggang Wang and Wenyu Liu},\n year = {2026},\n booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},\n doi = {10.1609/aaai.v40i13.38096},\n url = {https://ojs.aaai.org/index.php/AAAI/article/view/38096}\n}", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Detailed loss, complete evaluation protocols and code execution not audited.", "sections": [ "Introduction / Motivation", "Advances in Methods / Gait Learning Frameworks", "Challenges and Outlooks / Gait Foundation Models" ], "abstract": "CoD2 links a discriminative gait encoder with a diffusion-based generator. Identity-aware semantics and lower-level appearance and motion cues condition sequence generation, while generated examples support representation learning. Experiments on SUSTech1K, CCPG, GREW and Gait3D evaluate this reciprocal interaction and its compatibility with existing discriminative recognizers.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://ojs.aaai.org/index.php/AAAI/article/view/38096", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "lu2026walking-aaai-2026", "citationKey": "lu2026walking", "archiveKey": "lu2026walking", "method": "EMGaitNet", "title": "Walking Further: Semantic-Aware Multimodal Gait Recognition Under Long-Range Conditions", "authors": "Zhiyang Lu and Wen Jiang and Tianren Wu and Zhichao Wang and Changwang Zhang and Siqi Shen and Ming Cheng", "year": 2026, "venue": "AAAI", "motivation": "Distance changes make multimodal gait features difficult to align.", "idea": "Use semantic guidance to align and combine complementary observations.", "techniques": "Semantic-guided alignment; cross-attention; multimodal fusion.", "inputs": "Camera images; LiDAR", "backbone": "CLIP + symmetric cross-attention + spatiotemporal module", "loss": "", "datasets": "LRGait; others not extracted", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://ojs.aaai.org/index.php/AAAI/article/view/37703", "bibtex": "@inproceedings{lu2026walking,\n title = {Walking Further: Semantic-Aware Multimodal Gait Recognition Under Long-Range Conditions},\n author = {Zhiyang Lu and Wen Jiang and Tianren Wu and Zhichao Wang and Changwang Zhang and Siqi Shen and Ming Cheng},\n year = {2026},\n booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},\n doi = {10.1609/aaai.v40i9.37703},\n url = {https://ojs.aaai.org/index.php/AAAI/article/view/37703}\n}", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Detailed loss, complete evaluation protocols and code execution not audited.", "sections": [ "Advances in Methods / Ideal Gait Modalities", "Challenges and Outlooks / Real-World Gait Recognition" ], "abstract": "LRGait introduces a camera-and-LiDAR benchmark for long-range gait recognition. The proposed EMGaitNet uses body-part semantics to align image and point-cloud features, then combines them through symmetric cross-attention and temporal modeling. Experiments examine multimodal recognition across outdoor distances and environments where short-range training assumptions may fail.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://ojs.aaai.org/index.php/AAAI/article/view/37703", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "hou2026gat-aaai-2026", "citationKey": "hou2026gat", "archiveKey": "hou2026gat", "method": "GaT", "title": "Gait Transformer: End-to-End Transformer Backbone for Gait Recognition", "authors": "Saihui Hou and Wenpeng Lang and Jilong Wang and Yan Huang and Liang Wang and Yongzhen Huang", "year": 2026, "venue": "AAAI", "motivation": "Sparse silhouettes require effective spatial and temporal dependency modeling.", "idea": "Design a gait-specific Transformer that learns directly from silhouettes.", "techniques": "Shallow convolutional stem; spatiotemporal attention; end-to-end training.", "inputs": "Silhouette", "backbone": "Transformer + shallow convolutional stem", "loss": "", "datasets": "Gait3D; GREW; CCGR-Mini", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://ojs.aaai.org/index.php/AAAI/article/view/42479", "bibtex": "@inproceedings{hou2026gat,\n title = {Gait Transformer: End-to-End Transformer Backbone for Gait Recognition},\n author = {Saihui Hou and Wenpeng Lang and Jilong Wang and Yan Huang and Liang Wang and Yongzhen Huang},\n year = {2026},\n booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},\n doi = {10.1609/aaai.v40i6.42479},\n url = {https://ojs.aaai.org/index.php/AAAI/article/view/42479}\n}", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Detailed loss, complete evaluation protocols and code execution not audited.", "sections": [ "Advances in Benchmarks / State-of-the-Art Performance", "Advances in Methods / Strong Gait Backbones" ], "abstract": "GaT adapts a Transformer backbone to sparse silhouette sequences. Convolutional patch embedding, separate short- and long-range token mixing, and hybrid positional encoding support spatial and temporal gait modeling. The study evaluates training from scratch on Gait3D, GREW and CCGR-MINI, examining whether a gait-specific design can work without external pretraining.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://ojs.aaai.org/index.php/AAAI/article/view/42479", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "cosma2026scaling-aaai-2026", "citationKey": "cosma2026scaling", "archiveKey": "cosma2026scaling", "method": "GaitPT scaling study", "title": "On Model and Data Scaling for Skeleton-based Self-Supervised Gait Recognition", "authors": "Adrian Cosma and Andy Eduard Catruna and Emilian Radoi", "year": 2026, "venue": "AAAI", "motivation": "The effects of pretraining scale on gait transfer remain insufficiently characterized.", "idea": "Study data, model and compute scaling in skeleton-based pretraining.", "techniques": "Self-supervised skeleton modeling; scaling analysis; zero-shot evaluation.", "inputs": "Skeleton", "backbone": "Transformer", "loss": "", "datasets": "2.7M pretraining sequences; CASIA-B, PsyMo, GREW, Gait3D", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://ojs.aaai.org/index.php/AAAI/article/view/37340", "bibtex": "@inproceedings{cosma2026scaling,\n title = {On Model and Data Scaling for Skeleton-based Self-Supervised Gait Recognition},\n author = {Adrian Cosma and Andy Eduard Catruna and Emilian Radoi},\n year = {2026},\n booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},\n doi = {10.1609/aaai.v40i5.37340},\n url = {https://ojs.aaai.org/index.php/AAAI/article/view/37340}\n}", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Sequence count does not establish unique identity count. Scaling trends depend on the tested representation, objectives and compute budget.", "sections": [ "Introduction / Motivation", "Advances in Methods / Gait Learning Frameworks", "Challenges and Outlooks / Gait Foundation Models" ], "abstract": "This study measures how skeleton-based self-supervised gait recognition changes with training data, model capacity and computation. GaitPT variants are pretrained on 2.7 million walking sequences and evaluated without target-domain training across four benchmarks. Controlled comparisons with GaitFormer separate architectural effects from scale, providing empirical guidance on allocating data and compute.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://ojs.aaai.org/index.php/AAAI/article/view/37340", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "wu2026lmgait-aaai-2026", "citationKey": "wu2026lmgait", "archiveKey": "wu2026lmgait", "method": "LMGait", "title": "Language-Guided and Motion-Aware Gait Representation for Generalizable Recognition", "authors": "ZhengXian Wu and Chuanrui Zhang and Shen'Ao Jiang and Hangrui Xu and Zirui Liao and Luyuan Zhang and Li Huaqiu and Peng Jiao and Haoqian Wang", "year": 2026, "venue": "AAAI", "motivation": "Clothing changes can obscure identity-related motion.", "idea": "Use language semantics to guide motion representation learning.", "techniques": "Language-guided motion alignment; semantic supervision; clothing-robust features.", "inputs": "Gait images; text priors (exact visual representation not extracted)", "backbone": "Motion Awareness Module + Motion Temporal Capture Module", "loss": "", "datasets": "CCPG; SUSTech1K; CASIA-B*", "codeAvailable": "Yes", "codeUrl": "https://github.com/DingWu1021/LMGait", "url": "https://ojs.aaai.org/index.php/AAAI/article/view/38063", "bibtex": "@inproceedings{wu2026lmgait,\n title = {Language-Guided and Motion-Aware Gait Representation for Generalizable Recognition},\n author = {ZhengXian Wu and Chuanrui Zhang and Shen'Ao Jiang and Hangrui Xu and Zirui Liao and Luyuan Zhang and Li Huaqiu and Peng Jiao and Haoqian Wang},\n year = {2026},\n booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},\n doi = {10.1609/aaai.v40i13.38063},\n url = {https://ojs.aaai.org/index.php/AAAI/article/view/38063}\n}", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "CASIA-B* must remain distinct from CASIA-B; abstract's 'first' claim not adopted as survey conclusion.", "sections": [ "Advances in Methods / Gait Learning Frameworks" ], "abstract": "LMGait uses gait-related language as a semantic guide for recognizing identity under covariate changes. A motion-awareness module adapts textual features to visual observations, while temporal motion capture emphasizes discriminative movement. Evaluations across several datasets examine whether this guidance reduces reliance on static appearance cues.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://ojs.aaai.org/index.php/AAAI/article/view/38063", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "hamgait2026-array-2026", "citationKey": "hamgait2026", "archiveKey": "hamgait2026", "method": "HAMGait", "title": "HAMGait: A Hybrid Attention-Mamba Network for Cross-View Gait Recognition", "authors": "Hyebin Kim and Junhyug Noh and Taeyong Lee", "year": 2026, "venue": "Array", "motivation": "Local body dynamics and long-range dependencies must be modeled together.", "idea": "Combine attention with state-space sequence modeling.", "techniques": "Hybrid attention; Mamba blocks; local and global feature aggregation.", "inputs": "Silhouette", "backbone": "Mamba + attention", "loss": "", "datasets": "CASIA-B; SUSTech1K", "codeAvailable": "Yes", "codeUrl": "https://github.com/hyebinny/HAMGait", "url": "https://www.sciencedirect.com/science/article/pii/S2590005626003838", "bibtex": "", "scope": "Related research", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "September 2026 issue; exact online date not extracted; outside historical venue scope.", "sections": [ "IV.B" ], "authorsSource": "Public scholarly metadata", "abstract": "HAMGait combines attention and state-space sequence modeling to capture long-range walking patterns without indiscriminately expanding the temporal receptive field. A hybrid module processes whole-sequence and body-part tokens, while dual gated Mamba branches learn complementary representations. Evaluations on CASIA-B and SUSTech1K examine viewpoint, clothing and carrying variations.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.sciencedirect.com/science/article/pii/S2590005626003838", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "glcgait2026-bspc-2026", "citationKey": "glcgait2026", "archiveKey": "glcgait2026", "method": "GLC-Gait", "title": "GLC-Gait: A lightweight framework for cross-view gait recognition via global-local feature fusion", "authors": "Junhao Cheng and Chao Tang and Huosheng Hu and Anyang Tong and Miaohui Zhang and Wenjian Wang", "year": 2026, "venue": "BSPC", "motivation": "Fine-grained gait modeling can introduce excessive computation.", "idea": "Refine local motion while retaining global context in a compact model.", "techniques": "Global-local fusion; regional 3D processing; entropy regularization.", "inputs": "", "backbone": "Residual backbone + region-aware 3D convolution", "loss": "", "datasets": "CASIA-B; OU-MVLP; CASIA-C", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://www.sciencedirect.com/science/article/pii/S1746809426012176", "bibtex": "", "scope": "Related research", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Issue 2026-09-01; benchmark protocol and modality require full-text audit.", "sections": [ "IV.B", "IV.E" ], "authorsSource": "Public scholarly metadata", "abstract": "GLC-Gait targets compact cross-view recognition by combining residual global features with region-specific 3D convolutions. Lightweight channel attention selects complementary features, followed by sequence aggregation and BNNeck normalization. An entropy penalty promotes confident predictions without extra inference parameters. Experiments cover CASIA-B, OU-MVLP and CASIA-C to assess accuracy alongside model size.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.sciencedirect.com/science/article/pii/S1746809426012176", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "wang2026flowgait-chi-2026", "citationKey": "wang2026flowgait", "archiveKey": "wang2026flowgait", "method": "FlowGait", "title": "FlowGait: Enabling Robust Long-Term Gait Recognition Across Real-World Covariates with mmWave Radar", "authors": "Dequan Wang and Chenming He and Lingyu Wang and Chengzhen Meng and Xiaoran Fan and Yanyong Zhang", "year": 2026, "venue": "CHI", "motivation": "Radar gait models encounter shifts across days, paths and walking conditions.", "idea": "Adapt continually using newly observed walking data.", "techniques": "Millimeter-wave sensing; self-training; continual adaptation.", "inputs": "mmWave spectrogram", "backbone": "Transformer", "loss": "", "datasets": "47 volunteers; cross-covariate / route / day datasets", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://doi.org/10.1145/3772318.3790623", "bibtex": "", "scope": "Related research", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Nonvisual scope extension. Abstract and conclusion have inconsistent cross-route/day percentages; do not insert metrics into SOTA.", "sections": [ "IV.C", "V.A" ], "abstract": "FlowGait studies the degradation of radar-based identity recognition as a person's walking conditions change over time. It combines a spectrogram Transformer with continual self-training and hierarchical pseudo-labeling of unlabeled observations. Data collected from 47 volunteers across routes, covariates and days are used to evaluate adaptation and the retention of recognition performance.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://doi.org/10.1145/3772318.3790623", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "xu2026eventgait-cvpr-2026", "citationKey": "xu2026eventgait", "archiveKey": "xu2026eventgait", "method": "EventGait", "title": "EventGait: Towards Robust Gait Recognition with Event Streams", "authors": "Senyan Xu and Shuai Chen and Chuanfu Shen and Kean Liu and Zhijing Sun and Chengzhi Cao and Xueyang Fu", "year": 2026, "venue": "CVPR", "motivation": "Exploit event-based motion cues", "idea": "Complement dynamics and shape", "techniques": "Spiking experts and cross-modal alignment", "inputs": "Event", "backbone": "Spiking experts + dense shape encoder", "loss": "", "datasets": "SUSTech1K-E; CCGR-Mini-E; real-event benchmarks (names not fully extracted)", "codeAvailable": "Yes", "codeUrl": "https://github.com/QUEAHREN/EventGait", "url": "https://openaccess.thecvf.com/content/CVPR2026/html/Xu_EventGait_Towards_Robust_Gait_Recognition_with_Event_Streams_CVPR_2026_paper.html", "bibtex": "@inproceedings{xu2026eventgait,\n title = {EventGait: Towards Robust Gait Recognition with Event Streams},\n author = {Senyan Xu and Shuai Chen and Chuanfu Shen and Kean Liu and Zhijing Sun and Chengzhi Cao and Xueyang Fu},\n year = {2026},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n url = {https://openaccess.thecvf.com/content/CVPR2026/html/Xu_EventGait_Towards_Robust_Gait_Recognition_with_Event_Streams_CVPR_2026_paper.html}\n}", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Manuscript Table I" } ], "summaryStatus": "Manuscript summary", "notes": "SUSTech1K-E and CCGR-Mini-E are synthetic event derivatives. Training uses a vision foundation teacher; event input at inference is separate from training supervision.", "sections": [ "Background / Historical Evolution", "Advances in Benchmarks / Popular Datasets", "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Handling Real-World Noise" ], "abstract": "EventGait separates event-stream dynamics from body-shape information. Spiking experts model motion at different timescales, while alignment with visual foundation features supports structural representation. Synthetic event benchmarks supplement real recordings. The study evaluates performance under normal and difficult lighting, examining information lost by conventional long-window event aggregation.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2026/html/Xu_EventGait_Towards_Robust_Gait_Recognition_with_Event_Streams_CVPR_2026_paper.html", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "cai2026barbiegait-cvpr-2026", "citationKey": "cai2026barbiegait", "archiveKey": "cai2026barbiegait", "method": "GaitCLIF", "title": "BarbieGait: An Identity-Consistent Synthetic Human Dataset with Versatile Cloth-Changing for Gait Recognition", "authors": "Qingyuan Cai and Saihui Hou and Xuecai Hu and Yongzhen Huang", "year": 2026, "venue": "CVPR", "motivation": "Identity under clothing changes", "idea": "Clothing-invariant feature learning", "techniques": "Gait normalization; BarbieGait resource", "inputs": "Silhouette", "backbone": "GON-enhanced P3D/3D", "loss": "", "datasets": "BarbieGait; CCPG; SUSTech1K", "codeAvailable": "Yes", "codeUrl": "https://github.com/BarbieGait/BarbieGait", "url": "https://arxiv.org/abs/2604.12221", "bibtex": "@inproceedings{cai2026barbiegait,\n title = {BarbieGait: An Identity-Consistent Synthetic Human Dataset with Versatile Cloth-Changing for Gait Recognition},\n author = {Qingyuan Cai and Saihui Hou and Xuecai Hu and Yongzhen Huang},\n year = {2026},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n url = {https://arxiv.org/abs/2604.12221}\n}", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Paper first page + project method description" }, { "source": "Manuscript Table I" } ], "summaryStatus": "Manuscript summary", "notes": "Dataset modalities include RGB, 2D pose and silhouette; GaitCLIF input must not be conflated with all released modalities. Real-to-synthetic identities are linked.", "sections": [ "Background / Historical Evolution", "Background / Broader Applications", "Advances in Benchmarks / Popular Datasets", "Advances in Benchmarks / State-of-the-Art Performance", "Advances in Benchmarks / Public Codebases", "Advances in Methods / Gait Learning Frameworks", "Advances in Methods / Handling Real-World Noise", "Challenges and Outlooks / Gait Foundation Models" ], "abstract": "BarbieGait transfers real subjects into a controllable virtual environment to generate extensive clothing changes while retaining their walking identity. The dataset enables experiments that are difficult to stage with physical clothing collections. A GaitCLIF baseline learns clothing-invariant features and is evaluated on both synthetic and existing gait benchmarks.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2604.12221", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "huang2026gaitmax-cvpr-2026", "citationKey": "huang2026gaitmax", "archiveKey": "huang2026gaitmax", "method": "GaitMax", "title": "Unlocking Motion from Large Vision Models with a Semantic and Kinematic Duality for Gait Recognition", "authors": "Zhanbo Huang and Dingqiang Ye and Xiaoming Liu and Yu Kong", "year": 2026, "venue": "CVPR", "motivation": "Recover motion from large-model features", "idea": "Semantic and kinematic duality", "techniques": "Part trajectories; conditional decorrelation", "inputs": "RGB; language supervision", "backbone": "Large vision model + spatiotemporal attention", "loss": "Conditional Decorrelation Loss (auxiliary); identity loss not extracted", "datasets": "GCaption (annotation resource); full evaluation list not extracted", "codeAvailable": "Yes", "codeUrl": "https://zbhuang.com/gait-max", "url": "https://openaccess.thecvf.com/content/CVPR2026/papers/Huang_Unlocking_Motion_from_Large_Vision_Models_with_a_Semantic_and_CVPR_2026_paper.pdf", "bibtex": "@inproceedings{huang2026gaitmax,\n title = {Unlocking Motion from Large Vision Models with a Semantic and Kinematic Duality for Gait Recognition},\n author = {Zhanbo Huang and Dingqiang Ye and Xiaoming Liu and Yu Kong},\n year = {2026},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n url = {https://openaccess.thecvf.com/content/CVPR2026/papers/Huang_Unlocking_Motion_from_Large_Vision_Models_with_a_Semantic_and_CVPR_2026_paper.pdf}\n}", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary paper first page" }, { "source": "Manuscript Table I" } ], "summaryStatus": "Manuscript summary", "notes": "GCaption adds annotations to existing videos, not an independent new acquisition dataset; precise inference inputs need final full-text audit.", "sections": [ "Background / Historical Evolution", "Advances in Benchmarks / State-of-the-Art Performance", "Advances in Methods / Gait Learning Frameworks", "Advances in Methods / Gait Metric Designs", "Challenges and Outlooks / Real-World Gait Recognition", "Challenges and Outlooks / Gait Foundation Models" ], "abstract": "GaitMax combines global semantic features with ordered, part-level motion trajectories. Because detailed temporal features may also encode clothing or viewpoint shortcuts, the framework uses language-supervised conditional decorrelation. Its GCaption resource supplies richer descriptions of nuisance conditions across gait datasets, enabling both the proposed training objective and further context-aware analysis.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2026/papers/Huang_Unlocking_Motion_from_Large_Vision_Models_with_a_Semantic_and_CVPR_2026_paper.pdf", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "zheng2026hypergait-cvpr-2026", "citationKey": "zheng2026hypergait", "archiveKey": "zheng2026hypergait", "method": "HyperGait", "title": "HyperGait: Unleashing the Power of Parsing for Gait Recognition in the Wild via Hypergraph", "authors": "Jinkai Zheng and Jiaqing Wei and Xinxiang Jin and Yaoqi Sun and Xichun Sheng and Ming Li and Liangqiong Qu and Xinchen Liu and Wu Liu", "year": 2026, "venue": "CVPR", "motivation": "Higher-order body-region relations", "idea": "Parsing-based relational modeling", "techniques": "Spatial and temporal hypergraphs", "inputs": "Parsing", "backbone": "Hypergraph convolution", "loss": "", "datasets": "Gait3D; SUSTech1K", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2026/html/Zheng_HyperGait_Unleashing_the_Power_of_Parsing_for_Gait_Recognition_in_CVPR_2026_paper.html", "bibtex": "@inproceedings{zheng2026hypergait,\n title = {HyperGait: Unleashing the Power of Parsing for Gait Recognition in the Wild via Hypergraph},\n author = {Jinkai Zheng and Jiaqing Wei and Xinxiang Jin and Yaoqi Sun and Xichun Sheng and Ming Li and Liangqiong Qu and Xinchen Liu and Wu Liu},\n year = {2026},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n url = {https://openaccess.thecvf.com/content/CVPR2026/html/Zheng_HyperGait_Unleashing_the_Power_of_Parsing_for_Gait_Recognition_in_CVPR_2026_paper.html}\n}", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Manuscript Table I" } ], "summaryStatus": "Manuscript summary", "notes": "Detailed loss, complete evaluation protocols and code execution not audited.", "sections": [ "Background / Historical Evolution", "Advances in Methods / Ideal Gait Modalities" ], "abstract": "HyperGait models higher-order relationships in human parsing sequences. Spatial hypergraph convolutions connect body-part features, temporal hypergraph convolutions relate their evolution, and a global head provides complementary information. Evaluations on Gait3D and SUSTech1K test whether these interactions improve recognition from parsing alone in challenging scenes.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2026/html/Zheng_HyperGait_Unleashing_the_Power_of_Parsing_for_Gait_Recognition_in_CVPR_2026_paper.html", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "xu2026ms2gait-cvpr-2026", "citationKey": "xu2026ms2gait", "archiveKey": "xu2026ms2gait", "method": "MS2Gait", "title": "MS^2Gait: A Multi-Scale Spatio-Temporal Fusion Network for LiDAR-based Gait Recognition", "authors": "Shenyin Xu and Yishan Wang and Xinyu Li and Rui Liu and Zhongyuan Wang and Xin Tian", "year": 2026, "venue": "CVPR", "motivation": "Uniform feature processing overlooks spatial scales and temporal variation.", "idea": "Combine hierarchical spatial semantics with selective temporal enhancement.", "techniques": "Multi-scale spatial extraction; nonuniform temporal modeling; feature aggregation.", "inputs": "Point (LiDAR)", "backbone": "Transformer + point feature extractor", "loss": "", "datasets": "SUSTech1K; FreeGait", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2026/papers/Xu_MS2Gait_A_Multi-Scale_Spatio-Temporal_Fusion_Network_for_LiDAR-based_Gait_Recognition_CVPR_2026_paper.pdf", "bibtex": "@inproceedings{xu2026ms2gait,\n title = {MS$^2$Gait: A Multi-Scale Spatio-Temporal Fusion Network for LiDAR-based Gait Recognition},\n author = {Shenyin Xu and Yishan Wang and Xinyu Li and Rui Liu and Zhongyuan Wang and Xin Tian},\n year = {2026},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n url = {https://openaccess.thecvf.com/content/CVPR2026/papers/Xu_MS2Gait_A_Multi-Scale_Spatio-Temporal_Fusion_Network_for_LiDAR-based_Gait_Recognition_CVPR_2026_paper.pdf}\n}", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary paper first page" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Detailed loss, complete evaluation protocols and code execution not audited.", "sections": [ "Advances in Methods / Ideal Gait Modalities" ], "abstract": "MS2Gait learns directly from LiDAR point clouds. Hierarchical spatial interactions connect body regions at several scales, while a similarity-driven temporal Transformer weights observations according to motion coherence. Evaluations on SUSTech1K and FreeGait examine recognition accuracy and robustness when spatial information is incomplete or temporal observations are heterogeneous.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2026/papers/Xu_MS2Gait_A_Multi-Scale_Spatio-Temporal_Fusion_Network_for_LiDAR-based_Gait_Recognition_CVPR_2026_paper.pdf", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "wang2026mmgait-cvpr-2026", "citationKey": "wang2026mmgait", "archiveKey": "wang2026mmgait", "method": "OmniGait", "title": "MMGait: Towards Multi-Modal Gait Recognition", "authors": "Chenye Wang and Qingyuan Cai and Saihui Hou and Aoqi Li and Yongzhen Huang", "year": 2026, "venue": "CVPR", "motivation": "Heterogeneous modality availability", "idea": "Shared multimodal embedding", "techniques": "Unified modality configurations", "inputs": "Silhouette; Depth; IR; LiDAR projected depth; Radar projected depth", "backbone": "Modal-specific encoders + shared layers", "loss": "Cross-entropy + triplet (cross-modal baseline; configuration-specific)", "datasets": "MMGait; SUSTech1K", "codeAvailable": "Yes", "codeUrl": "https://github.com/BNU-IVC/MMGait", "url": "https://arxiv.org/abs/2604.15979", "bibtex": "@inproceedings{wang2026mmgait,\n title = {MMGait: Towards Multi-Modal Gait Recognition},\n author = {Chenye Wang and Qingyuan Cai and Saihui Hou and Aoqi Li and Yongzhen Huang},\n year = {2026},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n url = {https://arxiv.org/abs/2604.15979}\n}", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Full-text HTML §§2–4" }, { "source": "Manuscript Table I" } ], "summaryStatus": "Manuscript summary", "notes": "334,060 counts modality-specific sequences, not independent walks. OmniGait uses projected image representations. Dataset includes 12 modalities from 5 sensors; do not count 12 inputs for every method.", "sections": [ "Background / Historical Evolution", "Background / Broader Applications", "Advances in Benchmarks / Popular Datasets", "Advances in Benchmarks / Public Codebases", "Advances in Methods / Ideal Gait Modalities", "Challenges and Outlooks / Real-World Gait Recognition" ], "abstract": "MMGait combines five sensor types into a benchmark containing twelve modalities. It evaluates recognition within a modality, across modalities and through multimodal fusion. The OmniGait baseline learns a shared embedding for varying input configurations, enabling systematic analysis of sensor complementarity and heterogeneous probe-gallery conditions.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2604.15979", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "lu2026textguided-cvpr-2026", "citationKey": "lu2026textguided", "archiveKey": "lu2026textguided", "method": "TCFDNet", "title": "Text-guided Feature Disentanglement for Cross-modal Gait Recognition", "authors": "Zhiyang Lu and Ming Cheng", "year": 2026, "venue": "CVPR", "motivation": "Multimodal features mix shared identity cues with modality-specific information.", "idea": "Use text prototypes to guide cross-modal feature decomposition.", "techniques": "Text-guided disentanglement; CLIP semantics; feature stability constraints.", "inputs": "Camera images; LiDAR; text priors", "backbone": "CLIP + feature stability module", "loss": "", "datasets": "SUSTech1K; FreeGait", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2026/html/Lu_Text-guided_Feature_Disentanglement_for_Cross-modal_Gait_Recognition_CVPR_2026_paper.html", "bibtex": "@inproceedings{lu2026textguided,\n title = {Text-guided Feature Disentanglement for Cross-modal Gait Recognition},\n author = {Zhiyang Lu and Ming Cheng},\n year = {2026},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n url = {https://openaccess.thecvf.com/content/CVPR2026/html/Lu_Text-guided_Feature_Disentanglement_for_Cross-modal_Gait_Recognition_CVPR_2026_paper.html}\n}", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Camera representation and text use at inference require full-text coding; do not automatically treat text priors as a new captured modality.", "sections": [ "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Gait Learning Frameworks" ], "abstract": "TCFDNet uses language descriptions to align gait information across cameras and LiDAR. A modality-aware text dictionary guides separation of shared and sensor-specific features; orthogonality constraints and a stability module refine the shared representation. Cross-modal patch exchange provides further augmentation. SUSTech1K and FreeGait experiments assess heterogeneous probe-gallery matching.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2026/html/Lu_Text-guided_Feature_Disentanglement_for_Cross-modal_Gait_Recognition_CVPR_2026_paper.html", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "dao2026forces-cvprw-2026", "citationKey": "dao2026forces", "archiveKey": "dao2026forces", "method": "Physical force augmentation", "title": "Beyond Motion Patterns: An Empirical Study of Physical Forces for Human Motion Understanding", "authors": "Anh Dao and Manh Tran and Yufei Zhang and Xiaoming Liu and Zijun Cui", "year": 2026, "venue": "CVPR Workshops", "motivation": "Motion trajectories alone may omit useful physical information.", "idea": "Complement movement representations with force-related cues.", "techniques": "Physical-force augmentation; motion representation learning.", "inputs": "Motion + inferred forces", "backbone": "Multiple baseline networks", "loss": "", "datasets": "CASIA-B; action/captioning benchmarks", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2026W/PhysHuman/html/Dao_Beyond_Motion_Patterns_An_Empirical_Study_of_Physical_Forces_for_CVPRW_2026_paper.html", "bibtex": "", "scope": "Related research", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Workshop paper with several downstream tasks; separate from main-conference counts.", "sections": [ "IV.C", "V.C", "V.D" ], "abstract": "The study asks whether inferred physical forces add information beyond visual appearance and kinematics. It augments existing systems with force estimates and evaluates gait recognition, action recognition and video captioning across eight benchmarks. The analysis examines when biomechanical signals provide complementary information, including challenging clothing and motion conditions.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2026W/PhysHuman/html/Dao_Beyond_Motion_Patterns_An_Empirical_Study_of_Physical_Forces_for_CVPRW_2026_paper.html", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "shen2026childrengait-eccv-2026", "citationKey": "shen2026childrengait", "archiveKey": "shen2026childrengait", "method": "ChildGait-Video", "title": "Decoding Children's Gait Behavior", "authors": "Yifan Shen and Boyi Li and Meihuan Huang and Yuanzhe Liu and Xu Cao and Jinyang Jin and Zhengyuan Li and Anglin Liu and Junho Kim and Jingyuan Zhu and Fangzhou Lan and Jianguo Cao and Jintai Chen and Ismini Lourentzou and James M. Rehg", "year": 2026, "venue": "ECCV", "motivation": "Pediatric gait abnormalities require fine-grained behavioral evaluation.", "idea": "Study abnormal walking through a dedicated children's gait resource.", "techniques": "Video-based pediatric gait analysis; fine-grained behavior assessment.", "inputs": "RGB; pose annotations", "backbone": "Video framework", "loss": "", "datasets": "CGV", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://github.com/PediaMedAI/ChildrenGait", "bibtex": "", "scope": "Related research", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Author repository and lab acceptance announcement; clinical behavior task; >1,100 videos / 110 subjects; not an identity benchmark.", "sections": [ "II.D", "III.A", "V.C" ], "abstract": "ChildrenGait examines how video models capture children's walking behavior, including clinically relevant motor differences. The study introduces a multi-view collection spanning ages 3–17, with high-frame-rate videos and anonymized pose data. Evaluations identify limitations of existing gait foundation models and multimodal language models, motivating a dedicated video baseline for pediatric gait analysis.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://github.com/PediaMedAI/ChildrenGait", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ye2026foundationgait-eccv-2026", "citationKey": "ye2026foundationgait", "archiveKey": "ye2026foundationgait", "method": "FoundationGait", "title": "Silhouette-based Gait Foundation Model", "authors": "Dingqiang Ye and Chao Fan and Kartik Narayan and Bingzhe Wu and Chengwen Luo and Jianqiang Li and Vishal M. Patel", "year": 2026, "venue": "ECCV", "motivation": "Dataset-specific training limits reuse across gait tasks and domains.", "idea": "Pretrain a transferable gait representation at scale.", "techniques": "Self-supervised pretraining; cross-dataset transfer; task adaptation.", "inputs": "Silhouette", "backbone": "CNN (DeepGaitV2 family)", "loss": "", "datasets": "12-source pretraining corpus; Gait3D; OU-MVLP; others require version audit", "codeAvailable": "Yes", "codeUrl": "https://github.com/ShiqiYu/OpenGait", "url": "https://arxiv.org/abs/2512.00691", "bibtex": "@inproceedings{ye2026foundationgait,\n title = {Silhouette-based Gait Foundation Model},\n author = {Dingqiang Ye and Chao Fan and Kartik Narayan and Bingzhe Wu and Chengwen Luo and Jianqiang Li and Vishal M. Patel},\n year = {2026},\n booktitle = {European Conference on Computer Vision},\n url = {https://arxiv.org/abs/2512.00691}\n}", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "ECCV 2026 acceptance confirmed by author repository; keep separate from published counts. v2 dated 2026-08-20; first-publication year 2025.", "sections": [ "Advances in Benchmarks / Popular Datasets", "Advances in Benchmarks / State-of-the-Art Performance", "Advances in Methods / Gait Learning Frameworks", "Challenges and Outlooks / Gait Foundation Models" ], "abstract": "FoundationGait studies scalable, self-supervised pretraining for gait understanding. Models trained on more than two million sequences from twelve public datasets are transferred to identity recognition, health-related prediction and attribute estimation. Experiments examine both zero-shot use and fine-tuning, asking whether shared silhouette representations can support tasks previously treated separately.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2512.00691", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "xu2026msglfgcn-eswa-2026", "citationKey": "xu2026msglfgcn", "archiveKey": "xu2026msglfgcn", "method": "MS-GLFGCN", "title": "MS-GLFGCN: Multi-stream global-local fusion graph convolutional network for skeleton-based gait recognition", "authors": "Zhanhong Xu and Xiaoyan Wu and Bin Wu and Jianning Wu", "year": 2026, "venue": "ESWA", "motivation": "Skeleton models must capture local motion and whole-body coordination.", "idea": "Integrate multiple streams of global and local skeletal features.", "techniques": "Graph convolution; multi-stream processing; global-local fusion.", "inputs": "Skeleton", "backbone": "GCN", "loss": "", "datasets": "", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://www.sciencedirect.com/science/article/abs/pii/S0957417425042022", "bibtex": "", "scope": "Related research", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Issue 2026-03-15; DOI contains 2025; exact first-online date not extracted.", "sections": [ "IV.B" ], "abstract": "MS-GLFGCN represents skeletal locomotion through three complementary graph streams. Spatiotemporal graph convolutions, pooling, residual connections and a fusion module combine local and global joint interactions. Evaluation on Kinect recordings investigates recognition with limited samples and compares both classification accuracy and computation against alternative models.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.sciencedirect.com/science/article/abs/pii/S0957417425042022", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "fonua2026acoustic-electronics-2026", "citationKey": "fonua2026acoustic", "archiveKey": "fonua2026acoustic", "method": "Acoustic CNN benchmark", "title": "CNN-Based Acoustic Gait Recognition: A Benchmarking Framework", "authors": "Ilaisaane Tilisa Fonua and Shahram Latifi", "year": 2026, "venue": "Electronics", "motivation": "Acoustic identity recognition needs evaluation as the subject population grows.", "idea": "Assess footstep-based recognition under varying population sizes.", "techniques": "Acoustic gait signals; convolutional classification; population-scale evaluation.", "inputs": "Audio mel-spectrogram", "backbone": "CNN", "loss": "", "datasets": "AFPILD", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://www.mdpi.com/2079-9292/15/12/2658", "bibtex": "", "scope": "Related research", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Nonvisual modality; identity pools 10–40; keep separate from visual benchmark counts.", "sections": [ "III.A", "IV.A" ], "abstract": "This study evaluates person identification from recorded footsteps using a compact convolutional network and mel-spectrogram inputs. Experiments on AFPILD vary the number of identities, repeat cross-validation across random seeds, and compare against conventional audio and recurrent baselines. Noise experiments assess how recognition changes when acoustic conditions become less favorable.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.mdpi.com/2079-9292/15/12/2658", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "xu2026psgait-icassp-2026", "citationKey": "xu2026psgait", "archiveKey": "xu2026psgait", "method": "PSGait", "title": "PSGait: Gait Recognition using Parsing Skeleton", "authors": "Hangrui Xu and Zhengxian Wu and Chuanrui Zhang and Zhuohong Chen and Zhifang Liu and Peng Jiao and Haoqian Wang", "year": 2026, "venue": "ICASSP", "motivation": "Parsing maps and skeletons capture complementary body information.", "idea": "Use skeleton guidance to enrich parsing-based gait representations.", "techniques": "Skeleton-guided body-part modeling; representation fusion.", "inputs": "Parsing skeleton; Silhouette", "backbone": "Plug-in representation; downstream gait model", "loss": "", "datasets": "Not fully extracted", "codeAvailable": "Yes", "codeUrl": "https://github.com/realHarryX/PSGait", "url": "https://www.cmsworkshops.com/ICASSP2026/view_session.php?SessionID=1464&bare=1", "bibtex": "", "scope": "Related research", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "ICASSP official program confirms title and authors; earlier arXiv 2503.12047 title/authors differ, use proceedings version for final citation.", "sections": [ "IV.A", "IV.C" ], "abstract": "PSGait introduces a parsing-skeleton representation that uses skeletal structure to guide fine-grained human parsing. It aims to preserve more articulated body information than binary silhouettes or sparse joints alone. The representation can be incorporated into existing gait pipelines, and the study evaluates its contribution to recognition under challenging observation conditions.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://github.com/realHarryX/PSGait", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "qin2026ggcn-jvcir-2026", "citationKey": "qin2026ggcn", "archiveKey": "qin2026ggcn", "method": "GGCN", "title": "GGCN: Gait Recognition with Generate Network and Convolutional Neural Network", "authors": "Hao Qin and Zhenxue Chen and Qingqiang Guo and Q. M. Jonathan Wu and Mengxu Lu", "year": 2026, "venue": "JVCIR", "motivation": "Covariates distort the mapping from gait observations to identity.", "idea": "Combine generation and supervised feature mapping to reduce nuisance variation.", "techniques": "Generative modeling; supervised representation mapping.", "inputs": "Multiple gait sequence types (not fully extracted)", "backbone": "CNN", "loss": "", "datasets": "", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://www.sciencedirect.com/science/article/pii/S1047320326000854", "bibtex": "", "scope": "Related research", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "April 2026 issue; outside historical venue list.", "sections": [ "IV.C", "IV.D" ], "abstract": "GGCN addresses covariate interference with a three-stage recognition pipeline. A generative component refines low-level sequence information, an encoder forms higher-level features, and supervised mapping produces the identity representation. Experiments on CASIA-B, OU-LP and OU-MVLP assess whether this combination improves recognition across varying walking conditions.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.sciencedirect.com/science/article/pii/S1047320326000854", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "zou2026gallery-kbs-2026", "citationKey": "zou2026ccgr3d", "archiveKey": "zou2026gallery", "method": "GaitNet-G3D", "title": "3D Gallery for cross-covariate gait recognition", "authors": "Shinan Zou and Chengyu Long and Fan Guo and Jin Tang", "year": 2026, "venue": "KBS", "motivation": "A fixed 2D gallery provides limited coverage of probe viewpoints.", "idea": "Project a 3D gallery toward the predicted probe view before matching.", "techniques": "View prediction; 3D-to-2D projection; gallery-based retrieval.", "inputs": "Silhouette; 3D gallery", "backbone": "Silhouette network + view predictor", "loss": "", "datasets": "CCGR-3D", "codeAvailable": "Yes", "codeUrl": "https://github.com/ShinanZou/CCGR", "url": "https://www.sciencedirect.com/science/article/pii/S0950705126003187", "bibtex": "@article{zou2026ccgr3d,\n title = {3D Gallery for Cross-Covariate Gait Recognition},\n author = {Zou, Shinan and Long, Chengyu and Guo, Fan and Tang, Jin},\n year = {2026},\n journal = {Knowledge-Based Systems},\n volume = {339},\n pages = {115576},\n doi = {10.1016/j.knosys.2026.115576}\n}", "scope": "Related research", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "KBS not in latest main-table venue list. CCGR-3D reports 228 subjects / 12,312 sequences; evaluation snippet says 115 train + 86 test, requiring split clarification.", "sections": [ "Advances in Benchmarks / Popular Datasets" ], "abstract": "This work replaces a conventional 2D gallery with reconstructed 3D bodies that can be projected to match a probe's viewpoint. GaitNet-G3D predicts the view before selecting the corresponding projection. The accompanying CCGR-3D dataset supports evaluation under cross-view and full-body clothing changes, including analysis of retrieval cost and dependence on cross-view training pairs.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.sciencedirect.com/science/article/pii/S0950705126003187", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "zhang2026disrec-pr-2026", "citationKey": "zhang2026disrec", "archiveKey": "zhang2026disrec", "method": "GDN / DCRN / DisRec", "title": "Understanding gait recognition through silhouette sequence disentanglement and fine-grained visualization", "authors": "Shaoxiong Zhang and Yixiu Liu and Jinkai Zheng and Liangqiong Qu and Ming Li and Chenggang Yan", "year": 2026, "venue": "PR", "motivation": "The contributions of static, dynamic and residual information are difficult to separate.", "idea": "Explicitly decompose gait features into interpretable components.", "techniques": "Koopman-constrained dynamics; static-dynamic decomposition; residual modeling.", "inputs": "Silhouette", "backbone": "Decomposition and reconstruction framework", "loss": "", "datasets": "OU-MVLP; Gait3D", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://www.sciencedirect.com/science/article/abs/pii/S0031320326005844", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "November 2026 issue; first-online date needs verification before cutoff inclusion.", "sections": [ "IV.C", "V.A" ], "abstract": "The paper investigates which silhouette information contributes to identity decisions. GDN uses Koopman-inspired constraints to separate static, dynamic and residual components; DCRN evaluates their recognition value. DisRec then visualizes influential pixels within these components. Case studies on OU-MVLP and Gait3D connect quantitative recognition tests with more detailed visual explanations.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.sciencedirect.com/science/article/pii/S0031320326005844", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "freire2026multiyear-pr-2026", "citationKey": "freire2026multiyear", "archiveKey": "freire2026multiyear", "method": "Gait + HAR fusion", "title": "Multi-year long-term person re-identification using gait and HAR features", "authors": "David Freire-Obregon and Oliverio J. Santana and Javier Lorenzo-Navarro and Daniel Hernandez-Sosa and Modesto Castrillon-Santana", "year": 2026, "venue": "PR", "motivation": "Identity recognition must remain useful across long time intervals.", "idea": "Combine gait and action representations for longitudinal recognition.", "techniques": "Cross-attention fusion; gait and activity features; multiyear evaluation.", "inputs": "Gait sequences; video HAR features", "backbone": "BiLSTM + pretrained video encoder", "loss": "Triplet loss (primary paper first page)", "datasets": "Outdoor ultradistance-runner dataset (2020/2023)", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://www.sciencedirect.com/science/article/pii/S0031320325012907", "bibtex": "@article{freire2026multiyear,\n title = {Multi-year long-term person re-identification using gait and HAR features},\n author = {David Freire-Obregon and Oliverio J. Santana and Javier Lorenzo-Navarro and Daniel Hernandez-Sosa and Modesto Castrillon-Santana},\n year = {2026},\n journal = {Pattern Recognition},\n doi = {10.1016/j.patcog.2025.112627},\n url = {https://www.sciencedirect.com/science/article/pii/S0031320325012907}\n}", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "2026 issue / 2025 online. Adjacent gait-assisted Re-ID fits existing FLAG/MOJO scope, but cannot be mixed into standard walking-protocol SOTA.", "sections": [ "Challenges and Outlooks / Real-World Gait Recognition" ], "abstract": "This study examines person re-identification across multiple years using complementary gait and human-action features. A recurrent gait encoder represents walking dynamics, while pretrained video models supply distilled behavioral embeddings. Cross-attention connects the two streams, supporting an investigation of identity information that persists over extended observation gaps.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://accedacris.ulpgc.es/bitstream/10553/150431/1/1-s2.0-S0031320325012907-main.pdf", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "wang2026gaitadapt-pr-2026", "citationKey": "wang2026gaitadapt", "archiveKey": "wang2026gaitadapt", "method": "GaitAdapter", "title": "GaitAdapt: Continual learning for evolving gait recognition", "authors": "Jingjie Wang and Shunli Zhang and Xiang Wei and Senmao Tian", "year": 2026, "venue": "PR", "motivation": "Adapting to new gait domains can erase previously learned knowledge.", "idea": "Update the recognizer through replay-free continual learning.", "techniques": "Continual adaptation; forgetting control; gait adapters.", "inputs": "", "backbone": "GPAK + EDSN", "loss": "", "datasets": "", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://www.sciencedirect.com/science/article/pii/S0031320326007818", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Issue November 2026, later than cutoff; need first-online date <=2026-09-09 before counting. Search crawl date is not publication evidence.", "sections": [ "IV.C", "V.A" ], "abstract": "GaitAdapt introduces continual gait-recognition settings in which models encounter new identities or domains without discarding earlier knowledge. Its replay-free GaitAdapter accumulates part-level patterns through graph-based knowledge storage. A distance-stability constraint preserves relationships among classes as tasks change. Experiments evaluate both adaptation to new tasks and retention on previous ones.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.sciencedirect.com/science/article/pii/S0031320326007818", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ryu2026gwcpj-pr-2026", "citationKey": "ryu2026gwcpj", "archiveKey": "ryu2026gwcpj", "method": "GwCPJ", "title": "Individual identification using a novel 3D gait model from human joints and Joint-Features Set Abstraction", "authors": "Jegoon Ryu and Tatsuya Kawakami and Sei-ichiro Kamata", "year": 2026, "venue": "PR", "motivation": "Frame loss and sparse pose observations weaken sequence modeling.", "idea": "Represent walking through complementary joint-feature sets.", "techniques": "Joint-feature descriptors; set abstraction; frame-drop robustness.", "inputs": "3D joints", "backbone": "", "loss": "", "datasets": "CASIA-B; OU-MVLP; Gait3D", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://www.sciencedirect.com/science/article/pii/S0031320326008344", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "December 2026 issue; first-online date needs verification before counting.", "sections": [ "IV.A", "IV.B", "IV.D" ], "abstract": "This framework represents 3D joints through complementary spatial and temporal structures. Convergent and progressive joint models combine anatomical relationships with motion trajectories, while a joint-feature set-abstraction network learns identity descriptors. The study evaluates these representations and frame dropping under viewpoint, clothing and carrying variations.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.sciencedirect.com/science/article/abs/pii/S0031320326008344", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "wang2026synthesis-pr-2026", "citationKey": "wang2026synthesis", "archiveKey": "wang2026synthesis", "method": "Skeleton-to-silhouette synthesis", "title": "Silhouette generation from identity-level skeleton synthesis for improved gait recognition", "authors": "Yang Wang and Yan Huang and Caifeng Shan and Yan Huang and Liang Wang", "year": 2026, "venue": "PR", "motivation": "Training data offer limited identity diversity and plausible shape variation.", "idea": "Synthesize additional identities while preserving body structure.", "techniques": "Skeleton-guided silhouette synthesis; identity-level data augmentation.", "inputs": "Skeleton -> Silhouette", "backbone": "Skeleton synthesizer + GAN", "loss": "", "datasets": "", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://www.sciencedirect.com/science/article/pii/S0031320326012938", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "December 2026 issue; first-online date needs verification. Two distinct authors named Yan Huang appear in metadata.", "sections": [ "IV.C", "V.B" ], "abstract": "Ske2Sil augments gait training data at the identity level. A parameter-free synthesizer modifies skeletal proportions to create proxy identities; a dual-encoder adversarial network converts their skeleton sequences into silhouettes. Identity supervision supports discriminative synthesis. Tests on four gait benchmarks examine the benefit to existing recognizers, including settings with limited training data.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.sciencedirect.com/science/article/pii/S0031320326012938", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "zheng2026gaitpr-prl-2026", "citationKey": "zheng2026gaitpr", "archiveKey": "zheng2026gaitpr", "method": "GaitPR", "title": "GaitPR: A high-confidence multi-granular residual attention network for gait recognition", "authors": "Xiaona Zheng and Qintai Hu and Shuping Zhao and Jigang Wu", "year": 2026, "venue": "PRL", "motivation": "Fine motion and multi-scale spatial cues are not equally captured by one stream.", "idea": "Fuse complementary spatial and temporal representations.", "techniques": "Parallel multi-scale fusion; residual attention; fine-grained motion modeling.", "inputs": "", "backbone": "CNN + attention", "loss": "", "datasets": "CASIA-B*; SUSTech1K", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://www.sciencedirect.com/science/article/pii/S0167865526001492", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Issue July 2026. Do not label the backbone ViT based only on the keyword list.", "sections": [ "IV.B", "IV.D" ], "abstract": "GaitPR combines parallel convolutions at different scales with attention-based feature fusion. Residual attention emphasizes regions relevant to identity while preserving global body structure and local motion. Experiments on CASIA-B* and SUSTech1K assess the resulting representations under appearance changes, complex scenes and differing viewpoints.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.sciencedirect.com/science/article/abs/pii/S0167865526001492", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "privacy2026beyondface-sensors-2026", "citationKey": "privacy2026beyondface", "archiveKey": "privacy2026beyondface", "method": "Gait privacy evaluation", "title": "Privacy Beyond the Face: Assessing Gait Privacy Through Realistic Anonymization in Industrial Monitoring", "authors": "Sarah Weiß and Christopher Bonenberger and Tobias Niedermaier and Maik Knof and Markus Schneider", "year": 2026, "venue": "Sensors", "motivation": "Removing appearance details may leave recognizable walking patterns.", "idea": "Evaluate gait identity leakage after full-body anonymization.", "techniques": "Anonymization evaluation; gait-based identity attacks; privacy-utility analysis.", "inputs": "Original / anonymized body videos", "backbone": "DeepPrivacy2 + gait recognition evaluation", "loss": "", "datasets": "Custom industrial CCTV-like dataset", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://www.mdpi.com/1424-8220/26/1/187", "bibtex": "", "scope": "Related research", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "2026 issue; exact online date not extracted. Evidence about gait identity leakage, not proof all anonymization fails.", "sections": [ "VI" ], "authorsSource": "Public scholarly metadata", "abstract": "This study tests whether realistic full-body anonymization also hides identity carried by pose and gait. Original and DeepPrivacy2-anonymized industrial recordings are compared through keypoint changes, recognition performance and cross-domain matching. The results indicate that removing visual appearance alone may leave substantial gait-based linkability.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.mdpi.com/1424-8220/26/1/187", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "adeli2026gaitgen-wacv-2026", "citationKey": "adeli2026gaitgen", "archiveKey": "adeli2026gaitgen", "method": "GAITGen", "title": "GAITGen: Disentangled Motion-Pathology Impaired Gait Generative Model -- Bringing Motion Generation to the Clinical Domain", "authors": "Vida Adeli and Soroush Mehraban and Majid Mirmehdi and Alan Whone and Benjamin Filtjens and Amirhossein Dadashzadeh and Alfonso Fasano and Andrea Iaboni and Babak Taati", "year": 2026, "venue": "WACV", "motivation": "Clinical gait generation needs control over pathological variation.", "idea": "Separate motion content from pathology severity during generation.", "techniques": "Disentangled motion representation; pathology-conditioned generation.", "inputs": "Clinical motion sequences", "backbone": "Conditional residual VQ-VAE + transformers", "loss": "", "datasets": "", "codeAvailable": "Yes", "codeUrl": "https://github.com/TaatiTeam/GAITGen", "url": "https://openaccess.thecvf.com/content/WACV2026/html/Adeli_GAITGen_Disentangled_Motion-Pathology_Impaired_Gait_Generative_Model_--_Bringing_Motion_WACV_2026_paper.html", "bibtex": "@inproceedings{adeli2026gaitgen,\n title = {GAITGen: Disentangled Motion-Pathology Impaired Gait Generative Model -- Bringing Motion Generation to the Clinical Domain},\n author = {Vida Adeli and Soroush Mehraban and Majid Mirmehdi and Alan Whone and Benjamin Filtjens and Amirhossein Dadashzadeh and Alfonso Fasano and Andrea Iaboni and Babak Taati},\n year = {2026},\n booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},\n url = {https://openaccess.thecvf.com/content/WACV2026/html/Adeli_GAITGen_Disentangled_Motion-Pathology_Impaired_Gait_Generative_Model_--_Bringing_Motion_WACV_2026_paper.html}\n}", "scope": "Related research", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Clinical motion generation; exclude from identity-method denominator.", "sections": [ "Challenges and Outlooks / Gait Representation Learning for Healthcare" ], "abstract": "GAITGen generates gait sequences conditioned on impairment severity. A conditional quantized autoencoder separates motion and pathology factors, and Transformer components generate new sequences. Experiments on PD-GaM evaluate reconstruction, generation and downstream severity estimation; a clinical user study assesses the plausibility of generated walking patterns.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://openaccess.thecvf.com/content/WACV2026/html/Adeli_GAITGen_Disentangled_Motion-Pathology_Impaired_Gait_Generative_Model_--_Bringing_Motion_WACV_2026_paper.html", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "sayera2026robustgait-wacv-2026", "citationKey": "sayera2026robustgait", "archiveKey": "sayera2026robustgait", "method": "RobustGait", "title": "RobustGait: Robustness Analysis for Appearance Based Gait Recognition", "authors": "Reeshoon Sayera and Akash Kumar and Sirshapan Mitra and Prudvi Kamtam and Yogesh S Rawat", "year": 2026, "venue": "WACV", "motivation": "Video corruption can propagate through segmentation into gait features.", "idea": "Evaluate robustness across corruption and preprocessing choices.", "techniques": "Controlled corruptions; segmentation-aware evaluation; preprocessing analysis.", "inputs": "RGB corruption -> Silhouette", "backbone": "Six gait systems + multiple silhouette extractors", "loss": "", "datasets": "CASIA-B; CCPG; SUSTech1K; MEVID", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://arxiv.org/abs/2511.13065", "bibtex": "@inproceedings{sayera2026robustgait,\n title = {RobustGait: Robustness Analysis for Appearance Based Gait Recognition},\n author = {Reeshoon Sayera and Akash Kumar and Sirshapan Mitra and Prudvi Kamtam and Yogesh S Rawat},\n year = {2026},\n booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},\n url = {https://arxiv.org/abs/2511.13065}\n}", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "15 corruption types x 5 severities; benchmark variants are not independent acquired datasets. First preprint 2025; proceedings year 2026.", "sections": [ "Advances in Methods / Handling Real-World Noise", "Challenges and Outlooks / Real-World Gait Recognition" ], "abstract": "RobustGait evaluates how appearance-based recognizers respond to digital, environmental, temporal and occlusion perturbations. It varies corruption severity, silhouette extraction and recognition architecture, tracing degradation from RGB input through preprocessing. Evaluations also investigate noise-aware training and distillation as ways to improve robustness under deployment-like conditions.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2511.13065", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "vishwakarma2026scoligaitx-wacv-2026", "citationKey": "vishwakarma2026scoligaitx", "archiveKey": "vishwakarma2026scoligaitx", "method": "ScoliGaitX", "title": "ScoliGaitX: A Deep Multi-Modal Fusion Network for Scoliosis Assessment via Gait Video Analysis", "authors": "Kaushik Vishwakarma and Aditya Nigam", "year": 2026, "venue": "WACV", "motivation": "Scoliosis assessment may benefit from complementary walking cues.", "idea": "Combine multiple gait representations for scoliosis classification.", "techniques": "Clinical multimodal fusion; video-based gait analysis.", "inputs": "Silhouette; Optical flow; GEI-subtracted sequences", "backbone": "Multi-stream network + AGF", "loss": "", "datasets": "Scoliosis1K", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/WACV2026/html/Vishwakarma_ScoliGaitX_A_Deep_Multi-Modal_Fusion_Network_for_Scoliosis_Assessment_via_WACV_2026_paper.html", "bibtex": "@inproceedings{vishwakarma2026scoligaitx,\n title = {ScoliGaitX: A Deep Multi-Modal Fusion Network for Scoliosis Assessment via Gait Video Analysis},\n author = {Kaushik Vishwakarma and Aditya Nigam},\n year = {2026},\n booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},\n url = {https://openaccess.thecvf.com/content/WACV2026/html/Vishwakarma_ScoliGaitX_A_Deep_Multi-Modal_Fusion_Network_for_Scoliosis_Assessment_via_WACV_2026_paper.html}\n}", "scope": "Related research", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Clinical classification, not biometric identity recognition; reported accuracy is not a Rank-1 identity metric.", "sections": [ "Challenges and Outlooks / Gait Representation Learning for Healthcare" ], "abstract": "ScoliGaitX studies scoliosis assessment from gait videos using silhouettes, optical flow and sequences with the average gait template removed. An alignment and gating module combines their complementary appearance and motion cues. Evaluation on Scoliosis1K investigates classification performance and the contribution of each stream to non-invasive screening research.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://openaccess.thecvf.com/content/WACV2026/html/Vishwakarma_ScoliGaitX_A_Deep_Multi-Modal_Fusion_Network_for_Scoliosis_Assessment_via_WACV_2026_paper.html", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "asad2026tkan-arxiv-2026", "citationKey": "asad2026tkan", "archiveKey": "asad2026tkan", "method": "CNN + TKAN", "title": "Gait Recognition with Temporal Kolmogorov-Arnold Networks", "authors": "Mohammed Asad and Dinesh Kumar Vishwakarma", "year": 2026, "venue": "arXiv", "motivation": "Compact recognizers still need to capture gait cycles and long-range timing.", "idea": "Combine spatial extraction with two levels of temporal memory.", "techniques": "Convolutional features; temporal Kolmogorov-Arnold network; hierarchical memory.", "inputs": "Silhouette", "backbone": "CNN + TKAN", "loss": "", "datasets": "CASIA-B", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://arxiv.org/abs/2604.09990", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Single-benchmark preprint; generalization and efficiency claims need protocol-matched validation.", "sections": [ "IV.B" ], "abstract": "This paper applies Temporal Kolmogorov-Arnold Networks to silhouette-based recognition. Learnable one-dimensional functions replace fixed edge weights, while short- and long-term memory pathways model gait cycles and broader context. A CNN supplies spatial features. CASIA-B experiments investigate recognition under the paper's stated protocol using a compact temporal backbone.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2604.09990", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "gaitprotector2026-arxiv-2026", "citationKey": "duan2026gaitprotector", "archiveKey": "gaitprotector2026", "method": "GaitProtector", "title": "GaitProtector: Impersonation-Driven Gait De-Identification via Training-Free Diffusion Latent Optimization", "authors": "Duan, Huiran and Zhou, Qian and Guo, Zhongliang and Dong, Junhao and Li, Yuqi and Zhao, Guoying and Tian, Yingli", "year": 2026, "venue": "arXiv", "motivation": "Visual anonymization may fail to conceal gait identity.", "idea": "Optimize generated walking sequences to suppress identity evidence.", "techniques": "Diffusion latent optimization; gait de-identification; utility preservation.", "inputs": "Gait imagery", "backbone": "Diffusion model", "loss": "", "datasets": "CASIA-B", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://arxiv.org/abs/2605.12431", "bibtex": "@inproceedings{duan2026gaitprotector,\n title = {GaitProtector: Impersonation-Driven Gait De-Identification via Training-Free Diffusion Latent Optimization},\n author = {Huiran Duan and Qian Zhou and Zhongliang Guo and Junhao Dong and Yuqi Li and Guoying Zhao and Yingli Tian},\n year = {2026},\n booktitle = {IEEE International Conference on Automatic Face and Gesture Recognition},\n url = {https://arxiv.org/abs/2605.12431}\n}", "scope": "Related research", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Privacy mechanism paper; do not count as a new recognition method or combine attack success with recognition accuracy.", "sections": [ "Background / Broader Applications", "Privacy Concerns and Potential Solutions" ], "authorsSource": "Public scholarly metadata", "abstract": "GaitProtector investigates gait de-identification through optimization in a diffusion model's latent space. The method targets an alternative identity without retraining the generator, seeking to alter identifying motion cues while maintaining usable video structure. The paper evaluates privacy protection alongside visual and behavioral utility.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2605.12431", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "gabet2026health-arxiv-2026", "citationKey": "gabet2026health", "archiveKey": "gabet2026health", "method": "Health gait foundation model", "title": "A Gait Foundation Model Predicts Multi-System Health Phenotypes from 3D Skeletal Motion", "authors": "Adam Gabet and Sarah Kohn and Guy Lutsker and Shira Gelman and Anastasia Godneva and Gil Sasson and Arad Zulti and David Krongauz and Rotem Shaulitch and Assaf Rotem and Ohad Doron and Yuval Brodsky and Adina Weinberger and Eran Segal", "year": 2026, "venue": "arXiv", "motivation": "Gait representations could support several health phenotypes beyond a single task.", "idea": "Transfer a skeletal foundation representation to health-related prediction.", "techniques": "Skeleton-based representation learning; transfer across health phenotypes.", "inputs": "3D skeletal motion", "backbone": "", "loss": "", "datasets": "3,414 adults", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://arxiv.org/abs/2603.25283", "bibtex": "@misc{gabet2026health,\n title = {A Gait Foundation Model Predicts Multi-System Health Phenotypes from 3D Skeletal Motion},\n author = {Adam Gabet and Sarah Kohn and Guy Lutsker and Shira Gelman and Anastasia Godneva and Gil Sasson and Arad Zulti and David Krongauz and Rotem Shaulitch and Assaf Rotem and Ohad Doron and Yuval Brodsky and Adina Weinberger and Eran Segal},\n year = {2026},\n url = {https://arxiv.org/abs/2603.25283},\n eprint = {2603.25283},\n archiveprefix = {arXiv}\n}", "scope": "Related research", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Observational preprint; predictive association is not clinical deployment validation; no identity accuracy comparison.", "sections": [ "Challenges and Outlooks / Gait Representation Learning for Healthcare" ], "abstract": "The authors train a representation model on 3D skeletal recordings from 3,414 adults and investigate associations with a broad set of health measurements. Learned embeddings are compared with engineered features and assessed after adjustment for demographic and body-composition variables. Anatomical ablations explore which body regions support different prediction targets.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2603.25283", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-su2026localscore", "citationKey": "su2026localscore", "archiveKey": "su2026localscore", "method": "LocalScore", "title": "LocalScore: Local Density-Aware Similarity Scoring for Biometrics", "authors": "Yiyang Su and Minchul Kim and Jie Zhu and Christopher Perry and Feng Liu and Anil Jain and Xiaoming Liu", "year": 2026, "venue": "arXiv", "motivation": "Global template scores ignore local gallery density in open-set recognition.", "idea": "Adjust similarity using neighboring gallery features.", "techniques": "K-nearest-neighbor density estimation; architecture-independent score normalization.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://arxiv.org/abs/2602.01012", "bibtex": "@misc{su2026localscore,\n title = {LocalScore: Local Density-Aware Similarity Scoring for Biometrics},\n author = {Yiyang Su and Minchul Kim and Jie Zhu and Christopher Perry and Feng Liu and Anil Jain and Xiaoming Liu},\n year = {2026},\n url = {https://arxiv.org/abs/2602.01012},\n eprint = {2602.01012},\n archiveprefix = {arXiv}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Advances in Methods / Gait Metric Designs" ], "abstract": "LocalScore adjusts biometric similarities using the local density of gallery features rather than representing each identity through a single global template. Its nearest-neighbor-based normalization is independent of the recognition architecture and training loss. Experiments and analysis investigate gains in verification and open-set identification, especially with multiple gallery observations.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2602.01012", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "duan2026mhrgait-arxiv-2026", "citationKey": "duan2026mhrgait", "archiveKey": "duan2026mhrgait", "method": "MHRGait / MHRGait++", "title": "MHRGait: Gait Recognition from Momentum Human Rig Pose", "authors": "Huiran Duan and Qian Zhou and Xianda Guo and Hua Zou and Guoying Zhao and Zhongyuan Wang and Yingli Tian", "year": 2026, "venue": "arXiv", "motivation": "Geometric appearance alone incompletely describes individual articulation.", "idea": "Combine compact human-rig controls with complementary geometry.", "techniques": "Rig-space modeling; joint-control parameters; modality-balanced distance fusion.", "inputs": "MHR pose; Silhouette for ++", "backbone": "Anatomical and temporal modeling", "loss": "", "datasets": "CCPG; SUSTech1K; other two not extracted", "codeAvailable": "Yes", "codeUrl": "https://github.com/duanhuiran/MHRGait", "url": "https://arxiv.org/abs/2607.29083", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "184 body/hand parameters per frame; variants count as one paper. No conference acceptance verified.", "sections": [ "IV.A", "IV.C", "IV.E" ], "abstract": "MHRGait represents movement through articulated controls estimated from monocular video. Anatomically grouped body and hand parameters are modeled across space and time to form compact identity descriptors. MHRGait++ combines these with silhouettes using balanced distance fusion. Four benchmark evaluations investigate accuracy, transfer and computational cost relative to geometric representations.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2607.29083", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "wang2026rrsgait-arxiv-2026", "citationKey": "wang2026rrsgait", "archiveKey": "wang2026rrsgait", "method": "RRS-Gait", "title": "Learning Geometric Invariance for Gait Recognition", "authors": "Zengbin Wang and Junjie Li and Saihui Hou and Xu Liu and Chunshui Cao and Yongzhen Huang and Muyi Sun and Siye Wang and Man Zhang", "year": 2026, "venue": "arXiv", "motivation": "Reflection, rotation and scale changes alter observed gait geometry.", "idea": "Model geometric transformations explicitly to improve consistency.", "techniques": "Reflect-rotate-scale equivariance and invariance; geometric augmentation.", "inputs": "", "backbone": "Adaptive convolution kernels", "loss": "", "datasets": "Gait3D; GREW; CCPG; SUSTech1K", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://arxiv.org/abs/2601.05604", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "2026 literature update", "evidence": "Primary abstract / metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "notes": "Detailed loss, complete evaluation protocols and code execution not audited.", "sections": [ "IV.B", "IV.D" ], "abstract": "RRS-Gait approximates some changes in walking observations through reflection, rotation and scaling. Transformation-aware convolutions seek equivariant intermediate features, and pooling converts them into invariant descriptors. Experiments on Gait3D, GREW, CCPG and SUSTech1K examine whether explicit geometric structure complements data-driven learning of identity under covariate changes.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2601.05604", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "wang2025ra-aaai-2025", "citationKey": "wang2025ra", "archiveKey": "wang2025ra", "method": "CLIP-GAR", "title": "RA-GAR: A Richly Annotated Benchmark for Gait Attribute Recognition", "authors": "Wang, Chenye and Hou, Saihui and Li, Aoqi and Cai, Qingyuan and Huang, Yongzhen", "year": 2025, "venue": "AAAI", "motivation": "Improve gait attribute recognition robustness", "idea": "Two-stage CLIP-guided fusion", "techniques": "Gait-text multimodal contrastive pretraining", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet, Contrastive", "datasets": "RA-GAR", "codeAvailable": "Yes", "codeUrl": "", "url": "", "bibtex": "@inproceedings{wang2025ra,\n title = {RA-GAR: A Richly Annotated Benchmark for Gait Attribute Recognition},\n author = {Wang, Chenye and Hou, Saihui and Li, Aoqi and Cai, Qingyuan and Huang, Yongzhen},\n year = {2025},\n booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},\n volume = {39},\n number = {7},\n pages = {7591--7599}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 9 } ], "summaryStatus": "Author collection", "sections": [ "Challenges and Outlooks / Real-World Gait Recognition" ], "abstract": "Gait attracts growing interest from researchers due to its advantages as a non-invasive and non-cooperative biometric feature. Current gait-based attribute recognition methods primarily focus on estimating attributes such as gender, age, and emotions. However, there is insufficient attention to diverse gait attributes in various covariate scenarios. In this paper, we design and collect a Richly Annotated benchmark for 15 gait attributes, named RA-GAR, comprising data from 533 individuals with over 120,000 sequences. To our knowledge, RA-GAR represents the largest and most diverse benchmark of gait attributes currently available. Furthermore, to fully leverage the semantic information and enhance attribute-specific local perception, we propose a two-stage CLIP-based method for Gait Attribute Recognition, named CLIP-GAR. Experiments on the RA-GAR and MA-Gait datasets demonstrate the effectiveness of CLIP-GAR, showing significant improvements in mean accuracy and F1 score. Datasets — https://github.com/BNU-IVC/RA-GAR", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "jin2025exploring-aaai-2025", "citationKey": "jin2025exploring", "archiveKey": "jin2025exploring", "method": "MultiGait++", "title": "Exploring more from multiple gait modalities for human identification", "authors": "Jin, Dongyang and Fan, Chao and Chen, Weihua and Yu, Shiqi", "year": 2025, "venue": "AAAI", "motivation": "Unify diverse gait modalities", "idea": "C2Fusion-based MultiGait++", "techniques": "Multimodal gait fusion network", "inputs": "Silhouette, Parsing, Optical", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "GREW, Gait3D, CCPG, SUSTech1K", "codeAvailable": "OpenGait", "codeUrl": "", "url": "https://ojs.aaai.org/index.php/AAAI/article/view/32432", "bibtex": "@inproceedings{jin2025exploring,\n title = {Exploring more from multiple gait modalities for human identification},\n author = {Jin, Dongyang and Fan, Chao and Chen, Weihua and Yu, Shiqi},\n year = {2025},\n booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},\n volume = {39},\n number = {4},\n pages = {4120--4128}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 8 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Benchmarks / State-of-the-Art Performance", "Advances in Methods / Ideal Gait Modalities" ], "abstract": "The gait, as a kind of soft biometric characteristic, can reflect the distinct walking patterns of individuals at a distance, exhibiting a promising technique for unrestrained human identification. With largely excluding gait-unrelated cues hidden in RGB videos, the silhouette and skeleton, though visually compact, have acted as two of the most prevailing gait modalities for a long time. Recently, several attempts have been made to introduce more informative data forms like human parsing and optical flow images to capture gait characteristics, along with multi-branch architectures. However, due to the inconsistency within model designs and experiment settings, we argue that a comprehensive and fair comparative study among these popular gait modalities, involving the representational capacity and fusion strategy exploration, is still lacking. From the perspectives of fine vs. coarse-grained shape and whole vs. pixel-wise motion modeling, this work presents an in-depth investigation of three popular gait representations, i.e., silhouette, human parsing, and optical flow, with various fusion evaluations, and experimentally exposes their similarities and differences. Based on the obtained insights, we further develop a C 2Fusion strategy, consequently building our new framework MultiGait++. C2Fusion preserves commonalities while highlighting differences to enrich the learning of gait features. To verify our findings and conclusions, extensive experiments on Gait3D, GREW, CCPG, and SUSTech1K are conducted. Code — https://github.com/ShiqiYu/OpenGait", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "wang2025waveloss-aaai-2025", "citationKey": "wang2025waveloss", "archiveKey": "wang2025waveloss", "method": "WaveLoss", "title": "WaveLoss: An Adaptive Dynamic Loss for Deep Gait Recognition", "authors": "Wang, Zicheng and Wu, Qiuxia", "year": 2025, "venue": "AAAI", "motivation": "Improve loss design for gait", "idea": "adaptive part-level loss fusion", "techniques": "Dynamic norm-fusion loss function", "inputs": "Silhouette", "backbone": "CNN", "loss": "WaveLoss", "datasets": "CASIA-B, OU-MVLP, Gait3D", "codeAvailable": "No", "codeUrl": "", "url": "", "bibtex": "@inproceedings{wang2025waveloss,\n title = {WaveLoss: An Adaptive Dynamic Loss for Deep Gait Recognition},\n author = {Wang, Zicheng and Wu, Qiuxia},\n year = {2025},\n booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},\n volume = {39},\n number = {8},\n pages = {8259--8267}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 10 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Gait Metric Designs" ], "abstract": "Designing an appropriate loss function can enhance the discriminative power on gait recognition. However, previous research focuses on improving network structure and enriching input modalities but overlooks the loss functions. Although transferring loss functions from face recognition can address sample-level loss, additional design is needed for part-level loss. Therefore, we have designed a new loss function called Waveloss, aimed at adaptively and dynamically changing the preference for parts of different difficulties. First, the previous method treats the loss of different parts equally, which brings the problems of difficult convergence or susceptibility to noise interference, so we propose norm-fusion to adaptively learn samples of different difficulties. Additionally, since we find the exponential value represents preference for learning different samples, we introduce the Dynamic Learning Process, which dynamically adjusts the exponential value during iteration to focus on samples of varying difficulties at different training stages. Finally, as the changes of the exponential value leads to significant fluctuations in the gradient, we introduce the gradient truncation and normalization to avoid getting trapped in local optima and gradient vanishing or exploding by adaptively adjusting the gradient. Experimental results demonstrate that our proposed Waveloss achieves state-of-the-art performance on various gait recognition datasets and can improve the performance of different backbones as well.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "lang2025dpgait-acmmm-2025", "citationKey": "lang2025dpgait", "archiveKey": "lang2025dpgait", "method": "DPGait", "title": "Beyond Sparse Keypoints: Dense Pose Modeling for Robust Gait Recognition", "authors": "Lang, Wenpeng and Hou, Saihui and Huang, Yongzhen", "year": 2025, "venue": "ACM MM", "motivation": "Advance pose-based gait recognition", "idea": "Dense keypoints with divide-and-conquer modeling", "techniques": "Dense keypoint extraction, grouped encoding, attention integration", "inputs": "Skeleton", "backbone": "CNN, Attention", "loss": "CE, Triplet", "datasets": "CASIA-B, CCPG, SUSTech1K", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1145/3746027.3755685", "bibtex": "@inproceedings{lang2025dpgait,\n title = {Beyond Sparse Keypoints: Dense Pose Modeling for Robust Gait Recognition},\n author = {Lang, Wenpeng and Hou, Saihui and Huang, Yongzhen},\n year = {2025},\n booktitle = {Proceedings of the 33nd ACM International Conference on Multimedia}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 84 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait recognition has emerged as a promising biometric technology due to its ability to operate at a distance without subject cooperation. While pose-based methods offer advantages over appearance-based approaches in robustness and interpretability, their performance has been limited by the sparse keypoint representations of current pose estimation frameworks. We identify two critical limitations: (1) incomplete motion representation due to insufficient keypoints for dynamic body parts, and (2) lack of shape information from minimal skeleton points. This paper presents DPGait, a novel framework that addresses these challenges through innovations in both upstream processing and downstream modeling. First, we enhance pose estimation by extending the standard COCO keypoint format with additional motion-sensitive points and shape-descriptive keypoints inspired by human mesh estimation. Second, we propose a divide-and-conquer modeling strategy that processes dense keypoints through group convolution with cross-group attention, coupled with multi-granularity supervision for improved training. Our comprehensive experiments demonstrate state-of-the-art performance in pose-based gait recognition, achieving 85.8% rank-1 accuracy on SUSTech1K—surpassing leading silhouette-based methods for the first time. The results validate that dense pose representation combined with our novel modeling approach significantly advances the field of gait recognition.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1145/3746027.3755685", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "meng2025mirrorgait-acmmm-2025", "citationKey": "meng2025mirrorgait", "archiveKey": "meng2025mirrorgait", "method": "MirrorGait", "title": "Seeing from Magic Mirror: Contrastive Learning from Reconstruction for Pose-based Gait Recognition", "authors": "Meng, Shibei and Hou, Saihui and Fu, Yang and Hu, Xuecai and Huang, Junzhou and Huang, Yongzhen", "year": 2025, "venue": "ACM MM", "motivation": "Self-supervised pose gait recognition", "idea": "3D-aware contrastive learning", "techniques": "Dataset curation, viewpoint synthesis, spatio-temporal modeling", "inputs": "Skeleton", "backbone": "Transformer", "loss": "Contrastive", "datasets": "OU-MVLP, GREW, Gait3D, StreamGait", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1145/3746027.3755066", "bibtex": "@inproceedings{meng2025mirrorgait,\n title = {Seeing from Magic Mirror: Contrastive Learning from Reconstruction for Pose-based Gait Recognition},\n author = {Meng, Shibei and Hou, Saihui and Fu, Yang and Hu, Xuecai and Huang, Junzhou and Huang, Yongzhen},\n year = {2025},\n booktitle = {Proceedings of the 33nd ACM International Conference on Multimedia}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 86 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "While recent advancements in supervised gait recognition have yielded promising results, these approaches rely heavily on annotated walking data, limiting their generalizability to complex environments. This paper presents a self-supervised gait recognition framework using human poses as input to address this challenge, focusing on high-quality pretrained data and self-supervised learning strategies. We first introduce StreamGait, a large-scale, unlabelled dataset that captures in-the-wild distributions of walking sequences. This dataset is curated from Internet livestreams across diverse geographic and environmental scenarios, reflecting variations in real-world camera angles, weather, and pedestrian behavior. Our framework, MirrorGait, conducts self-supervised learning by integration with 2D-to-3D pose reconstruction to synthesize multiview perspectives for effective 3D-aware contrastive learning. With specific designs of temporal position embedding and gait partition head on a Transformer backbone, the encoder can readily adapt to the periodic and fine-grained nature of gait. Extensive experiments on three widely used gait datasets, Gait3D, GREW, and OUMVLP- Pose, demonstrate that our method, with minimal fine-tuning on the pretrained model, achieves state-of-the-art performance among pose-based gait recognition approaches. The dataset, code, and models are available at https://github.com/BNU-IVC/StreamGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1145/3746027.3755066", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "jin2025denoising-cvpr-2025", "citationKey": "jin2025denoising", "archiveKey": "jin2025denoising", "method": "DenoisingGait", "title": "On Denoising Walking Videos for Gait Recognition", "authors": "Jin, Dongyang and Fan, Chao and Ma, Jingzhe and Zhou, Jingkai and Chen, Weihua and Yu, Shiqi", "year": 2025, "venue": "CVPR", "motivation": "Walking videos contain appearance and geometric noise unrelated to identity.", "idea": "Recover gait-relevant information through video denoising.", "techniques": "Diffusion-based denoising; geometric constraints; feature matching.", "inputs": "Silhouette, RGB", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, CCPG, SUSTech1K", "codeAvailable": "OpenGait", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2025/html/Jin_On_Denoising_Walking_Videos_for_Gait_Recognition_CVPR_2025_paper.html", "bibtex": "@inproceedings{jin2025denoising,\n title = {On Denoising Walking Videos for Gait Recognition},\n author = {Jin, Dongyang and Fan, Chao and Ma, Jingzhe and Zhou, Jingkai and Chen, Weihua and Yu, Shiqi},\n year = {2025},\n booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference},\n pages = {12347--12357}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 31 }, { "source": "Manuscript Table I" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution", "Advances in Benchmarks / State-of-the-Art Performance", "Advances in Methods / Gait Learning Frameworks", "Challenges and Outlooks / Gait Foundation Models" ], "abstract": "To capture individual gait patterns, excluding identityirrelevant cues in walking videos, such as clothing texture and color, remains a persistent challenge for vision-based gait recognition. Traditional silhouette- and pose-based methods, though theoretically effective at removing such distractions, often fall short of high accuracy due to their sparse and less informative inputs. Emerging end-to-end methods address this by directly denoising RGB videos using human priors. Building on this trend, we propose DenoisingGait, a novel gait denoising method. Inspired by the philosophy that “what I cannot create, I do not understand”, we turn to generative diffusion models, uncovering how they partially filter out irrelevant factors for gait understanding. Additionally, we introduce a geometry-driven Feature Matching module, which, combined with background removal via human silhouettes, condenses the multichannel diffusion features at each foreground pixel into a two-channel direction vector. Specifically, the proposed within- and cross-frame matching respectively capture the local vectorized structures of gait appearance and motion, producing a novel flow-like gait representation termed Gait Feature Field, which further reduces residual noise in diffusion features. Experiments on the CCPG, CASIA-B*, and SUSTech1K datasets demonstrate that DenoisingGait achieves a new SoTA performance in most cases for both within- and cross-domain evaluations. Code is available at https://github.com/ShiqiYu/OpenGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2025/html/Jin_On_Denoising_Walking_Videos_for_Gait_Recognition_CVPR_2025_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "yang2025bridging-cvpr-2025", "citationKey": "yang2025bridging", "archiveKey": "yang2025bridging", "method": "GaitLLM", "title": "Bridging gait recognition and large language models sequence modeling", "authors": "Yang, Shaopeng and Wang, Jilong and Hou, Saihui and Liu, Xu and Cao, Chunshui and Wang, Liang and Huang, Yongzhen", "year": 2025, "venue": "CVPR", "motivation": "Gait sequence modeling can benefit from pretrained sequence knowledge.", "idea": "Translate gait features into tokens compatible with a language model.", "techniques": "Gait tokenization; pretrained LLM sequence modeling; feature projection.", "inputs": "Silhouette", "backbone": "CNN, Transformer", "loss": "CE, Triplet", "datasets": "GREW, Gait3D, CCPG, SUSTech1K", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2025/html/Yang_Bridging_Gait_Recognition_and_Large_Language_Models_Sequence_Modeling_CVPR_2025_paper.html", "bibtex": "@inproceedings{yang2025bridging,\n title = {Bridging gait recognition and large language models sequence modeling},\n author = {Yang, Shaopeng and Wang, Jilong and Hou, Saihui and Liu, Xu and Cao, Chunshui and Wang, Liang and Huang, Yongzhen},\n year = {2025},\n booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference},\n pages = {3460--3469}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 33 }, { "source": "Manuscript Table I" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution", "Advances in Methods / Gait Learning Frameworks" ], "abstract": "Gait sequences exhibit sequential structures and contextual relationships similar to those in natural language, where each element—whether a word or a gait step—is connected to its predecessors and successors. This similarity enables the transformation of gait sequences into “texts” containing identity-related information. Large Language Models (LLMs), designed to understand and generate sequential data, can thus be utilized for gait sequence modeling to enhance gait recognition performance. Leveraging these insights, we make a pioneering effort to apply LLMs to gait recognition, which we refer to as GaitLLM. Specifically, we propose the Gait-to-Language (G2L) module, which converts gait sequences into a textual format suitable for LLMs, and the Language-to-Gait (L2G) module, which maps the LLM’s output back to the gait feature space, thereby bridging the gap between LLM outputs and gait recognition. Notably, GaitLLM leverages the powerful modeling capabilities of LLMs without relying on complex architectural designs, improving gait recognition performance with only a small number of trainable parameters. Our method achieves state-of-the-art results on four popular gait datasets—SUSTech1K, CCPG, Gait3D, and GREW—demonstrating the effectiveness of applying LLMs in this domain. This work highlights the potential of LLMs to significantly enhance gait recognition, paving the way for future research and practical applications.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2025/html/Yang_Bridging_Gait_Recognition_and_Large_Language_Models_Sequence_Modeling_CVPR_2025_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "shen2025lidargait-cvpr-2025", "citationKey": "shen2025lidargait++", "archiveKey": "shen2025lidargait++", "method": "LidarGait++", "title": "LidarGait++: Learning Local Features and Size Awareness from LiDAR Point Clouds for 3D Gait Recognition", "authors": "Shen, Chuanfu and Wang, Rui and Duan, Lixin and Yu, Shiqi", "year": 2025, "venue": "CVPR", "motivation": "Point-cloud projection can lose local geometry and individual body-size cues.", "idea": "Jointly encode local structure and size-aware identity information.", "techniques": "Multi-scale point features; local aggregation; size encoding.", "inputs": "Point", "backbone": "MLP", "loss": "CE, Triplet", "datasets": "SUSTech1K, FreeGait", "codeAvailable": "OpenGait", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2025/html/Shen_LidarGait_Learning_Local_Features_and_Size_Awareness_from_LiDAR_Point_CVPR_2025_paper.html", "bibtex": "@inproceedings{shen2025lidargait++,\n title = {LidarGait++: Learning Local Features and Size Awareness from LiDAR Point Clouds for 3D Gait Recognition},\n author = {Shen, Chuanfu and Wang, Rui and Duan, Lixin and Yu, Shiqi},\n year = {2025},\n booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference},\n pages = {6627--6636}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 32 }, { "source": "Manuscript Table I" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution", "Advances in Benchmarks / State-of-the-Art Performance", "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Handling Real-World Noise" ], "abstract": "Point clouds have gained growing interest in gait recognition. However, current methods, which typically convert point clouds into 3D voxels, often fail to extract essential gait-specific features. In this paper, we explore gait recognition within 3D point clouds from the perspectives of architectural designs and gait representation modeling. We indicate the significance of local and body size features in 3D gait recognition and introduce LidarGait++, a novel framework combining advanced local representation learning techniques with a novel size-aware learning mechanism. Specifically, LidarGait++ utilizes Set Abstraction (SA) layer and Pyramid Point Pooling (P3) layer for learning locally fine-grained gait representations from 3D point clouds directly. Both the SA and P3 layers can be further enhanced with size-aware learning to make the model aware of the actual size of the subjects. In the end, LidarGait++ not only outperforms current state-of-the-art methods, but it also consistently demonstrates robust performance and great generalizability on two benchmarks. Our extensive experiments validate the effectiveness of size and local features in 3D gait recognition.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2025/html/Shen_LidarGait_Learning_Local_Features_and_Size_Awareness_from_LiDAR_Point_CVPR_2025_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-kim2025sapiensid", "citationKey": "kim2025sapiensid", "archiveKey": "kim2025sapiensid", "method": "SapiensID", "title": "SapiensID: Foundation for Human Recognition", "authors": "Kim, Minchul and Ye, Dingqiang and Su, Yiyang and Liu, Feng and Liu, Xiaoming", "year": 2025, "venue": "CVPR", "motivation": "Separate face and body models struggle with changing pose and subject scale.", "idea": "Learn a unified human recognition model with adaptive tokenization.", "techniques": "Retina Patch; masked recognition modeling; semantic attention; WebBody4M.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://arxiv.org/abs/2504.04708", "bibtex": "@inproceedings{kim2025sapiensid,\n title = {SapiensID: Foundation for Human Recognition},\n author = {Kim, Minchul and Ye, Dingqiang and Su, Yiyang and Liu, Feng and Liu, Xiaoming},\n year = {2025},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n pages = {13937--13947},\n url = {https://arxiv.org/abs/2504.04708}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Challenges and Outlooks / Real-World Gait Recognition" ], "abstract": "SapiensID unifies face and body recognition within a common model. Adaptive patch generation handles changes in subject scale, masked recognition training accommodates variable token counts, and semantic attention aggregates body regions. WebBody4M supports pretraining; evaluations cover image-based identification and challenging changes in pose, visibility and scale.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2504.04708", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-jager2025expanding", "citationKey": "jager2025expanding", "archiveKey": "jager2025expanding", "method": "", "title": "Expanding on the briar dataset: A comprehensive whole body biometric recognition resource at extreme distances and real-world scenarios (collections 1-4)", "authors": "Jager, Gavin and Cornett, David and Glenn, Gavin and Aykac, Deniz and Johnson, Christi and Zhang, Robert and Shivers, Ryan and Bolme, David and Davies, Laura and Dolvin, Scott and others", "year": 2025, "venue": "FG", "motivation": "Whole-body recognition needs diverse long-range and real-world evaluation data.", "idea": "Extend BRIAR through additional collection campaigns.", "techniques": "Multimodal biometric acquisition; extreme-range protocols; collections 1–4.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@inproceedings{jager2025expanding,\n title = {Expanding on the briar dataset: A comprehensive whole body biometric recognition resource at extreme distances and real-world scenarios (collections 1-4)},\n author = {Jager, Gavin and Cornett, David and Glenn, Gavin and Aykac, Deniz and Johnson, Christi and Zhang, Robert and Shivers, Ryan and Bolme, David and Davies, Laura and Dolvin, Scott and others},\n year = {2025},\n booktitle = {2025 IEEE 19th International Conference on Automatic Face and Gesture Recognition (FG)},\n pages = {1--9}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Introduction / Motivation", "Background / Broader Applications", "Advances in Benchmarks / Popular Datasets" ], "abstract": "This paper extends the BRIAR resource for recognizing people at extreme distances and from elevated or aerial cameras. It documents collections 1–4, including acquisition, curation and annotation. The dataset targets whole-body biometric research under operational conditions that differ substantially from conventional close-range recognition benchmarks.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2501.14070", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "habib2025cargait-iccv-2025", "citationKey": "habib2025cargait", "archiveKey": "habib2025cargait", "method": "CarGait", "title": "CarGait: Cross-Attention based Re-ranking for Gait recognition", "authors": "Habib, Gavriel and Barzilay, Noa and Shimshi, Or and Ben-Ari, Rami and Darshan, Nir", "year": 2025, "venue": "ICCV", "motivation": "Improve top-rank gait accuracy", "idea": "Fine-grained re-ranking with attention", "techniques": "Body-part matching, metric learning", "inputs": "Silhouette", "backbone": "Attention, MLP", "loss": "CE, Ranking Loss", "datasets": "OU-MVLP, GREW, Gait3D", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/ICCV2025/html/Habib_CarGait_Cross-Attention_based_Re-ranking_for_Gait_recognition_ICCV_2025_paper.html", "bibtex": "@inproceedings{habib2025cargait,\n title = {CarGait: Cross-Attention based Re-ranking for Gait recognition},\n author = {Habib, Gavriel and Barzilay, Noa and Shimshi, Or and Ben-Ari, Rami and Darshan, Nir},\n year = {2025},\n booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision},\n url = {https://openaccess.thecvf.com/content/ICCV2025/html/Habib_CarGait_Cross-Attention_based_Re-ranking_for_Gait_recognition_ICCV_2025_paper.html}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 54 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Gait Metric Designs" ], "abstract": "Gait recognition is a computer vision task that identifies individuals based on their walking patterns. Gait recognition performance is commonly evaluated by ranking a gallery of candidates and measuring the accuracy at the top Rank- K. Existing models are typically single-staged, i.e. searching for the probe’s nearest neighbors in a gallery using a single global feature representation. Although these models typically excel at retrieving the correct identity within the top- K predictions, they struggle when hard negatives appear in the top short-list, leading to relatively low performance at the highest ranks ( e.g., Rank-1). In this paper, we introduce CarGait, a Cross-Attention Re-ranking method for gait recognition, that involves re-ordering the top-K list leveraging the fine-grained correlations between pairs of gait sequences through cross-attention between gait strips. This re-ranking scheme can be adapted to existing single-stage models to enhance their final results. We demonstrate the capabilities of CarGait by extensive experiments on three common gait datasets, Gait3D, GREW, and OU-MVLP , and seven different gait models, showing consistent improvements in Rank-1,5 accuracy, superior results over existing re-ranking methods, and strong baselines.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/ICCV2025/html/Habib_CarGait_Cross-Attention_based_Re-ranking_for_Gait_recognition_ICCV_2025_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "wang2025gaitx-acmmm-2025", "citationKey": "wang2025gaitx", "archiveKey": "wang2025gaitx", "method": "Gait-X", "title": "Gait-X: Exploring X modality for Generalized Gait Recognition", "authors": "Wang, Zengbin and Hou, Saihui and Li, Junjie and Liu, Xu and Cao, Chunshui and Huang, Yongzhen and Wang, Siye and Zhang, Man", "year": 2025, "venue": "ICCV", "motivation": "Generalized gait-specific modality exploration", "idea": "Frequency-based stable modality customization", "techniques": "Frequency decomposition, noise suppression, guided training", "inputs": "RGB", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, CCPG, SUSTech1K", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/ICCV2025/html/Wang_Gait-X_Exploring_X_modality_for_Generalized_Gait_Recognition_ICCV_2025_paper.html", "bibtex": "@inproceedings{wang2025gaitx,\n title = {Gait-X: Exploring X modality for Generalized Gait Recognition},\n author = {Wang, Zengbin and Hou, Saihui and Li, Junjie and Liu, Xu and Cao, Chunshui and Huang, Yongzhen and Wang, Siye and Zhang, Man},\n year = {2025},\n booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision},\n pages = {13259--13269},\n url = {https://openaccess.thecvf.com/content/ICCV2025/html/Wang_Gait-X_Exploring_X_modality_for_Generalized_Gait_Recognition_ICCV_2025_paper.html}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 85 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Ideal Gait Modalities" ], "abstract": "Modality exploration has been repeatedly mentioned in gait recognition, evolving from silhouette to parsing, mesh, point clouds, etc. These latest modalities agree that silhouette is less affected by background and clothing noises, but argue it loses too much valuable discriminative information. They seek to retain the strengths of silhouette while extracting more semantic or structural information through upstream estimation for better recognition. We agree with this principle but argue that these upstream estimations are usually unstable and the resulted modalities rely on pre-defined design. Moreover, the crucial aspect of modality generalization remains underexplored. To address this, we propose Gait-X to explore how to flexibly and stably develop a gaitspecific generalized X modality from a frequency perspective. Specifically, 1) We replace upstream estimation with stable frequency decomposition and conduct a comprehensive analysis of how different frequencies impact modality and within-/cross-domain performance; 2) To enable flexible modality customization and mitigate the influence of noise and domain variations, we propose to remove irrelevant low-frequency noise and suppress high-frequency domain-specific information to form our X modality; 3) To further improve model generalization, we expand the representation across multiple frequencies to guide the model in balancing whole frequencies for enhanced generalization. Extensive experiments on CCPG, SUSTech1K, and CASIA- B datasets show superior generalization.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/ICCV2025/html/Wang_Gait-X_Exploring_X_modality_for_Generalized_Gait_Recognition_ICCV_2025_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "huang2025origins-iccv-2025", "citationKey": "huang2025origins", "archiveKey": "huang2025origins", "method": "Origins-S", "title": "Learning A Unified Template for Gait Recognition", "authors": "Huang, Panjian and Hou, Saihui and Huang, Junzhou and Huang, Yongzhen", "year": 2025, "venue": "ICCV", "motivation": "Unify diverse gait representations", "idea": "Generate unified semantic gait space", "techniques": "Generative Template Diffusion, semantic alignment", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet, MSE Loss", "datasets": "CASIA-B, GREW, Gait3D, CCPG, SUSTech1K, CCGR-MINI", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/ICCV2025/html/Huang_Learning_A_Unified_Template_for_Gait_Recognition_ICCV_2025_paper.html", "bibtex": "@inproceedings{huang2025origins,\n title = {Learning A Unified Template for Gait Recognition},\n author = {Huang, Panjian and Hou, Saihui and Huang, Junzhou and Huang, Yongzhen},\n year = {2025},\n booktitle = {2025 IEEE/CVF International Conference on Computer Vision (ICCV)}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 55 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Benchmarks / State-of-the-Art Performance" ], "abstract": "“What I cannot create, I do not understand.” Human wisdom reveals that creation is one of the highest forms of learning. For example, Diffusion Models have demonstrated remarkable semantic structure and memory in image generation, understanding, and restoration, which intuitively benefits representation learning. However, current gait networks rarely embrace this perspective, relying primarily on learning by contrasting gait samples under varying complex conditions, leading to semantic inconsistency and uniformity issues. To address these issues, we propose Origins with generative capabilities whose underlying philosophy is that different entities are generated from a unified template , inherently regularizing gait representations within a consistent and diverse semantic space to capture accurate gait differences. Admittedly, learning this unified template is exceedingly challenging, as it requires the comprehensiveness of the template to encompass gait representations with various conditions. Inspired by Diffusion Models, Origins diffuses the unified template into timestep templates for gait generative learning, and meanwhile transfers the unified template for gait representation learning. Especially, gait generative and representation learning serve as a unified framework for end-to-end joint training. Extensive experiments on CASIA-B, CCPG, SUSTech1K, Gait3D, GREW and CCGR-MINI demonstrate that Origins performs unified generative and representation learning, achieving superior performance.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-zhu2025qme", "citationKey": "zhu2025qme", "archiveKey": "zhu2025qme", "method": "QME", "title": "A Quality-Guided Mixture of Score-Fusion Experts Framework for Human Recognition", "authors": "Zhu, Jie and Su, Yiyang and Kim, Minchul and Jain, Anil K. and Liu, Xiaoming", "year": 2025, "venue": "ICCV", "motivation": "Biometric scores differ in distribution and reliability across modalities.", "idea": "Condition learned score-fusion experts on estimated input quality.", "techniques": "Modality quality estimation; mixture of experts; score triplet loss.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://arxiv.org/abs/2508.00053", "bibtex": "@inproceedings{zhu2025qme,\n title = {A Quality-Guided Mixture of Score-Fusion Experts Framework for Human Recognition},\n author = {Zhu, Jie and Su, Yiyang and Kim, Minchul and Jain, Anil K. and Liu, Xiaoming},\n year = {2025},\n booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision},\n url = {https://arxiv.org/abs/2508.00053}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Challenges and Outlooks / Real-World Gait Recognition" ], "abstract": "Whole-body biometric recognition is a challenging multimodal task that integrates various biometric modalities, including face, gait, and body. This integration is essential for overcoming the limitations of unimodal systems. Traditionally, whole-body recognition involves deploying different models to process multiple modalities, achieving the final outcome by score-fusion (e.g., weighted averaging of similarity matrices from each model). However, these conventional methods may overlook the variations in score distributions of individual modalities, making it challenging to improve final performance. In this work, we present Quality-guided Mixture of score-fusion Experts (QME), a novel framework designed for improving whole-body biometric recognition performance through a learnable score-fusion strategy using a Mixture of Experts (MoE). We introduce a novel pseudo-quality loss for quality estimation with a modality-specific Quality Estimator (QE), and a score triplet loss to improve the metric performance. Extensive experiments on multiple whole-body biometric datasets demonstrate the effectiveness of our proposed approach, achieving state-of-the-art results across various metrics compared to baseline methods. Our method is effective for multimodal and multi-model, addressing key challenges such as model misalignment in the similarity score domain and variability in data quality. Code is available at the Project Link.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://arxiv.org/abs/2508.00053", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "zhou2025cgtgait-ijcb-2025", "citationKey": "Zhou2025CGTGait", "archiveKey": "Zhou2025CGTGait", "method": "CGTGait", "title": "CGTGait Collaborative Graph and Transformer for Gait Emotion Recognition", "authors": "Junjie Zhou and Haijun Xiong and Junhao Lu and Ziyu Lin and Bin Feng", "year": 2025, "venue": "IJCB", "motivation": "Long-range gait emotion modeling", "idea": "Cross-stream global temporal fusion", "techniques": "Graph-transformer dual-stream fusion", "inputs": "Silhouette", "backbone": "GCN, Transformer", "loss": "CE, Triplet", "datasets": "EMOGAIT, ELMD", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1109/IJCB65343.2025.11411435", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 65 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "CGTGait_Collaborative_Graph_and_Transformer_for_Gait_Emotion_Recognition.pdf", "abstract": "Skeleton-based gait emotion recognition has received significant attention due to its wide-ranging applications. However, existing methods primarily focus on extracting spatial and local temporal motion information, failing to capture long-range temporal representations. In this paper, we propose CGTGait, a novel framework that collaboratively integrates graph convolution and transformers to extract discriminative spatiotemporal features for gait emotion recognition. Specifically, CGTGait consists of multiple CGT blocks, where each block employs graph convolution to capture frame-level spatial topology and the transformer to model global temporal dependencies. Additionally, we introduce a Bidirectional Cross-Stream Fusion (BCSF) module to effectively aggregate posture and motion spatiotemporal features, facilitating the exchange of complementary information between the two streams. We evaluate our method on two widely used datasets, Emotion-Gait and ELMD, demonstrating that our CGTGait achieves state-ofthe-art or at least competitive performance while reducing computational complexity by approximately82.2% (only requiring 0.34G FLOPs) during testing. Code is available at https://github.com/githubzjj1/CGTGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/IJCB65343.2025.11411435", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "lu2025csca-ijcb-2025", "citationKey": "Lu2025CSCA", "archiveKey": "Lu2025CSCA", "method": "CSCA", "title": "Collaborative Spatial and Channel Attention for Structural Vibration-based Gait Recognition", "authors": "Junhao Lu and Haijun Xiong and Junjie Zhou and Ziyu Lin and Bin Feng", "year": 2025, "venue": "IJCB", "motivation": "Global vibration gait modeling", "idea": "Linear channel multi-scale spatial", "techniques": "Collaborative spatial-channel attention", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE", "datasets": "VIBEID", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1109/IJCB65343.2025.11410963", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 64 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Collaborative_Spatial_and_Channel_Attention_for_Structural_Vibration-based_Gait_Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Structural vibration-based gait recognition aims to identify pedestrians through their unique vibration patterns and has gained significant attention for its non-invasive nature. However, existing methods primarily rely on manually crafted features and convolutional neural networks (CNNs) to extract local details, without considering global gait features in vibration signals. To address this limitation, we propose CSCA, a novel framework designed to capture discriminative global gait features through collaborative spatial and channel attention. Specifically, CSCA consists of two key components: Mamba-inspired Linear Channel Attention (MLCA) and Strip Pooling-guided Spatial Attention (SPSA). MLCA integrates Mamba with linear attention to enhance discriminative attention in the channel dimension by capturing global gait features. Additionally, SPSA integrates axial strip pooling with multikernel depth-wise convolutions to extract multi-level spatial semantic features. Experimental results on the VIBEID dataset demonstrate that CSCA achieves state-of-the-art performance across diverse environmental conditions, with identity recognition accuracy exceeding 96%. Code is available at https://github.com/xiaomush/CSCA.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/IJCB65343.2025.11410963", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "li2025cvvnet-ijcb-2025", "citationKey": "Li2025CVVNet", "archiveKey": "Li2025CVVNet", "method": "CVVNet", "title": "CVVNet A Cross-Vertical-View Network for Gait Recognition", "authors": "Xiangru Li and Wei Song and Yingda Huang and Wei Meng and Le Chang and Hongyang Li", "year": 2025, "venue": "IJCB", "motivation": "Robust vertical view generalization", "idea": "Multi-frequency gated feature fusion", "techniques": "Cross-vertical-view frequency aggregation", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "Gait3D, DroneGait", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/ijcb65343.2025.11410871", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 62 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "CVVNet_A Cross-Vertical-View Network for Gait Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition enables contact-free, long-range person identification that is robust to clothing variations and non-cooperative scenarios. While existing methods perform well in controlled environments, they struggle with crossvertical view scenarios where elevation changes induce severe perspective distortions and self-occlusions. Our analysis reveals conventional CNN-based approaches relying on single-scale receptive fields fail to capture hierarchical frequency features, while standard self-attention mechanisms tend to over-smooth critical high-frequency edge patterns. To tackle this challenge, we propose Cross-Vertical- View Network(CVVNet), a frequency aggregation architecture specifically designed for robust cross-vertical-view gait recognition. CVVNet employs a Multi-Scale Attention Gated Aggregation (MSAGA) module which consists of a High-Low Frequency Extraction module (HLFE) and a Dynamic Gated Aggregation (DGA) mechanism. Specifically, HLFE adopts parallel multi-scale convolution/max-pooling path and self-attention path as high- and low-frequency mixers for effective multi-frequency feature extraction from input silhouettes, while DGA is introduced to adaptively adjust the fusion ratio of frequency and spatial features. The integration of HLFE and DGA enables CVVNet to effectively handle distortions from view changes, significantly improving the recognition robustness across different vertical views. Experimental results show that our CVVNet achieves state-of-the-art performance on both datasets, delivering an 8.6% improvement on DroneGait while retaining the leading position on Gait3D.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/ijcb65343.2025.11410871", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "cao2025dyuda-ijcb-2025", "citationKey": "Cao2025DyUDA", "archiveKey": "Cao2025DyUDA", "method": "DyUDA", "title": "Learning Dynamic Gait Regions Adaptive Part Weighting UDA for Cross-domain Gait Recognition", "authors": "Yicheng Cao and Hanqi Lyu and Min Zhang and Yan Sun", "year": 2025, "venue": "IJCB", "motivation": "Robust cross-domain gait adaptation", "idea": "Region weighting hard-sample segmentation", "techniques": "Dynamic part-weighted gait UDA", "inputs": "Silhouette", "backbone": "CNN", "loss": "Triplet", "datasets": "CASIA-B, OU-MVLP, Gait3D, GREW", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/IJCB65343.2025.11410992", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 63 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Learning_Dynamic_Gait_Regions_Adaptive_Part_Weighting_UDA_for_Cross-domain_Gait_Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition has emerged as a promising biometric identification technology due to its non-invasive nature and long-distance capture capabilities. However, its practical deployment faces challenges in cross-domain scenarios, where models trained on labeled source domain data struggle to generalize to unlabeled target domains with divergent distributions caused by variations in illumination, clothing, camera angles, and environmental factors. Existing Unsupervised Domain Adaptation (UDA) techniques often fail to address key issues such as hard sample mining and robustness to dynamic gait features. In this work, we propose DyUDA, a novel framework that effectively addresses these challenges in cross-domain gait recognition. First, an Adaptive Part Weighting (APW) module dynamically prioritizes motion-relevant body regions (e.g., head and feet), reducing the impact of appearance variation. Second, a Mix-sim Sampler categorizes samples into high-similarity (high-sim) and low-similarity (low-sim) groups based on cosine similarity, leveraging both to optimize pseudo-label quality and mitigate training instability. Third, a Temporal Segmentation approach splits gait sequences into subsegments, enriching feature diversity and improving robustness against viewpoint changes by expanding positive sample pairs. Extensive experiments demonstrate that our DyUDA framework achieves state-of-the-art performance on benchmark datasets, significantly outperforming existing UDA methods in cross-domain accuracy while maintaining stability under complex real-world conditions.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/IJCB65343.2025.11410992", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-hid2025", "citationKey": "hid2025", "archiveKey": "hid2025", "method": "", "title": "Human Identification at a Distance: Challenges, Methods and Results on the Competition HID 2025", "authors": "Ma, Jingzhe and Zhang, Meng and Yu, Jianlong and Liu, Kun and Xu, Zunxiao and Cheng, Xue and Zhou, Junjie and Wang, Yanfei and Li, Jiahang and Wang, Zepeng and Osamura, Kazuki and Liu, Rujie and Abe, Narishige and Wang, Jingjie and Zhang, Shunli and Xie, Haojun and Wu, Jiajun and Wu, Weiming and Kang, Wenxiong and Gao, Qingshuo and Xiong, Jiaming and Ben, Xianye and Chen, Lei and Song, Lichen and Cui, Junjian and Xiong, Haijun and Lu, Junhao and Feng, Bin and Liu, Mengyuan and Zhou, Ji and Zhao, Baoquan and Xu, Ke and Huang, Yongzhen and Wang, Liang and Marin-Jimenez, Manuel J. and Ahad, Md Atiqur Rahman and Yu, Shiqi", "year": 2025, "venue": "IJCB", "motivation": "Practical gait recognition requires shared evaluation under challenging observation conditions.", "idea": "Compare community submissions through the HID 2025 competition.", "techniques": "Common protocols; benchmark evaluation; analysis of participating systems.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/IJCB65343.2025.11411399", "bibtex": "@inproceedings{hid2025,\n title = {Human Identification at a Distance: Challenges, Methods and Results on the Competition HID 2025},\n author = {Ma, Jingzhe and Zhang, Meng and Yu, Jianlong and Liu, Kun and Xu, Zunxiao and Cheng, Xue and Zhou, Junjie and Wang, Yanfei and Li, Jiahang and Wang, Zepeng and Osamura, Kazuki and Liu, Rujie and Abe, Narishige and Wang, Jingjie and Zhang, Shunli and Xie, Haojun and Wu, Jiajun and Wu, Weiming and Kang, Wenxiong and Gao, Qingshuo and Xiong, Jiaming and Ben, Xianye and Chen, Lei and Song, Lichen and Cui, Junjian and Xiong, Haijun and Lu, Junhao and Feng, Bin and Liu, Mengyuan and Zhou, Ji and Zhao, Baoquan and Xu, Ke and Huang, Yongzhen and Wang, Liang and Marin-Jimenez, Manuel J. and Ahad, Md Atiqur Rahman and Yu, Shiqi},\n year = {2025},\n booktitle = {2025 IEEE International Joint Conference on Biometrics (IJCB)},\n doi = {10.1109/IJCB65343.2025.11411399}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Advances in Benchmarks / Competitions" ], "abstract": "The HID 2025 report reviews a gait-recognition competition using SUSTech-Competition without organizer-supplied training data. It describes evaluation splits, leading systems and cross-domain challenges involving clothing, carried objects and views. The strongest submission reached 94.2% accuracy on the competition protocol; the report also discusses technical trends across participating methods.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2602.07565", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "tian2025qgait-ijcb-2025", "citationKey": "tian2025qgait", "archiveKey": "tian2025qgait", "method": "QGait", "title": "QGait Toward Accurate Quantization for Gait Recognition", "authors": "Senmao Tian and Haoyu Gao and Gangyi Hong and Shuyun Wang and JingJie Wang and Xin Yu and Shunli Zhang", "year": 2025, "venue": "IJCB", "motivation": "Efficient edge gait deployment", "idea": "Two-stage soft distance calibration", "techniques": "Quantized gait recognition", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, Gait3D, GREW", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/IJCB65343.2025.11411264", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 60 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "QGait_Toward_Accurate_Quantization_for_Gait_Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Existing deep learning methods have made significant progress in gait recognition. Quantization can facilitate the application of gait models as a model-agnostic general compression technique. Typically, appearance-based models binarize inputs into silhouette sequences. However, mainstream quantization methods prioritize minimizing task loss over quantization error, which is detrimental to gait recognition with binarized inputs. To address this, we propose a differentiable soft quantizer, which better simulates the gradient of the round function during backpropagation. This enables the network to learn from subtle input perturbations. However, our theoretical analysis and empirical studies reveal that directly applying the soft quantizer can hinder network convergence. We addressed this issue by adopting a two-stage training strategy, introducing a soft quantizer during the fine-tuning phase. However, in the first stage of training, we observed a significant change in the output distribution of different samples in the feature space compared to the full-precision network. It is this change that led to a loss in performance. Based on this, we propose an Inter-class Distance-guided Calibration (IDC) strategy to preserve the relative distance between the embeddings of samples with different labels. Extensive experiments validate the effectiveness of our approach, demonstrating state-of-the-art accuracy across various settings and datasets. The code will be made publicly available.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/IJCB65343.2025.11411264", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "gupta2025mind-ijcb-2025", "citationKey": "gupta2025mind", "archiveKey": "gupta2025mind", "method": "RG-Gait", "title": "Mind the Gap Bridging Occlusion in Gait Recognition via Residual Gap Correction", "authors": "Ayush Gupta and Siyuan Huang and Rama Chellappa", "year": 2025, "venue": "IJCB", "motivation": "Occlusion without holistic forgetting", "idea": "Learn and gate residuals", "techniques": "Residual correction adaptive fusion", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "Gait3D, GREW, BRIAR", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1109/ijcb65343.2025.11411559", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 59 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Mind the Gap_Bridging Occlusion in Gait Recognition via Residual Gap Correction.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait is becoming popular as a method of person reidentification because of its ability to identify people at a distance. However, most current works in gait recognition do not address the practical problem of occlusions. Among those which do, some require paired tuples of occluded and holistic sequences, which are impractical to collect in the real world. Further, these approaches work on occlusions but fail to retain performance on holistic inputs. To address these challenges, we propose RG-Gait, a method for residual correction for occluded gait recognition with holistic retention. We model the problem as a residual learning task, conceptualizing the occluded gait signature as a residual deviation from the holistic gait representation. Our proposed network adaptively integrates the learned residual, significantly improving performance on occluded gait sequences without compromising the holistic recognition accuracy. We evaluate our approach on the challenging Gait3D, GREW and BRIAR datasets and show that learning the residual can be an effective technique to tackle occluded gait recognition with holistic retention. We release our code publicly at https://github.com/Ayush-00/rg-gait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/ijcb65343.2025.11411559", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "zhou2025tdgait-ijcb-2025", "citationKey": "Zhou2025TDGait", "archiveKey": "Zhou2025TDGait", "method": "TDGait", "title": "TDGait Modeling Temporal Dynamics for Gait Recognition", "authors": "Yuhao Zhou and Mingyang Wang and Xianjie Wang", "year": 2025, "venue": "IJCB", "motivation": "Capture gait temporal dynamics", "idea": "Adjacent-frame difference residual learning", "techniques": "Explicit temporal difference modeling", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "Gait3d, GREW, SUSTech1K", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/IJCB65343.2025.11411492", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 61 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "TDGait_Modeling_Temporal_Dynamics_for_Gait_Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition has emerged as a promising biometric identification method due to its non-cooperative and nonintrusive nature. However, existing methods often exhibit limitations in effectively modeling the temporal dynamics of human gait, which are crucial for robust identification in real-world scenarios. This paper introduces TDGait, a novel network specifically designed to address these limitations by explicitly modeling temporal dynamics of gait patterns. The core innovations of TDGait are twofold: 1) a Temporal Difference Module (TDM) is proposed to efficiently capture frame-to-frame variations, focusing on dynamic changes rather than static appearance. 2) a Residual Dynamic Block (RDB) is introduced to seamlessly integrate TDM without impairing spatial feature learning capacity. This block preserves spatial learning via a residual pathway while enhancing dynamic information through the embedded TDM. TDGait hierarchically combines spatial feature extraction with this explicit temporal dynamics modeling using TDM within RDBs. Comprehensive experiments on three benchmark datasets demonstrate state-of-the-art performance, achieving Rank-1 accuracies of 83.38% on SUSTech1K, 72.5% on Gait3D, and 79.1% on GREW. Extensive ablation studies validate the effectiveness and efficiency of the TDM and RDB components in capturing temporal dynamics. This work provides a new perspective on modeling the temporal dynamics of human gait, contributing to more robust biometric systems resilient to real-world variations.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/IJCB65343.2025.11411492", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "hou2025edge-ijcv-2024", "citationKey": "hou2025edge", "archiveKey": "hou2025edge", "method": "Edge-Oriented Adversarial Attack", "title": "Edge-oriented adversarial attack for deep gait recognition", "authors": "Hou, Saihui and Wang, Zengbin and Zhang, Man and Cao, Chunshui and Liu, Xu and Huang, Yongzhen", "year": 2025, "venue": "IJCV", "motivation": "Evaluate gait model robustness", "idea": "Sparse edge-focused adversarial attack", "techniques": "Spatial sparsity, temporal sparsity, targeted perturbation", "inputs": "Silhouette, GEI", "backbone": "CNN", "loss": "Untargeted Loss, Targeted Loss, Imperceptible Loss", "datasets": "CASIA-B, OU-MVLP, Gait3D, GREW", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1007/s11263-024-02225-1", "bibtex": "@article{hou2025edge,\n title = {Edge-oriented adversarial attack for deep gait recognition},\n author = {Hou, Saihui and Wang, Zengbin and Zhang, Man and Cao, Chunshui and Liu, Xu and Huang, Yongzhen},\n year = {2025},\n journal = {International Journal of Computer Vision},\n volume = {133},\n number = {4},\n pages = {1549--1563},\n publisher = {Springer}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 96 }, { "source": "Author paper summaries", "sheet": "journals", "row": 3 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Privacy Concerns and Potential Solutions" ], "yearNote": "Archive: 2024; bibliography: 2025", "abstract": "Gait recognition is a non-intrusive method that captures unique walking patterns without subject cooperation, which has emerged as a promising technique across various fields. Recent studies based on Deep Neural Networks (DNNs) have notably improved the performance, however, the potential vulnerability inherent in DNNs and their resistance to interference in practical gait recognition systems remain under-explored. To fill the gap, in this paper, we focus on imperceptible adversarial attack for deep gait recognition and propose an edge-oriented attack strategy tailored for silhouette-based approaches. Specifically, we make a pioneering attempt to explore the intrinsic characteristics of binary silhouettes, with a primary focus on injecting noise perturbations into the edge area. This simple yet effective solution enables sparse attack in both the spatial and temporal dimensions, which largely ensures imperceptibility and simultaneously achieves high success rate. In particular, our solution is built on a unified framework, allowing seamless switching between untargeted and targeted attack modes. Extensive experiments conducted on in-the-lab and in-the-wild benchmarks validate the effectiveness of our attack strategy and emphasize the necessity to study adversarial attack and defense strategy in the near future.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1007/s11263-024-02225-1", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-ye2025biggergait", "citationKey": "ye2025biggergait", "archiveKey": "ye2025biggergait", "method": "BiggerGait", "title": "BiggerGait: Unlocking Gait Recognition with Layer-wise Representations from Large Vision Models", "authors": "Dingqiang Ye and Chao Fan and Zhanbo Huang and Chengwen Luo and Jianqiang Li and Shiqi Yu and Xiaoming Liu", "year": 2025, "venue": "NeurIPS", "motivation": "Different layers of large vision models retain different gait-relevant information.", "idea": "Extract complementary gait evidence from layer-wise representations.", "techniques": "Large vision models; multi-layer feature extraction; gait representation learning.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://proceedings.neurips.cc/paper_files/paper/2025/file/6a5c23219f401f3efd322579002dbb80-Paper-Conference.pdf", "bibtex": "@inproceedings{ye2025biggergait,\n title = {BiggerGait: Unlocking Gait Recognition with Layer-wise Representations from Large Vision Models},\n author = {Dingqiang Ye and Chao Fan and Zhanbo Huang and Chengwen Luo and Jianqiang Li and Shiqi Yu and Xiaoming Liu},\n year = {2025},\n booktitle = {Advances in Neural Information Processing Systems},\n url = {https://proceedings.neurips.cc/paper_files/paper/2025/file/6a5c23219f401f3efd322579002dbb80-Paper-Conference.pdf}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Advances in Benchmarks / Popular Datasets", "Advances in Benchmarks / State-of-the-Art Performance", "Advances in Benchmarks / Public Codebases", "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Gait Learning Frameworks", "Challenges and Outlooks / Gait Foundation Models" ], "abstract": "Large vision models (LVM) based gait recognition has achieved impressive performance. However, existing LVM-based approaches may overemphasize gait priors while neglecting the intrinsic value of LVM itself, particularly the rich, distinct representations across its multi-layers. To adequately unlock LVM’s potential, this work investigates the impact of layer-wise representations on downstream recognition tasks. Our analysis reveals that LVM’s intermediate layers offer complementary properties across tasks, integrating them yields an impressive improvement even without rich well-designed gait priors. Building on this insight, we propose a simple and universal baseline for LVM-based gait recognition, termed BiggerGait. Comprehensive evaluations on CCPG, CAISA-B*, SUSTech1K, and CCGR_MINI validate the superiority of BiggerGait across both within- and cross-domain tasks, establishing it as a simple yet practical baseline for gait representation learning. All the models and code will be publicly available.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://proceedings.neurips.cc/paper_files/paper/2025/file/6a5c23219f401f3efd322579002dbb80-Paper-Conference.pdf", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "he2025radar-pr-2025", "citationKey": "he2025radar", "archiveKey": "he2025radar", "method": "DST-A2F", "title": "Radar Gait Recognition Using Dual-branch Swin Transformer with Asymmetric Attention Fusion", "authors": "Wentao He and Jianfeng Ren and Ruibin Bai and Xudong Jiang", "year": 2025, "venue": "PR", "motivation": "Overcome limitations of camera-based gait recognition", "idea": "Dual-branch transformer for radar signals", "techniques": "Spectrogram–CVD fusion; hierarchical attention", "inputs": "Spectrogram, CVD", "backbone": "Swin Transformer, Attention", "loss": "CE", "datasets": "NTU-RGR, MMRGait-1.0", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2024.111101", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "journals", "row": 17 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Radar-gait-recognition-using-Dual-branch-Swin-Transformer-_2025_Pattern-Reco.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Video-based gait recognition suffers from potential privacy issues and performance degradation due to dim environments, partial occlusions, or camera view changes. Radar has recently become increasingly popular and overcome various challenges presented by vision sensors. To capture tiny differences in radar gait signatures of different people, a dual-branch Swin Transformer is proposed, where one branch captures the time variations of the radar micro-Doppler signature and the other captures the repetitive frequency patterns in the spectrogram. Unlike natural images where objects can be translated, rotated, or scaled, the spatial coordinates of spectrograms and CVDs have unique physical meanings, and there is no affine transformation for radar targets in these synthetic images. The patch splitting mechanism in Vision Transformer makes it ideal to extract discriminant information from patches, and learn the attentive information across patches, as each patch carries some unique physical properties of radar targets. Swin Transformer consists of a set of cascaded Swin blocks to extract semantic features from shallow to deep representations, further improving the classification performance. Lastly, to highlight the branch with larger discriminant power, an Asymmetric Attention Fusion is proposed to optimally fuse the discriminant features from the two branches. To enrich the research on radar gait recognition, a large-scale NTU-RGR dataset is constructed, containing 45,768 radar frames of 98 subjects. The proposed method is evaluated on the NTU-RGR dataset and the MMRGait-1.0 database. It consistently and significantly outperforms all the compared methods on both datasets. The codes are available at: https://github.com/wentaoheunnc/NTU-RGR.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2024.111101", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "peng2025rsanet-pr-2025", "citationKey": "peng2025rsanet", "archiveKey": "peng2025rsanet", "method": "RSANet", "title": "RSANet: Relative-sequence quality assessment network for gait recognition in the wild", "authors": "Peng, Guozhen and Wang, Yunhong and Zhang, Shaoxiong and Li, Rui and Zhao, Yuwei and Li, Annan", "year": 2025, "venue": "PR", "motivation": "Handle low-quality wild silhouettes", "idea": "Relative sequence quality assessment", "techniques": "Inter-frame similarity weighting; receptive field enhancement", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "Gait3D, GREW, BUAA-Duke-Gait", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2024.111219", "bibtex": "@article{peng2025rsanet,\n title = {RSANet: Relative-sequence quality assessment network for gait recognition in the wild},\n author = {Peng, Guozhen and Wang, Yunhong and Zhang, Shaoxiong and Li, Rui and Zhao, Yuwei and Li, Annan},\n year = {2025},\n journal = {Pattern Recognition},\n volume = {161},\n pages = {111219},\n publisher = {Elsevier}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 111 }, { "source": "Author paper summaries", "sheet": "journals", "row": 18 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Handling Real-World Noise" ], "abstract": "Gait recognition in the wild has received increasing attention since the gait pattern is hard to disguise and can be captured in a long distance. However, due to occlusions and segmentation errors, low-quality silhouettes are common and inevitable. To mitigate this low-quality problem, some prior arts propose absolute-single quality assessment models. Although these methods obtain a good performance, they only focus on the silhouette quality of a single frame, lacking consideration of the variation state of the entire sequence. In this paper, we propose a Relative-Sequence Quality Assessment Network, named RSANet. It uses the Average Feature Similarity Module (AFSM) to evaluate silhouette quality by calculating the similarity between one silhouette and all other silhouettes in the same silhouette sequence. The silhouette quality is based on the sequence, reflecting a relative quality. Furthermore, RSANet uses Multi-Temporal-Receptive-Field Residual Blocks (MTB) to extend temporal receptive fields without parameter increases. It achieves a Rank-1 accuracy of 75.2% on Gait3D, 81.8% on GREW, and 77.6% on BUAA-Duke-Gait datasets respectively. The code is available at https://github.com/PGZ-Sleepy/RSANet.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2024.111219", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "shen2024comprehensive-tbiom-2025", "citationKey": "shen2024comprehensive", "archiveKey": "shen2024comprehensive", "method": "", "title": "A comprehensive survey on deep gait recognition: Algorithms, datasets, and challenges", "authors": "Shen, Chuanfu and Yu, Shiqi and Wang, Jilong and Huang, George Q and Wang, Liang", "year": 2025, "venue": "TBIOM", "motivation": "Survey deep learning gait recognition", "idea": "Systematic taxonomy and evaluation", "techniques": "Representation learning and architectures", "inputs": "", "backbone": "", "loss": "", "datasets": "CASIA-B, OU-MVLP, Gait3D, GREW", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TBIOM.2024.3486345", "bibtex": "@article{shen2024comprehensive,\n title = {A comprehensive survey on deep gait recognition: Algorithms, datasets, and challenges},\n author = {Shen, Chuanfu and Yu, Shiqi and Wang, Jilong and Huang, George Q and Wang, Liang},\n year = {2025},\n journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},\n volume = {7},\n number = {2},\n pages = {270--292},\n publisher = {IEEE},\n doi = {10.1109/TBIOM.2024.3486345}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 130 }, { "source": "Author paper summaries", "sheet": "journals", "row": 37 } ], "summaryStatus": "Author collection", "sections": [ "Introduction / Context", "Privacy Concerns and Potential Solutions" ], "abstract": "Gait recognition aims to identify a person at a distance, serving as a promising solution for long-distance and less-cooperation pedestrian recognition. Recently, significant advances in gait recognition have achieved inspiring success in many challenging scenarios by utilizing deep learning techniques. Against the backdrop that deep gait recognition has achieved almost perfect performance in laboratory datasets, much recent research has introduced new challenges for gait recognition, including robust deep representation modeling, in-the-wild gait recognition, and even recognition from new visual sensors such as infrared and depth cameras. Meanwhile, the increasing performance of gait recognition might also reveal concerns about biometrics security and privacy prevention for society. We provide a comprehensive survey on recent literature using deep learning and a discussion on the privacy and security of gait biometrics. This survey reviews the existing deep gait recognition methods through a novel view based on our proposed taxonomy. The proposed taxonomy differs from the conventional taxonomy of categorizing available gait recognition methods into the modelor appearance-based methods, while our taxonomic hierarchy considers deep gait recognition from two perspectives: deep representation learning and deep network architectures, illustrating the current approaches from both micro and macro levels. We also include up-to-date reviews of datasets and performance evaluations on diverse scenarios. Finally, we introduce privacy and security concerns on gait biometrics and discuss outstanding challenges and potential directions for future research.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TBIOM.2024.3486345", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "liao2024denseposegait-tbiom-2025", "citationKey": "liao2024denseposegait", "archiveKey": "liao2024denseposegait", "method": "DensePoseGait", "title": "DensePoseGait Dense Human Pose Part-Guided for Gait Recognition", "authors": "Rijun Liao and Zhu Li and Shuvra S. Bhattacharyya and George York", "year": 2025, "venue": "TBIOM", "motivation": "Robust 3D gait recognition", "idea": "Leverage dense 3D body surfaces", "techniques": "Part-guided feature learning alignment", "inputs": "DensePose", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, Gait3D, GREW", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TBIOM.2024.3486732", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 131 }, { "source": "Author paper summaries", "sheet": "journals", "row": 38 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "DensePoseGait_Dense_Human_Pose_Part-Guided_for_Gait_Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition is a technology that identifies human ID according to the human unique biometric gait feature. It has two popular categories, appearance-based and modelbased algorithms. Appearance-based algorithms generally use human silhouettes as the initial input data. External factors such as clothing and physical carrying can drastically alter human silhouettes. In contrast, model-based algorithms tend to be more robust in regard to appearances, with human skeletons providing the initial input data in general. However, human skeletons suffer from limited information which causes an obstacle to increasing performance. In this paper, we, therefore, address this challenge by presenting two new databases, named CASIA-B-DensePose and MoBo-DensePose, which are based on the publicly available multiview database, CASIA-B and MoBo. They exploit UV coordinates of body surface and human semantic segmentation as the initial gait feature. It is less sensitive to human shape compared with human silhouettes, and has richer semantic information compared with human skeletons. In addition, we also introduce a novel model-based framework, DensePoseGait, to take full advantage of databases. Unlike traditional algorithms which either extract isolated local features or combine them with global features, DensePoseGait uses a novel way to exploit partial features. That is, human pose parts are employed as a regulator to guide the learning of global features in the training stage. Its core idea is to establish better representative features with the assistance of partial features, but not require additional calculation in the inference stage. We believe these databases and framework can offer researchers a fresh perspective on model-based gait recognition and inspire further exploration and advancements in this area.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TBIOM.2024.3486732", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "guo2025distillation-tbiom-2025", "citationKey": "guo2025distillation", "archiveKey": "guo2025distillation", "method": "H-GADER", "title": "Distillation-Guided Representation Learning for Unconstrained Video Human Authentication", "authors": "Yuxiang Guo and Siyuan Huang and Ram Prabhakar Kathirvel and Chun Pong Lau and Rama Chellappa and Cheng Peng", "year": 2025, "venue": "TBIOM", "motivation": "Improve unconstrained human authentication", "idea": "Combine gait and body recognition adaptively", "techniques": "Motion detection; cross-modal distillation; semantic feature extraction; conditional model fusion", "inputs": "Silhouette, RGB, Skeleton", "backbone": "CNN, GCN, Transformer", "loss": "CE, Triplet", "datasets": "CASIA-B, Gait3D, CCPG", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TBIOM.2025.3595366", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 134 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Distillation-Guided_Representation_Learning_for_Unconstrained_Video_Human_Authentication.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Human authentication is an important and challenging biometric task, particularly from unconstrained videos. While body recognition is a popular approach, gait recognition holds the promise of robustly identifying subjects based on walking patterns instead of appearance information. Previous gait-based approaches have performed well for curated indoor scenes; however, they tend to underperform in unconstrained situations. To address these challenges, we propose a framework, termed Holistic GAit DEtection and Recognition (H-GADER), for human authentication in challenging outdoor scenarios. Specifically, H-GADER leverages a Double Helical Signature to detect segments that contain human movement and builds discriminative features through a novel gait recognition method. To further enhance robustness, H-GADER encodes viewpoint information in its architecture, and distills learned representations from an auxiliary RGB recognition model; this allows H-GADER to learn from maximum amount of data at training time. At test time, H- GADER infers solely from the silhouette modality. Furthermore, we introduce a body recognition model through semantic, largescale, self-supervised training to complement gait recognition. By conditionally fusing gait and body representations based on the presence/absence of gait information as decided by the gait detection, we demonstrate significant improvements compared to when a single modality or a naive feature ensemble is used. We evaluate our method on multiple existing State-of-The-Arts (SoTA) gait baselines and demonstrate consistent improvements on indoor and outdoor datasets, especially on the BRIAR dataset, which features unconstrained, long-distance videos, achieving a 28.9% improvement.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TBIOM.2025.3595366", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ren2025unsupervised-tbiom-2025", "citationKey": "ren2025unsupervised", "archiveKey": "ren2025unsupervised", "method": "Selective Fusion (SF)", "title": "Unsupervised gait recognition with selective fusion", "authors": "Ren, Xuqian and Yang, Shaopeng and Hou, Saihui and Cao, Chunshui and Liu, Xu and Huang, Yongzhen", "year": 2025, "venue": "TBIOM", "motivation": "Unsupervised gait recognition improvement", "idea": "Gradually align cross-view and clothing", "techniques": "Selective cluster and sample fusion", "inputs": "Silhouette", "backbone": "CNN", "loss": "Contrastive, ClusterNCE Loss", "datasets": "CASIA-BN, Outdoor-Gait, GREW", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TBIOM.2025.3544805", "bibtex": "@article{ren2025unsupervised,\n title = {Unsupervised gait recognition with selective fusion},\n author = {Ren, Xuqian and Yang, Shaopeng and Hou, Saihui and Cao, Chunshui and Liu, Xu and Huang, Yongzhen},\n year = {2025},\n journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 133 }, { "source": "Author paper summaries", "sheet": "journals", "row": 40 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Gait Learning Frameworks" ], "abstract": "Previous gait recognition methods primarily relied on labeled datasets, which require a labor-intensive labeling process. To eliminate this dependency, we focus on a new task: Unsupervised Gait Recognition (UGR). We introduce a cluster-based baseline to solve UGR. However, we identify additional challenges in this task. First, sequences of the same person in different clothes tend to cluster separately due to significant appearance changes. Second, sequences captured from 0◦ and 180◦ views lack distinct walking postures and do not cluster with sequences from other views. To address these challenges, we propose a Selective Fusion method, consisting of Selective Cluster Fusion (SCF) and Selective Sample Fusion (SSF). SCF merges clusters of the same person wearing different clothes by updating the cluster-level memory bank using a multi-cluster update strategy. SSF gradually merges sequences taken from front/back views using curriculum learning. Extensive experiments demonstrate the effectiveness of our method in improving rank-1 accuracy under different clothing and view conditions.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TBIOM.2025.3544805", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "lu2024flag-tcsvt-2025", "citationKey": "lu2024flag", "archiveKey": "lu2024flag", "method": "FLAG", "title": "FLAG A Framework With Explicit Learning Based on Appearance and Gait for Video-Based Clothes-Changing Person Re-Identification", "authors": "Hengjie Lu and Yilin Gao and Shugong Xu", "year": 2025, "venue": "TCSVT", "motivation": "Handle clothes-changing ReID challenge", "idea": "Explicitly learn appearance and gait", "techniques": "Multi-modal fusion; multi-scale gait modeling; cross-modal knowledge distillation", "inputs": "RGB, Silhouette", "backbone": "CNN, Transformer", "loss": "CE, Triplet", "datasets": "CCVID", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TCSVT.2024.3483265", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 149 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "FLAG_A_Framework_With_Explicit_Learning_Based_on_Appearance_and_Gait_for_Video-Based_Clothes-Changing_Person_Re-Identification.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Person re-identification (ReID) aims to search for the target person among the non-overlapping surveillance cameras. Video-based clothes-changing person re-identification (VCC-ReID) has become an essential branch of ReID due to the rich spatial and temporal information in the videos and the broad application of the scenarios. Appearance and gait are discriminative features in the video-based ReID, but appearance information is limited due to the clothes changing, which makes the VCC-ReID challenging. To solve this challenge, we propose a Framework with explicit Learning based on Appearance and Gait (FLAG), which can explicitly extract two corresponding types of information and be combined with most existing video-based ReID methods. The FLAG includes a multi-modal and multigranularities Architecture (MGA), which is a large model, and a Cross-Modal Knowledge Distillation Scheme (CMKDS), which has a small model. They can be applied to devices with different computing resources. The MGA is designed to simultaneously take the visible light and silhouette modalities as input to explicitly learn the appearance and gait features, respectively. The silhouette modalities are composed of several levels of granularities to model global and local gait features and independently serve as input for MGA. The Embedding-Based parallel fusion module is proposed to fuse the appearance and multi-granularities gait feature efficiently. The CMKDS is present to distill the MGA to a small single-modal model that only uses the visible light modality as input. The Embedding-Based direct and indirect distillation strategies are designed in the CMKDS. Experimental results demonstrate that the FLAG combined with the existing video-based ReID methods can significantly improve their performance. In addition, when FLAG is combined with the AP3D method, the MGA can outperform state-of-the-art accuracy by 4.2%.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TCSVT.2024.3483265", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "chen2025fusiongait-tcsvt-2025", "citationKey": "Chen2025FusionGait", "archiveKey": "Chen2025FusionGait", "method": "FusionGait", "title": "Fusing Events and Frames for Robust Gait Recognition", "authors": "Liaogehao Chen and Zhenjun Zhang and Changchang Li and Yaonan Wang", "year": 2025, "venue": "TCSVT", "motivation": "Robust low-light gait recognition", "idea": "Self-supervised cost-volume fusion", "techniques": "Event-frame complementary fusion", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, EV-CASIA-B", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TCSVT.2025.3642699", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 150 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Fusing_Events_and_Frames_for_Robust_Gait_Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Reliable gait recognition under low-light conditions remains challenging for traditional cameras. Event cameras, with their high dynamic range and fine temporal resolution, offer a promising alternative but suffer from sparse signals under small motion amplitudes and vulnerability to noise or irrelevant movements (e.g., shadows). To address these issues, we propose FusionGait, a complementary fusion framework that integrates event with standard frames to achieve robust gait recognition. Specifically, we propose a Self-supervised Hierarchical Feature Extractor (SSL-HFE) built upon DINOv2, which employs learnable prompts to bridge the gap between gray frames and RGB frames, extracts multi-level semantic features, and enhances their discriminability through a self-supervised learning strategy. Then, we introduce the Complementary Fusion Learning Module (CFLM), which employs cross-cost volumes to explicitly model pixel-level correlations between frames and events, enabling effective cross-modal interaction and fusion. Furthermore, we propose EvGSimulator, a sensor-specific data augmentation strategy that simulates diverse illumination conditions based on physical properties. The framework maintains robustness even when real frames are unavailable by reconstructing frame-like representations from events, and it scales to ultra–high-frame-rate scenarios with hundreds of frames per second. We also collect DA VIS346-Gait-RGE, the multi-view semi-indoor & outdoor gait dataset captured with a DA VIS346 event camera, including three modalities: event streams, gray frames, and reconstructed frames. Experiments across multiple datasets show FusionGait achieves state-of-the-art performance, effectively surpassing single-modality methods.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TCSVT.2025.3642699", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "wang2025gaitc-tcsvt-2025", "citationKey": "wang2025gaitc", "archiveKey": "wang2025gaitc", "method": "GaitC3I", "title": "GaitC3I: Robust Cross-Covariate Gait Recognition via Causal Intervention", "authors": "Wang, Jilong and Hou, Saihui and Guo, Xianda and Huang, Yan and Huang, Yongzhen and Zhang, Tianzhu and Wang, Liang", "year": 2025, "venue": "TCSVT", "motivation": "Mitigate covariate bias in gait recognition", "idea": "Introduce causal intervention for robust metric learning", "techniques": "Structural causal modeling, backdoor adjustment, dynamic negative sampling", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet, Contrastive", "datasets": "Gait3D, GREW, OU-MVLP, CASIA-B, CCPG, CCGR", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TCSVT.2025.3545210", "bibtex": "@article{wang2025gaitc,\n title = {GaitC3I: Robust Cross-Covariate Gait Recognition via Causal Intervention},\n author = {Wang, Jilong and Hou, Saihui and Guo, Xianda and Huang, Yan and Huang, Yongzhen and Zhang, Tianzhu and Wang, Liang},\n year = {2025},\n journal = {IEEE Transactions on Circuits and Systems for Video Technology},\n volume = {35},\n number = {8},\n pages = {8057--8070},\n publisher = {IEEE},\n doi = {10.1109/TCSVT.2025.3545210}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 147 }, { "source": "Author paper summaries", "sheet": "journals", "row": 53 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Benchmarks / State-of-the-Art Performance", "Advances in Methods / Gait Learning Frameworks", "Advances in Methods / Handling Real-World Noise", "Advances in Methods / Gait Metric Designs", "Challenges and Outlooks / Real-World Gait Recognition" ], "abstract": "Cross-covariate gait recognition aims to analyze a pedestrian’s gait to extract an identity representation that is invariant across varying covariates. However, prevailing methods that have achieved good results on controlled in-the-lab datasets often perform poorly on realistic datasets. In this work, we find a significant cause is that the widely used pairwise metric learning paradigm cannot correctly handle the relationship between samples from different covariate conditions. Even worse, it may yield harmful signals that inadvertently mislead models to focus on covariate-related features, particularly when covariate distributions vary across subjects. To address this issue, we propose a Cross-Covariate Causal Intervention (GaitC 3I) framework, a unified causality-inspired approach aimed at enhancing the robustness of gait recognition across diverse conditions. Specifically, our method consists of two parts: 1) an effective causal intervention metric learning paradigm based on backdoor adjustment, which strategically mitigates spurious correlations induced by covariates, thus ensuring a more invariant gait representation; and 2) an annotation-free selection strategy that progressively matches each positive sample with negative samples from similar covariate conditions at various granularities. We demonstrate the effectiveness of our GaitC 3I through extensive evaluation on six popular gait datasets–Gait3D, GREW, OUMVLP, CASIA- B, CCPG, and CCGR–achieving substantial improvements. Our method not only outperforms existing state-of-the-art models but also provides a systematic solution to remove the spurious correlations in gait recognition.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TCSVT.2025.3545210", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "zhai2024multi-tcsvt-2025", "citationKey": "zhai2024multi", "archiveKey": "zhai2024multi", "method": "LMGCS", "title": "Multi-View Gait Recognition With Joint Local Multi-Scale and Global Contextual Spatio-Temporal Features", "authors": "Wenzhe Zhai and Haomiao Li and Chaoqun Zheng and Xianglei Xing", "year": 2025, "venue": "TCSVT", "motivation": "Improve spatio-temporal gait feature extraction", "idea": "Combine fine-grained local and global contextual features", "techniques": "Multi-resolution temporal extraction, feature fusion, self-attention contextual modeling", "inputs": "Silhouette", "backbone": "CNN, Transformer", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TCSVT.2024.3476384", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 148 }, { "source": "Author paper summaries", "sheet": "journals", "row": 54 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Multi-View_Gait_Recognition_With_Joint_Local_Multi-Scale_and_Global_Contextual_Spatio-Temporal_Features.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Existing gait recognition methods are capable of extracting rich spatial gait information but often overlook fine-grained temporal features within local regions and temporal contextual information across different sub-regions. Considering gait recognition as a fine-grained recognition task and each individual exhibits uniqueness in their movements across different temporal sequences, we propose a local multi-scale and global contextual spatio-temporal (LMGCS) network for gait recognition. It divides the whole gait sequence into sub-sequences with multiple spatio resolutions and extracts multi-scale temporal features. We extract the temporal context information of different sub-sequences with the transformer, and all sub-sequences are fused to form global features. Furthermore, the loss function that combines the triplet loss function and cross-entropy loss function is utilized to prompt the proposed model to fulfill the gait recognition. The proposed method achieved state-of-theart results on two popular public datasets. It achieved rank-1 accuracy of 98.0%, 95.4%, and 85.0% on the three walk states of the CASIA-B dataset and 90.9% on the OU-MVLP dataset.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TCSVT.2024.3476384", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "wang2025new-tifs-2025", "citationKey": "wang2025new", "archiveKey": "wang2025new", "method": "CCVReID", "title": "A New Benchmark and Algorithm for Clothes-Changing Video Person Re-Identification", "authors": "Likai Wang and Xiangqun Zhang and Ruize Han and Yanjie Wei and Song Wang and Wei Feng", "year": 2025, "venue": "TIFS", "motivation": "Enable robust clothes-changing video Re-ID", "idea": "Combine appearance and gait cues", "techniques": "Dual-branch fusion; confidence-based reweighting; ranking graph integration", "inputs": "RGB, Silhouette", "backbone": "CNN, GCN", "loss": "Triplet", "datasets": "", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/tifs.2025.3539079", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 164 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "A_New_Benchmark_and_Algorithm_for_Clothes-Changing_Video_Person_Re-Identification.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Person re-identification (Re-ID) is a classical computer vision task and has significant applications for public security and information forensics. Recently, long-term Re-ID with clothes-changing has attracted increasing attention. However, existing methods mainly focus on image-based setting, where richer temporal information is overlooked. In this paper, we focus on the relatively new yet practical problem of Clothes-Changing Video-based Re-ID (CCVReID), which is less studied. First, given the dataset shortage, we build two new benchmark datasets for CCVReID problem, including a large-scale synthetic video dataset and a real-world one, both containing human sequences with various clothing changes. Moreover, we systematically study this problem by simultaneously considering the classical appearance feature and temporal feature contained in the video. We develop a dual-branch fusion framework that makes use of the information from both clothes-aware appearance feature and clothes-free gait feature. For better information fusion, a confidence-guided re-ranking strategy is proposed to adaptively balance the weight of these two categories of features. We have released the benchmark and code proposed in this work to the public at https://github.com/kkw98/CCVReID.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/tifs.2025.3539079", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "yang2025multimodal-tifs-2025", "citationKey": "yang2025multimodal", "archiveKey": "yang2025multimodal", "method": "M3L", "title": "Multimodal Mutual Learning for Unsupervised Gait Recognition", "authors": "Yang, Shaopeng and Hou, Saihui and Liu, Xu and Cao, Chunshui and Ma, Kang and Huang, Yongzhen", "year": 2025, "venue": "TIFS", "motivation": "Robust unsupervised gait recognition", "idea": "Leverage complementary multimodal cues", "techniques": "Cross-modal positive pairs, mutual prediction, feature alignment, modality decoupling", "inputs": "Silhouette, Skeleton", "backbone": "CNN", "loss": "Contrastive", "datasets": "Gait3D, GREW, CASIA-B, SUSTech1K", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TIFS.2025.3602638", "bibtex": "@article{yang2025multimodal,\n title = {Multimodal Mutual Learning for Unsupervised Gait Recognition},\n author = {Yang, Shaopeng and Hou, Saihui and Liu, Xu and Cao, Chunshui and Ma, Kang and Huang, Yongzhen},\n year = {2025},\n journal = {IEEE Transactions on Information Forensics and Security},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 162 }, { "source": "Author paper summaries", "sheet": "journals", "row": 66 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Gait Learning Frameworks" ], "abstract": "The primary challenge in unsupervised gait recognition lies in generating meaningful and diverse supervisory signals to guide representation learning. The effectiveness of such methods largely depends on the richness of the supervisory signals. Unlike previous methods that construct supervisory signals solely from a single modality, we propose a novel framework, named Multimodal Mutual Learning (M 3L), that leverages the identity consistency and complementary nature of both silhouette and skeleton modalities to generate richer and more informative supervisory signals. To fully leverage the richer supervisory signals, M3L encourages mutual prediction between the silhouette and skeleton modalities, guiding the network toward modalityinvariant representations. However, mutual prediction alone is hindered by the inherent modality gap, so we introduce a Multimodal Collaborative Module to explicitly bridge this gap and promote cross-modal knowledge transfer. Moreover, to make the framework practical when only one modality is available at inference, we introduce a Multimodal Disentanglement Module. Multimodal Disentanglement Module decouples the two branches and distills a shared representation, preserving the gains of multimodal training while allowing the model to maintain robust performance under single-modality conditions. Extensive experiments on four widely used gait datasets—Gait3D, GREW, CASIA-B, and SUSTech1K—demonstrate the effectiveness of our approach and highlight its potential to advance unsupervised gait recognition.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIFS.2025.3602638", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "lu2025mojo-tifs-2025", "citationKey": "lu2025mojo", "archiveKey": "lu2025mojo", "method": "MOJO", "title": "MOJO MOtion Pattern Learning and JOint-Based Fine-Grained Mining for Person Re-Identification Based on 4D LiDAR Point Clouds", "authors": "Zhiyang Lu and Chenglu Wen and Ming Cheng and Cheng Wang", "year": 2025, "venue": "TIFS", "motivation": "Enable robust LiDAR-based identification", "idea": "Leverage motion patterns in 4D point clouds", "techniques": "Motion field modeling; cross-frame correction; joint-based graph learning; hierarchical feature aggregation", "inputs": "Point", "backbone": "GNN, MLP", "loss": "CE, Triplet", "datasets": "LReID, SUSTech1K", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1109/TIFS.2025.3614500", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 163 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "MOJO_MOtion_Pattern_Learning_and_JOint-Based_Fine-Grained_Mining_for_Person_Re-Identification_Based_on_4D_LiDAR_Point_Clouds.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Person Re-identification (ReID) primarily involves the extraction of discriminative representations derived from morphological characteristics, gait patterns, and related attributes. While camera-based Person ReID methods yield notable results, their reliability diminishes in scenarios involving long distances and limited illumination. LiDAR enables the precise acquisition of human point cloud sequences across extended distances, unaffected by variations in lighting or similar factors. Nevertheless, current LiDAR-based Person ReID techniques are limited to static measurements, rendering them susceptible to perturbations from attire variations, occlusions, and similar confounding factors. To address these issues, this manuscript introduces MOJO, which is applied to 4D LiDAR point clouds to extract unique motion patterns specific to individuals. To characterize the motion patterns across two consecutive point cloud frames, MOJO employs optimal transport to compute point-wise motion vectors, thereby enabling the identification of discriminative implicit motion information. To mitigate the attenuation of point cloud density induced by self-occlusion during dynamic motion, MOJO leverages inverse point-wise flow information to integrate forward frames, thereby yielding a comprehensive representation, whilst concurrently ameliorating the e ffects of heterogeneous density distribution within localized regions of the 4D point cloud data. Additionally, the inherent unordered nature and sparsity of 4D point clouds present significant obstacles to capturing discriminative features. We develop the 3D joint graph to extract scalable fine-grained traits and employ the joint pyramid pooling module to conduct hierarchical spatiotemporal aggregation across the 4D point clouds. Extensive experimental evaluations demonstrate that MOJO achieves state-of-the-art (SOTA) accuracy on the LReID dataset (for LiDAR-based Person Re-identification) and SUSTech1k dataset (for LiDAR-based Gait Recognition) without any pre-training while exhibiting robust performance across various point cloud densities. Our code will be available at https://github.com/O-VIGIA/MOJO", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIFS.2025.3614500", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "dou2024clash-tip-2025", "citationKey": "dou2024clash", "archiveKey": "dou2024clash", "method": "CLASH", "title": "CLASH Complementary Learning With Neural Architecture Search for Gait Recognition", "authors": "Huanzhang Dou and Pengyi Zhang and Yuhan Zhao and Lu Jin and Xi Li", "year": 2025, "venue": "TIP", "motivation": "Informative and robust gait representation", "idea": "Combine dense and sparse descriptors", "techniques": "Pixel-boundary distance encoding, foreground-background separation, complementary learning, automated architecture search", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, Gait3D, GREW", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TIP.2024.3360870", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 171 }, { "source": "Author paper summaries", "sheet": "journals", "row": 73 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "CLASH_Complementary_Learning_With_Neural_Architecture_Search_for_Gait_Recognition.pdf", "abstract": "Gait recognition, which aims at identifying individuals by their walking patterns, has achieved great success based on silhouette. The binary silhouette sequence encodes the walking pattern within the sparse boundary representation. Therefore, most pixels in the silhouette are under-sensitive to the walking pattern since the sparse boundary lacks dense spatialtemporal information, which is suitable to be represented with dense texture. To enhance the sensitivity to the walking pattern while maintaining the robustness of recognition, we present a Complementary Learning with neural Architecture SearcH (CLASH) framework, consisting of walking pattern sensitive gait descriptor named dense spatial-temporal field (DSTF) and neural architecture search based complementary learning (NCL). Specifically, DSTF transforms the representation from the sparse binary boundary into the dense distance-based texture, which is sensitive to the walking pattern at the pixel level. Further, NCL presents a task-specific search space for complementary learning, which mutually complements the sensitivity of DSTF and the robustness of the silhouette to represent the walking pattern effectively. Extensive experiments demonstrate the effectiveness of the proposed methods under both in-the-lab and in-thewild scenarios. On CASIA-B, we achieve rank-1 accuracy of 98.8%, 96.5%, and 89.3% under three conditions. On OU-MVLP, we achieve rank-1 accuracy of 91.9%. Under the latest in-thewild datasets, we outperform the latest silhouette-based methods by 16.3% and 19.7% on Gait3D and GREW, respectively.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIP.2024.3360870", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "peng2024masked-tmm-2025", "citationKey": "peng2024masked", "archiveKey": "peng2024masked", "method": "MADE", "title": "Masked Attribute Description Embedding for Cloth-Changing Person Re-Identification", "authors": "Chunlei Peng and Boyu Wang and Decheng Liu and Nannan Wang and Ruimin Hu and Xinbo Gao", "year": 2025, "venue": "TMM", "motivation": "Overcome clothes-change interference", "idea": "Remove cloth-sensitive appearance info", "techniques": "Attribute masking; feature-text fusion; shared latent mapping", "inputs": "RGB", "backbone": "Transformer, Attention", "loss": "CE, Triplet", "datasets": "PRCC, LTCC", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TMM.2024.3521730", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 180 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Masked_Attribute_Description_Embedding_for_Cloth-Changing_Person_Re-Identification.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Cloth-changing person re-identification (CC-ReID) aims to match persons who change clothes over long periods. The key challenge in CC-ReID is to extract cloth-irrelated features, such as face, hairstyle, body shape, and gait. Current research mainly focuses on modeling body shape using multi-modal biological features (such as silhouettes and sketches). However, it does not fully leverage the personal description information hidden in the original RGB image. Considering that there are certain attribute descriptions that remain unchanged after the changing of cloth, we propose a Masked Attribute Description Embedding (MADE) method that unifies personal visual appearance and attribute description for CC-ReID. Specifically, handling variable cloth-sensitive information, such as color and type, is challenging for effective modeling. To address this, we mask the clothes type and color information (upper body type, upper body color, lower body type, and lower body color) in the personal attribute description extracted through an attribute detection model. The masked attribute description is then connected and embedded into Transformer blocks at various levels, fusing it with the low-level to high-level features of the image. This approach compels the model to discard cloth information. Experiments are conducted on several CC-ReID benchmarks, including PRCC, LTCC, CelebreID-light, and LaST. Results demonstrate that MADE effectively utilizes attribute description, enhancing cloth-changing person reidentification performance, and compares favorably with state-ofthe-art methods.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TMM.2024.3521730", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "zhang2025trackletgait-tmm-2025", "citationKey": "Zhang2025TrackletGait", "archiveKey": "Zhang2025TrackletGait", "method": "TrackletGait", "title": "TrackletGait A Robust Framework for Gait Recognition in the Wild", "authors": "Shaoxiong Zhang and Jinkai Zheng and Shangdong Zhu and Chenggang Yan", "year": 2025, "venue": "TMM", "motivation": "Robust wild gait recognition", "idea": "Tracklets wavelets hard-triplet exclusion", "techniques": "Tracklet sampling wavelet downsampling", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, Gait3D, GREW, CCPG, SUSTech1K", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1109/TMM.2025.3607705", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 181 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "TrackletGait_A_Robust_Framework_for_Gait_Recognition_in_the_Wild.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition aims to identify individuals based on their body shape and walking patterns. Though much progress has been achieved driven by deep learning, gait recognition in realworld surveillance scenarios remains quite challenging to current methods. Conventional approaches, which rely on periodic gait cycles and controlled environments, struggle with the non-periodic and occluded silhouette sequences encountered in the wild. In this paper, we propose a novel framework, TrackletGait, designed to address these challenges in the wild. We propose Random Tracklet Sampling, a generalization of existing sampling methods, which strikes a balance between robustness and representation in capturing diverse walking patterns. Next, we introduce Haar Wavelet-based Downsampling to preserve information during spatial downsampling. Finally, we present a Hardness Exclusion Triplet Loss, designed to exclude low-quality silhouettes by discarding hard triplet samples. TrackletGait achieves state-ofthe-art results, with 77.8% and 80.4% rank-1 accuracy on the Gait3D and GREW datasets, respectively, while using only 10.3M backbone parameters. Extensive experiments are also conducted to further investigate the factors affecting gait recognition in the wild.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TMM.2025.3607705", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "chen2023adaptive-tnnls-2025", "citationKey": "chen2023adaptive", "archiveKey": "chen2023adaptive", "method": "AGFL", "title": "Adaptive Gait Feature Learning Using Mixed Gait Sequence", "authors": "Yifan Chen and Yang Zhao and Xuelong Li", "year": 2025, "venue": "TNNLS", "motivation": "Robust gait recognition under occlusion", "idea": "Leverage mixed sequences for learning", "techniques": "Gait sequence mixing, multigranularity feature extraction, feature distance alignment, global-local integration", "inputs": "Silhouette", "backbone": "CNN", "loss": "Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TNNLS.2023.3331050", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 184 }, { "source": "Author paper summaries", "sheet": "journals", "row": 84 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Adaptive_Gait_Feature_Learning_Using_Mixed_Gait_Sequence.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition has become a mainstream technology for identification, as it can recognize the identity of subjects from a distance without any cooperation. However, when subjects wear coats (CL) or backpacks (BG), their gait silhouette will be occluded, which will lose some gait information and bring great difficulties to the identification. Another important challenge in gait recognition is that the gait silhouette of the same subject captured by different camera angles varies greatly, which will cause the same subject to be misidentified as different individuals under different camera angles. In this article, we try to overcome these problems from three aspects: data augmentation, feature extraction, and feature refinement. Correspondingly, we propose gait sequence mixing (GSM), multigranularity feature extraction (MFE), and feature distance alignment (FDA). GSM is a method that belongs to data enhancement, which uses the gait sequences in NM to assist in learning the gait sequences in BG or CL, thus reducing the influence of lost gait information in abnormal gait sequences (BG or CL). MFE explores and fuses different granularity features of gait sequences from different scales, and it can learn as much useful information as possible from incomplete gait silhouettes. FDA refines the extracted gait features with the help of the distribution of gait features in real world and makes them more discriminative, thus reducing the influence of various camera angles. Extensive experiments demonstrate that our method has better results than some state-of-the-art methods on CASIA-B and mini-OUMVLP. We also embed the GSM module and FDA module into some state-of-the-art methods, and the recognition accuracy of these methods is greatly improved.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TNNLS.2023.3331050", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "li2025rethinking-tnnls-2025", "citationKey": "li2025rethinking", "archiveKey": "li2025rethinking", "method": "GaitBase", "title": "Rethinking Appearance-Based Deep Gait Recognition Reviews Analysis and Insights From Gait Recognition Evolution", "authors": "Jingqi Li and Yuzhen Zhang and Yi Zeng and Changxin Ye and Wenzheng Xu and Xianye Ben and Fei-Yue Wang and Junping Zhang", "year": 2025, "venue": "TNNLS", "motivation": "Generalized and stable gait recognition", "idea": "Analyze and improve gait representations", "techniques": "Global-local feature extraction, temporal modeling, pooling strategies, empirical evaluation, optimization analysis", "inputs": "", "backbone": "", "loss": "", "datasets": "CASIA-B, OU-MVLP, Gait3D, GREW", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TNNLS.2025.3526815", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 185 }, { "source": "Author paper summaries", "sheet": "journals", "row": 85 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Rethinking_Appearance-Based_Deep_Gait_Recognition_Reviews_Analysis_and_Insights_From_Gait_Recognition_Evolution.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition is a prominent biometric recognition technique extensively employed in public security. Appearance-based and model-based gait recognition are two categories of methods commonly used. Specifically, appearance-based methods, which use silhouettes to represent body information, typically outperform model-based methods that rely on skeleton data, making them more popular. Recently, the shift from single-frame templates to multiframe silhouettes has advanced appearance-based gait recognition with better spatiotemporal representation. However, there is a notable lack of comprehensive studies that deepen the understanding of multiframe appearancebased gait recognition methods. This article reviews various methods to trace the evolution of gait recognition. Furthermore, we unify various performant models in one framework, study the overlooked effects on data arrangement, and explore the scaling ability of existing methods. Besides the advancement in gait recognition, we also summarize the current challenges and future prospects to foster future research.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TNNLS.2025.3526815", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "meng2025fastposegait-tpami-2025", "citationKey": "meng2025fastposegait", "archiveKey": "meng2025fastposegait", "method": "GPGait++", "title": "From FastPoseGait to GPGait++: Bridging the Past and Future for Pose-based Gait Recognition", "authors": "Meng, Shibei and Fu, Yang and Hou, Saihui and Hu, Xuecai and Cao, Chunshui and Liu, Xu and Huang, Yongzhen", "year": 2025, "venue": "TPAMI", "motivation": "Improve pose-based gait recognition", "idea": "Unified, generalizable, part-aware framework", "techniques": "Human-oriented input, part-aware GCN", "inputs": "Skeleton", "backbone": "GCN, Transformer", "loss": "CE, Triplet", "datasets": "CASIA-B, CCPG, Gait3D, GREW, SUSTech1K, OU-MVLP", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1109/TPAMI.2025.3577594", "bibtex": "@article{meng2025fastposegait,\n title = {From FastPoseGait to GPGait++: Bridging the Past and Future for Pose-based Gait Recognition},\n author = {Meng, Shibei and Fu, Yang and Hou, Saihui and Hu, Xuecai and Cao, Chunshui and Liu, Xu and Huang, Yongzhen},\n year = {2025},\n journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 193 }, { "source": "Author paper summaries", "sheet": "journals", "row": 93 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Benchmarks / Public Codebases", "Advances in Methods / Ideal Gait Modalities" ], "abstract": "Recent studies on pose-based gait recognition have underscored the potential of utilizing such fundamental data to achieve superior outcomes. Nonetheless, the development of current pose-based methods faces significant obstacles due to several critical issues: (1) Misaligned Settings, which results in a lack of thorough and unbiased comparative analysis. (2) Inferior Performance, which causes diminished focus on posebased gait representations. (3) Limited Generalization, which hinders the effective application in real-world scenarios. Focused on tackling the aforementioned challenges, our study introduces a comprehensive benchmark and a versatile approach to bridge the past and future for pose-based gait recognition. First, we revisit previous pose-based methods and make great efforts to establish a unified framework, FastPoseGait, aiming at a fair and comprehensive comparison investigation with consistent experimental settings and a more stable training process. Then, within this framework, we propose GPGait++, a generalized posebased gait recognition method featuring a human-oriented input and part-aware modeling, intended to enhance the generalization ability and discriminative power across diverse environments and camera viewpoints. Experiments on six public gait recognition datasets reveal that our unified framework significantly enhances the performance of previous approaches, and GPGait++ exhibits state-of-the-art cross-domain capabilities compared to existing pose-based methods, marking a significant advancement in the field of pose-based gait recognition. The source code will be accessible at https://github.com/BNU-IVC/FastPoseGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TPAMI.2025.3577594", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "fan2025opengait-tpami-2025", "citationKey": "fan2025opengait", "archiveKey": "fan2025opengait", "method": "OpenGait", "title": "OpenGait: A Comprehensive Benchmark Study for Gait Recognition towards Better Practicality", "authors": "Fan, Chao and Hou, Saihui and Liang, Junhao and Shen, Chuanfu and Ma, Jingzhe and Jin, Dongyang and Huang, Yongzhen and Yu, Shiqi", "year": 2025, "venue": "TPAMI", "motivation": "Improve real-world gait recognition", "idea": "Unified benchmark with robust baselines", "techniques": "Multi-modal inputs, deep networks, contrastive evaluation", "inputs": "Silhouette, Skeleton, SMPL, RGB, Point, Parsing", "backbone": "CNN, Transformer", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, VG, CCPG, SUSTech1K, CCGR, GREW, Gait3D", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1109/TPAMI.2025.3576283", "bibtex": "@article{fan2025opengait,\n title = {OpenGait: A Comprehensive Benchmark Study for Gait Recognition towards Better Practicality},\n author = {Fan, Chao and Hou, Saihui and Liang, Junhao and Shen, Chuanfu and Ma, Jingzhe and Jin, Dongyang and Huang, Yongzhen and Yu, Shiqi},\n year = {2025},\n journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},\n publisher = {IEEE},\n doi = {10.1109/TPAMI.2025.3576283}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 194 }, { "source": "Author paper summaries", "sheet": "journals", "row": 94 } ], "summaryStatus": "Author collection", "sections": [ "Background / Historical Evolution", "Advances in Benchmarks / Public Codebases", "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Strong Gait Backbones" ], "abstract": "Gait recognition, a rapidly advancing vision technology for person identification from a distance, has made significant strides in indoor settings. However, evidence suggests that existing methods often yield unsatisfactory results when applied to newly released real-world gait datasets. Furthermore, conclusions drawn from indoor gait datasets may not easily generalize to outdoor ones. Therefore, the primary goal of this paper is to present a comprehensive benchmark study aimed at improving practicality rather than solely focusing on enhancing performance. To this end, we developed OpenGait, a flexible and efficient gait recognition platform. Using OpenGait, we conducted in-depth ablation experiments to revisit recent developments in gait recognition. Surprisingly, we detected some imperfect parts of some prior methods and thereby uncovered several critical yet previously neglected insights. These findings led us to develop three structurally simple yet empirically powerful and practically robust baseline models: DeepGaitV2, SkeletonGait, and Skeleton- Gait++, which represent the appearance-based, model-based, and multi-modal methodologies for gait pattern description, respectively. In addition to achieving state-of-the-art performance, our careful exploration provides new perspectives on the modeling experience of deep gait models and the representational capacity of typical gait modalities. In the end, we discuss the key trends and challenges in current gait recognition, aiming to inspire further advancements towards better practicality. The code is available at https://github.com/ShiqiYu/OpenGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TPAMI.2025.3576283", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "guo2025gait-tpami-2025", "citationKey": "guo2025gait", "archiveKey": "guo2025gait", "method": "SPOSGait", "title": "Gait recognition in the wild: A large-scale benchmark and NAS-based baseline", "authors": "Guo, Xianda and Zhu, Zheng and Yang, Tian and Lin, Beibei and Huang, Junjie and Deng, Jiankang and Huang, Guan and Zhou, Jie and Lu, Jiwen", "year": 2025, "venue": "TPAMI", "motivation": "Lack of real-world gait benchmarks", "idea": "Build large-scale wild gait dataset", "techniques": "Multi-camera data collection; manual annotation; automated preprocessing; neural architecture search", "inputs": "Silhouette, GEI, Flow, 2D Pose,3D Pose, 3D Mesh", "backbone": "CNN, GCN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, GREW, Gait3D", "codeAvailable": "No", "codeUrl": "", "url": "", "bibtex": "@article{guo2025gait,\n title = {Gait recognition in the wild: A large-scale benchmark and NAS-based baseline},\n author = {Guo, Xianda and Zhu, Zheng and Yang, Tian and Lin, Beibei and Huang, Junjie and Deng, Jiankang and Huang, Guan and Zhou, Jie and Lu, Jiwen},\n year = {2025},\n journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 195 }, { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Benchmarks / Popular Datasets", "Advances in Methods / Strong Gait Backbones", "Challenges and Outlooks / Real-World Gait Recognition" ], "abstract": "Gait benchmarks empower the research community to train and evaluate high-performance gait recognition systems. Even though growing efforts have been devoted to cross-view recognition, academia is restricted by current existing databases captured in the controlled environment. In this paper, we contribute a new benchmark and strong baseline for Gait REcognition in the Wild (GREW). The GREW dataset is constructed from natural videos, which contain hundreds of cameras and thousands of hours of streams in open systems. With tremendous manual annotations, the GREW consists of 26 K identities and 128 K sequences with rich attributes for unconstrained gait recognition. Moreover, we add a distractor set of over 233 K sequences, making it more suitable for real-world applications. Compared with prevailing predefined cross-view datasets, the GREW has diverse and practical view variations, as well as more naturally challenging factors. To the best of our knowledge, this is the first large-scale dataset for gait recognition in the wild. Equipped with this benchmark, we dissect the unconstrained gait recognition problem, where representative appearance-based and model-based methods are explored. The proposed GREW benchmark proves to be essential for both training and evaluating gait recognizers in unconstrained scenarios. In addition, we propose the Single Path One-Shot neural architecture search with uniform sampling for Gait recognition, named SPOSGait, which is the first NAS-based gait recognition model. In experiments, SPOSGait achieves state-of-the-art performance on the CASIA-B, OU-MVLP, Gait3D, and GREW benchmarks, outperforming existing approaches by a large margin.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "aliases": [ "ref-guo2025sposgait" ], "citationAliases": [ "guo2025sposgait" ], "statisticsEligible": true, "venueInScope": true }, { "id": "zhang2025gaitcloud-wacv-2025", "citationKey": "zhang2025gaitcloud", "archiveKey": "zhang2025gaitcloud", "method": "GaitCloud", "title": "Gaitcloud: Leveraging Spatial-Temporal Information for Lidar-Base Gait Recognition With a True-3D Gait Representation", "authors": "Zhang, Shaoxiong and Awano, Hiromitsu and Sato, Takashi", "year": 2025, "venue": "WACV", "motivation": "Overcome 2D limitations, leverage LiDAR", "idea": "Voxelized 3D point cloud representation", "techniques": "Point cloud voxelization, 3D residual encoding, horizontal pooling", "inputs": "Point", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "SUSTech1K", "codeAvailable": "Yes", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/WACV2025/html/Zhang_GaitCloud_Leveraging_Spatial-Temporal_Information_for_LiDAR-Base_Gait_Recognition_with_A_WACV_2025_paper.html", "bibtex": "@inproceedings{zhang2025gaitcloud,\n title = {Gaitcloud: Leveraging Spatial-Temporal Information for Lidar-Base Gait Recognition With a True-3D Gait Representation},\n author = {Zhang, Shaoxiong and Awano, Hiromitsu and Sato, Takashi},\n year = {2025},\n booktitle = {2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},\n pages = {2849--2858}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 94 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait recognition using point clouds captured by LiDAR (Light Detection And Ranging) sensors offers better adaptability to variations in walking conditions compared to camera-based methods, due to the precise spatial information captured. However, existing methods typically project the point clouds into a sequence of 2D depth images extended along the time dimension and adopt gait recognition networks optimized for camera-based approaches. This planar projection compromises the integrity of the 3D coordinates (length, width, and depth) and results in severe silhouette deformations with varied observation viewpoints, similar to the camera-based methods. To better utilize the spatial information in gait point clouds, we propose a true 3D gait representation using efficient point cloud voxelization, termed GaitCloud. Additionally, we explore the unique nature of LiDAR-captured point clouds and present two improved modules adapted to our method, called Layer Encoder (LE) and Horizontal Convolutional Pooling (HCP). Evaluation results using the open-access gait dataset SUSTech1K show that our method outperforms the state-of-the-art, achieving recognition accuracies of 93.1% and 89.2% in cross-view and variance experiments, respectively. These results demonstrate that 3D gait representation based on point cloud voxelization more effectively utilizes spatial information than depth images, offering new possibilities for high-performance LiDAR-based gait recognition. The source code is available at https://github.com/seagrgz/GaitCloud-master.git.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/WACV2025/html/Zhang_GaitCloud_Leveraging_Spatial-Temporal_Information_for_LiDAR-Base_Gait_Recognition_with_A_WACV_2025_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "guo2025gaitcontour-wacv-2025", "citationKey": "guo2025gaitcontour", "archiveKey": "guo2025gaitcontour", "method": "GaitContour", "title": "GaitContour: Efficient Gait Recognition Based on a Contour-Pose Representation", "authors": "Guo, Yuxiang and Shah, Anshul and Liu, Jiang and Gupta, Ayush and Chellappa, Rama and Peng, Cheng", "year": 2025, "venue": "WACV", "motivation": "Balance efficiency and accuracy", "idea": "Dense contour points with regional modeling", "techniques": "Contour extraction, local-global transformer, feature aggregation", "inputs": "Silhouette", "backbone": "CNN, Transformer", "loss": "Triplet", "datasets": "OU-MVLP, GREW, Gait3D, CCPG, SUSTech1K, BRIAR", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/WACV2025/html/Guo_GaitContour_Efficient_Gait_Recognition_Based_on_a_Contour-Pose_Representation_WACV_2025_paper.html", "bibtex": "@inproceedings{guo2025gaitcontour,\n title = {GaitContour: Efficient Gait Recognition Based on a Contour-Pose Representation},\n author = {Guo, Yuxiang and Shah, Anshul and Liu, Jiang and Gupta, Ayush and Chellappa, Rama and Peng, Cheng},\n year = {2025},\n booktitle = {2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},\n pages = {1051--1061}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 91 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait recognition holds the promise to robustly identify subjects based on walking patterns instead of appearance information. In recent years, this field has been dominated by learning methods based on two input formats: silhouette images and sparse keypoints. Compared to imagebased approaches, keypoint-based methods can achieve significantly higher efficiency due to their sparsity. However, sparsity also results in information loss, thereby reducing performance. In this work, we propose a novel, keypointbased Contour-Pose representation, which compactly encodes both body shape and parts information. We further propose a local-to-global architecture, called GaitContour, to leverage this novel representation and efficiently compute subject embedding in two stages. The first stage consists of a local transformer that extracts features from five different body regions. The second stage then aggregates the regional features to estimate a global human gait representation. Such a design significantly reduces the complexity of the attention operation and improves both efficiency and performance. Through large scale experiments, Gait- Contour is shown to perform significantly better than previous keypoint-based methods. Furthermore, the Contour- Pose representation also achieves new SoTA performances on fusion-based gait recognition methods.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/WACV2025/html/Guo_GaitContour_Efficient_Gait_Recognition_Based_on_a_Contour-Pose_Representation_WACV_2025_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "gupta2025mimicgait-wacv-2025", "citationKey": "gupta2025mimicgait", "archiveKey": "gupta2025mimicgait", "method": "MimicGait", "title": "MimicGait: A Model Agnostic approach for Occluded Gait Recognition Using Correlational Knowledge Distillation", "authors": "Gupta, Ayush and Chellappa, Rama", "year": 2025, "venue": "WACV", "motivation": "Robust gait recognition under occlusion", "idea": "Knowledge distillation with visibility estimation", "techniques": "Knowledge distillation, visibility estimation, feature augmentation", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "GREW, Gait3D, BRIAR", "codeAvailable": "Yes", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/WACV2025/html/Gupta_MimicGait_A_Model_Agnostic_Approach_for_Occluded_Gait_Recognition_using_WACV_2025_paper.html", "bibtex": "@inproceedings{gupta2025mimicgait,\n title = {MimicGait: A Model Agnostic approach for Occluded Gait Recognition Using Correlational Knowledge Distillation},\n author = {Gupta, Ayush and Chellappa, Rama},\n year = {2025},\n booktitle = {2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},\n pages = {4757--4766}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 92 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Handling Real-World Noise" ], "abstract": "Gait recognition is an important biometric technique over large distances. State-of-the-art gait recognition systems perform very well in controlled environments at close range. Recently, there has been an increased interest in gait recognition in the wild prompted by the collection of outdoor, more challenging datasets containing variations in terms of illumination, pitch angles and distances. An important problem in these environments is that of occlusion, where the subject is partially blocked from camera view. While important, this problem has received little attention. Thus, we propose MimicGait, a model-agnostic approach for gait recognition in the presence of occlusions. We train the network using a multi-instance correlational distillation loss to capture both inter-sequence and intra-sequence correlations in the occluded gait patterns of a subject, utilizing an auxiliary Visibility Estimation Network to guide the training of the proposed mimic network. We demonstrate the effectiveness of our approach on challenging real-world datasets like GREW, Gait3D and BRIAR. We release the code in https://github.com/Ayush-00/mimicgait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/WACV2025/html/Gupta_MimicGait_A_Model_Agnostic_Approach_for_Occluded_Gait_Recognition_using_WACV_2025_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "wang2025vm-wacv-2025", "citationKey": "wang2025vm", "archiveKey": "wang2025vm", "method": "VM-Gait", "title": "VM-Gait: Multi-Modal 3D Representation Based on Virtual Marker for Gait Recognition", "authors": "Wang, Zhao-Yang and Liu, Jiang and Chen, Jieneng and Chellappa, Rama", "year": 2025, "venue": "WACV", "motivation": "Enhance robustness using 3D SMPL", "idea": "Virtual Marker-based feature fusion", "techniques": "Multi-modal fusion, flexible framework, virtual markers", "inputs": "Silhouette, SMPL", "backbone": "CNN, Transformer", "loss": "CE, Triplet", "datasets": "OU-MVLP, Gait3D, BRIAR", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/WACV2025/html/Wang_VM-Gait_Multi-Modal_3D_Representation_Based_on_Virtual_Marker_for_Gait_WACV_2025_paper.html", "bibtex": "@inproceedings{wang2025vm,\n title = {VM-Gait: Multi-Modal 3D Representation Based on Virtual Marker for Gait Recognition},\n author = {Wang, Zhao-Yang and Liu, Jiang and Chen, Jieneng and Chellappa, Rama},\n year = {2025},\n booktitle = {2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},\n pages = {5326--5335}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 93 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait recognition plays a vital role in biometric applications by analyzing the unique characteristics of an individual’s walking pattern. Methods based on 2D representations, such as silhouettes and skeletons, are increasingly being developed to learn the shape features and joint dynamic movements. Nevertheless, the effectiveness of 2D representation-based methods is impeded by factors such as changes in viewpoint, partial occlusion, and noisy environments. 3D representation-based methods can complement 2D representation-based approaches by providing more precise dynamic body shapes and motion information, along with increased robustness against changes in viewpoint and partial occlusion. However, the complexity of acquiring accurate 3D representations and the challenges associated with extracting dynamic topological features from sequences of 3D representations hinder the development of 3D representations-based methods. In this paper, we present VM-Gait, a novel multi-modal gait recognition framework that harnesses the advantages of integrating both 2D and 3D representations. Furthermore, we introduce a new 3D representation, Virtual Marker, into gait recognition to efficiently learn topological features from 3D representations, avoiding the computational complexities inherent in directly learning from 3D representations like 3D meshes or 3D point clouds. Extensive experiments demonstrate that the proposed framework effectively learns and fuses discriminative information from different gait modalities, enhancing gait recognition performance.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/WACV2025/html/Wang_VM-Gait_Multi-Modal_3D_Representation_Based_on_Virtual_Marker_for_Gait_WACV_2025_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-huang2025vills", "citationKey": "huang2025vills", "archiveKey": "huang2025vills", "method": "", "title": "Vills: Video-image learning to learn semantics for person re-identification", "authors": "Huang, Siyuan and Kathirvel, Ram Prabhakar and Guo, Yuxiang and Chellappa, Rama and Peng, Cheng", "year": 2025, "venue": "WACV", "motivation": "Human recognition needs both local semantics and global video context.", "idea": "Learn complementary person semantics from videos and images.", "techniques": "Video-image representation learning; local-global semantic modeling.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/wacv61041.2025.00582", "bibtex": "@inproceedings{huang2025vills,\n title = {Vills: Video-image learning to learn semantics for person re-identification},\n author = {Huang, Siyuan and Kathirvel, Ram Prabhakar and Guo, Yuxiang and Chellappa, Rama and Peng, Cheng},\n year = {2025},\n booktitle = {Proceedings of the Winter Conference on Applications of Computer Vision},\n pages = {5969--5979}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Advances in Benchmarks / Research Projects", "Challenges and Outlooks / Real-World Gait Recognition" ], "abstract": "VILLS learns person representations jointly from images and videos through self-supervised pretraining. Local semantic extraction seeks consistent body-related features, while a shared learning and adaptation module aligns spatial and temporal representations. The study evaluates transfer across image- and video-based re-identification tasks, including challenging appearance changes.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2311.17074", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-simeoni2025dinov3", "citationKey": "simeoni2025dinov3", "archiveKey": "simeoni2025dinov3", "method": "", "title": "DINOv3", "authors": "Simeoni, Oriane and Vo, Huy V. and Seitzer, Maximilian and Baldassarre, Federico and Oquab, Maxime and others", "year": 2025, "venue": "arXiv", "motivation": "General-purpose dense visual features require scalable self-supervised learning.", "idea": "Scale visual pretraining while maintaining useful dense representations.", "techniques": "Self-supervised vision Transformers; large-scale data curation; dense feature learning.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://arxiv.org/abs/2508.10104", "bibtex": "@misc{simeoni2025dinov3,\n title = {DINOv3},\n author = {Simeoni, Oriane and Vo, Huy V. and Seitzer, Maximilian and Baldassarre, Federico and Oquab, Maxime and others},\n year = {2025},\n url = {https://arxiv.org/abs/2508.10104}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Introduction / Motivation" ], "abstract": "DINOv3 investigates scaling self-supervised visual learning through data preparation, larger models and training refinements. Gram anchoring addresses degradation of dense features during extended training. Additional adaptation steps support varying resolutions, model sizes and text alignment. The released model family is evaluated across diverse visual tasks and resource settings.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2508.10104", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-liu2025person", "citationKey": "liu2025person", "archiveKey": "liu2025person", "method": "FarSight", "title": "Person recognition at altitude and range: Fusion of face, body shape and gait", "authors": "Liu, Feng and Chimitt, Nicholas and Guo, Lanqing and Jain, Jitesh and Kane, Aditya and Kim, Minchul and Robbins, Wes and Su, Yiyang and Ye, Dingqiang and Zhang, Xingguang and others", "year": 2025, "venue": "arXiv", "motivation": "Altitude and range reduce the reliability of individual biometric cues.", "idea": "Combine face, body shape and gait in a whole-body recognition system.", "techniques": "FarSight system; complementary biometric features; multimodal fusion.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@article{liu2025person,\n title = {Person recognition at altitude and range: Fusion of face, body shape and gait},\n author = {Liu, Feng and Chimitt, Nicholas and Guo, Lanqing and Jain, Jitesh and Kane, Aditya and Kim, Minchul and Robbins, Wes and Su, Yiyang and Ye, Dingqiang and Zhang, Xingguang and others},\n year = {2025},\n journal = {arXiv preprint arXiv:2505.04616}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Advances in Benchmarks / Research Projects", "Challenges and Outlooks / Real-World Gait Recognition" ], "abstract": "FarSight integrates face, body-shape and gait cues for recognition at long distances and elevated viewpoints. Its pipeline combines multi-person tracking, recognition-aware restoration, modality-specific encoders and quality-guided fusion. BRIAR experiments evaluate verification and identification under degraded imagery and domain shifts, with additional assessment in the NIST FIVE evaluation.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2505.04616", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "dong2024hybridgait-aaai-2024", "citationKey": "dong2024hybridgait", "archiveKey": "dong2024hybridgait", "method": "HybridGait", "title": "HybridGait: A benchmark for spatial-temporal cloth-changing gait recognition with hybrid explorations", "authors": "Dong, Yilan and Yu, Chunlin and Ha, Ruiyang and Shi, Ye and Ma, Yuexin and Xu, Lan and Fu, Yanwei and Wang, Jingya", "year": 2024, "venue": "AAAI", "motivation": "Robust cloth-changing gait recognition", "idea": "3D mesh alignment and fusion", "techniques": "SMPL-aided hybrid multi-branch network", "inputs": "Silhouette, SMPL", "backbone": "CNN, Transformer", "loss": "CE, Triplet", "datasets": "Gait3D, CCGait", "codeAvailable": "Yes", "codeUrl": "", "url": "", "bibtex": "@inproceedings{dong2024hybridgait,\n title = {HybridGait: A benchmark for spatial-temporal cloth-changing gait recognition with hybrid explorations},\n author = {Dong, Yilan and Yu, Chunlin and Ha, Ruiyang and Shi, Ye and Ma, Yuexin and Xu, Lan and Fu, Yanwei and Wang, Jingya},\n year = {2024},\n booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},\n volume = {38},\n number = {2},\n pages = {1600--1608}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 4 } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Existing gait recognition benchmarks mostly include minor clothing variations in the laboratory environments, but lack persistent changes in appearance over time and space. In this paper, we propose the first in-the-wild benchmark CCGait for cloth-changing gait recognition, which incorporates diverse clothing changes, indoor and outdoor scenes, and multimodal statistics over 92 days. To further address the coupling effect of clothing and viewpoint variations, we propose a hybrid approach HybridGait that exploits both temporal dynamics and the projected 2D information of 3D human meshes. Specifically, we introduce a Canonical Alignment Spatial-Temporal Transformer (CA-STT) module to encode human joint position-aware features, and fully exploit 3D dense priors via a Silhouette-guided Deformation with 3D-2D Appearance Projection (SilD) strategy. Our contributions are twofold: we provide a challenging benchmark CCGait that captures realistic appearance changes across an expanded and space, and we propose a hybrid framework HybridGait that outperforms prior works on CCGait and Gait3D benchmarks. Our project page is available at https://github.com/HCVLab/HybridGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "zou2024cross-aaai-2024", "citationKey": "zou2024cross", "archiveKey": "zou2024cross", "method": "ParsingGait", "title": "Cross-covariate gait recognition: A benchmark", "authors": "Zou, Shinan and Fan, Chao and Xiong, Jianbo and Shen, Chuanfu and Yu, Shiqi and Tang, Jin", "year": 2024, "venue": "AAAI", "motivation": "Address covariate diversity limitations", "idea": "Large-scale dataset with diverse covariates", "techniques": "Parsing-based gait recognition baseline", "inputs": "Parsing", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CCGR", "codeAvailable": "OpenGait", "codeUrl": "", "url": "", "bibtex": "@inproceedings{zou2024cross,\n title = {Cross-covariate gait recognition: A benchmark},\n author = {Zou, Shinan and Fan, Chao and Xiong, Jianbo and Shen, Chuanfu and Yu, Shiqi and Tang, Jin},\n year = {2024},\n booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},\n volume = {38},\n number = {7},\n pages = {7855--7863}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 7 } ], "summaryStatus": "Author collection", "sections": [ "Introduction / Motivation", "Background / Broader Applications", "Advances in Benchmarks / Popular Datasets", "Advances in Methods / Handling Real-World Noise", "Challenges and Outlooks / Real-World Gait Recognition" ], "abstract": "Gait datasets are essential for gait research. However, this paper observes that present benchmarks, whether conventional constrained or emerging real-world datasets, fall short regarding covariate diversity. To bridge this gap, we undertake an arduous 20-month effort to collect a cross-covariate gait recognition (CCGR) dataset. The CCGR dataset has 970 subjects and about 1.6 million sequences; almost every subject has 33 views and 53 different covariates. Compared to existing datasets, CCGR has both population and individual-level diversity. In addition, the views and covariates are well labeled, enabling the analysis of the effects of different factors. CCGR provides multiple types of gait data, including RGB, parsing, silhouette, and pose, offering researchers a comprehensive resource for exploration. In order to delve deeper into addressing cross-covariate gait recognition, we propose parsingbased gait recognition (ParsingGait) by utilizing the newly proposed parsing data. We have conducted extensive experiments. Our main results show: 1) Cross-covariate emerges as a pivotal challenge for practical applications of gait recognition. 2) ParsingGait demonstrates remarkable potential for further advancement. 3) Alarmingly, existing SOTA methods achieve less than 43% accuracy on the CCGR, highlighting the urgency of exploring cross-covariate gait recognition. Link: https://github.com/ShinanZou/CCGR.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "wang2024qagait-aaai-2024", "citationKey": "wang2024qagait", "archiveKey": "wang2024qagait", "method": "QAGait", "title": "Qagait: Revisit gait recognition from a quality perspective", "authors": "Wang, Zengbin and Hou, Saihui and Zhang, Man and Liu, Xu and Cao, Chunshui and Huang, Yongzhen and Li, Peipei and Xu, Shibiao", "year": 2024, "venue": "AAAI", "motivation": "Handle silhouette quality variations", "idea": "Quality assessment and adaptive loss", "techniques": "Quality-aware gait recognition framework", "inputs": "Silhouette", "backbone": "CNN", "loss": "QA-CE, QA-Triplet", "datasets": "CASIA-B, GREW, Gait3D", "codeAvailable": "Yes", "codeUrl": "", "url": "", "bibtex": "@inproceedings{wang2024qagait,\n title = {Qagait: Revisit gait recognition from a quality perspective},\n author = {Wang, Zengbin and Hou, Saihui and Zhang, Man and Liu, Xu and Cao, Chunshui and Huang, Yongzhen and Li, Peipei and Xu, Shibiao},\n year = {2024},\n booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},\n volume = {38},\n number = {6},\n pages = {5785--5793}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 6 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Strong Gait Backbones", "Advances in Methods / Handling Real-World Noise" ], "abstract": "Gait recognition is a promising biometric method that aims to identify pedestrians from their unique walking patterns. Silhouette modality, renowned for its easy acquisition, simple structure, sparse representation, and convenient modeling, has been widely employed in controlled in-the-lab research. However, as gait recognition rapidly advances from in-thelab to in-the-wild scenarios, various conditions raise significant challenges for silhouette modality, including 1) unidentifiable low-quality silhouettes (abnormal segmentation, severe occlusion, or even non-human shape), and 2) identifiable but challenging silhouettes (background noise, non-standard posture, slight occlusion). To address these challenges, we revisit gait recognition pipeline and approach gait recognition from a quality perspective, namely QAGait. Specifically, we propose a series of cost-effective quality assessment strategies, including Maxmial Connect Area and Template Match to eliminate background noises and unidentifiable silhouettes, Alignment strategy to handle non-standard postures. We also propose two quality-aware loss functions to integrate silhouette quality into optimization within the embedding space. Extensive experiments demonstrate our QAGait can guarantee both gait reliability and performance enhancement. Furthermore, our quality assessment strategies can seamlessly integrate with existing gait datasets, showcasing our superiority. Code is available at https://github.com/wzb-bupt/QAGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "fan2024skegait-aaai-2024", "citationKey": "fan2024skeletongait", "archiveKey": "fan2024skegait", "method": "SkeletonGait++", "title": "Skeletongait: Gait recognition using skeleton maps", "authors": "Fan, Chao and Ma, Jingzhe and Jin, Dongyang and Shen, Chuanfu and Yu, Shiqi", "year": 2024, "venue": "AAAI", "motivation": "Robust gait recognition under variations", "idea": "Skeleton map and silhouette fusion", "techniques": "Skeleton-silhouette multi-branch fusion", "inputs": "Silhouette, Skeleton", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "OU-MVLP, GREW, Gait3D, CCPG, SUSTech1K", "codeAvailable": "OpenGait", "codeUrl": "", "url": "", "bibtex": "@inproceedings{fan2024skeletongait,\n title = {Skeletongait: Gait recognition using skeleton maps},\n author = {Fan, Chao and Ma, Jingzhe and Jin, Dongyang and Shen, Chuanfu and Yu, Shiqi},\n year = {2024},\n booktitle = {Proceedings of the AAAI conference on artificial intelligence},\n volume = {38},\n number = {2},\n pages = {1662--1669}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 5 } ], "summaryStatus": "Author collection", "sections": [ "Background / Task Definition", "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Strong Gait Backbones", "Advances in Methods / Handling Real-World Noise" ], "abstract": "The choice of the representations is essential for deep gait recognition methods. The binary silhouettes and skeletal coordinates are two dominant representations in recent literature, achieving remarkable advances in many scenarios. However, inherent challenges remain, in which silhouettes are not always guaranteed in unconstrained scenes, and structural cues have not been fully utilized from skeletons. In this paper, we introduce a novel skeletal gait representation named skeleton map, together with SkeletonGait, a skeleton-based method to exploit structural information from human skeleton maps. Specifically, the skeleton map represents the coordinates of human joints as a heatmap with Gaussian approximation, exhibiting a silhouette-like image devoid of exact body structure. Beyond achieving state-of-the-art performances over five popular gait datasets, more importantly, SkeletonGait uncovers novel insights about how important structural features are in describing gait and when they play a role. Furthermore, we propose a multi-branch architecture, named SkeletonGait++, to make use of complementary features from both skeletons and silhouettes. Experiments indicate that SkeletonGait++ outperforms existing state-of-theart methods by a significant margin in various scenarios. For instance, it achieves an impressive rank-1 accuracy of over 85% on the challenging GREW dataset. The source code is available at https://github.com/ShiqiYu/OpenGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "thapar2024gaitw-accv-2024", "citationKey": "thapar2024gaitw", "archiveKey": "thapar2024gaitw", "method": "GaitW", "title": "GaitW: Enhancing Gait Recognition in the Wild Using Dynamic Information", "authors": "Thapar, Daksh and Chaudhari, Jayesh and Manchanda, Sunny and Nigam, Aditya and Arora, Chetan", "year": 2024, "venue": "ACCV", "motivation": "Improve generalization in-the-wild", "idea": "Dynamic modeling with curriculum learning", "techniques": "Transformer with spatial-temporal encoders", "inputs": "Silhouette", "backbone": "CNN, Transformer", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, GREW, Gait3D", "codeAvailable": "Yes", "codeUrl": "", "url": "", "bibtex": "@inproceedings{thapar2024gaitw,\n title = {GaitW: Enhancing Gait Recognition in the Wild Using Dynamic Information},\n author = {Thapar, Daksh and Chaudhari, Jayesh and Manchanda, Sunny and Nigam, Aditya and Arora, Chetan},\n year = {2024},\n booktitle = {Proceedings of the Asian Conference on Computer Vision},\n pages = {268--285}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 14 } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Successofmoderndeepneuralnetworks( DNNs)forgaitrecognition on in-the-lab datasets such as CASIA-B and OU-MVLP have encouraged the community to aim for more challenging, and in-the-wild datasets such asGREW and Gait3D. The new datasets contain large variations in silhouettes due to change in camera pose, clothing, accessories, as well as occlusion, thus posing huge challenges to existing techniques and training strategies for gait recognition. We posit that to achieve high accuracy in in-the-wild datasets, explicitly leveraging dynamic information in gait samples during training is imperative. We propose a novel transformer based architecture for gait recognition specifically leveraging such dynamic information. The novel contributions include: (1) We propose interleaved spatial and temporal encoders to attend to positioning of various body parts in a frame, and movement of a body part across the sample, respectively. (2) We propose a novel dynamic information inspired curriculum, where we first determine the hardness of a sample based on the disparity between representations of its frame-wise silhouettes (FWSs) and GEI. The model is trained using easier samples first, followed by progressively difficult samples. (3) We propose maskannealing for silhouettes using Gait Energy Images (GEIs), which attends to silhouette contours and allows a model to learn robust silhouette shape representation. We report a significant improvement in accuracy (in %) of 96.9, 92.9, 81.2, and 67.7 on benchmarkCASIA-B, OU-MVLP, GREW, and Gait3D datasets respectively using our technique, against the current state-of-the-art (SOTA) accuracy of 96.9, 92.4, 77.4, and 67.0 by MSGR [43], HSTGait [41], SkeletonGait [13], andQAGait [45] respectively. In a significant departure from the current trend, and as evident from the above numbers, the proposed technique sets up simultaneousSOTA on most prominentin-the-lab as well as in-the-wild datasets. Complete source code and trained models of our method will be publicly available.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "li2024aerialgait-acmmm-2024", "citationKey": "li2024aerialgait", "archiveKey": "li2024aerialgait", "method": "AGGNet", "title": "AerialGait: Bridging aerial and ground views for gait recognition", "authors": "Li, Aoqi and Hou, Saihui and Wang, Chenye and Cai, Qingyuan and Huang, Yongzhen", "year": 2024, "venue": "ACM MM", "motivation": "Enable aerial-ground gait recognition", "idea": "Dataset creation and AGG-Net", "techniques": "Large-scale dataset; uncertainty learning; prototype-based domain alignment", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "AerialGait, DroneGait", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1145/nnnnnnn.nnnnnnn", "bibtex": "@inproceedings{li2024aerialgait,\n title = {AerialGait: Bridging aerial and ground views for gait recognition},\n author = {Li, Aoqi and Hou, Saihui and Wang, Chenye and Cai, Qingyuan and Huang, Yongzhen},\n year = {2024},\n booktitle = {Proceedings of the 32nd ACM International Conference on Multimedia},\n pages = {1139--1147}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 80 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Background / Broader Applications", "Advances in Methods / Handling Real-World Noise" ], "abstract": "In this work, we present AerialGait, a comprehensive dataset for aerial-ground gait recognition. This dataset comprises 82,454 sequences totaling over 10 million frames from 533 subjects, captured from both aerial and ground perspectives. To align with real-life scenarios of aerial and ground surveillance, we utilize a drone and a ground surveillance camera for data acquisition. The drone is operated at various speeds, directions, and altitudes. Meanwhile, we conduct data collection across five diverse surveillance sites to ensure a comprehensive simulation of real-world settings. AerialGait has several unique features: 1) The gait sequences exhibit significant variations in views, resolutions, and illumination across five distinct scenes. 2) It incorporates challenges of motion blur and frame discontinuity due to drone mobility. 3) The dataset reflects the domain gap caused by the view disparity between aerial and ground views, presenting a realistic challenge for drone-based gait recognition. Moreover, we perform a comprehensive analysis of existing gait recognition methods on AerialGait dataset and propose the Aerial-Ground Gait Network (AGG-Net). AGG-Net effectively learns discriminative features from aerial views by uncertainty learning and clusters features across aerial and ground views through prototype learning. Our model achieves state-of-theart performance on both AerialGait and DroneGait datasets. The dataset and code will be made available upon acceptance.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1145/nnnnnnn.nnnnnnn", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "peng2024glgait-acmmm-2024", "citationKey": "peng2024glgait", "archiveKey": "peng2024glgait", "method": "GLGait", "title": "Glgait: a global-local temporal receptive field network for gait recognition in the wild", "authors": "Peng, Guozhen and Wang, Yunhong and Zhao, Yuwei and Zhang, Shaoxiong and Li, Annan", "year": 2024, "venue": "ACM MM", "motivation": "Improve gait recognition in-the-wild", "idea": "Global-local temporal modeling", "techniques": "Attention-enhanced temporal convolution, center-aware loss", "inputs": "Silhouette", "backbone": "CNN, Attention", "loss": "CE, Center-Aug Triplet", "datasets": "GREW, Gait3D", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1145/3664647.3680812", "bibtex": "@inproceedings{peng2024glgait,\n title = {Glgait: a global-local temporal receptive field network for gait recognition in the wild},\n author = {Peng, Guozhen and Wang, Yunhong and Zhao, Yuwei and Zhang, Shaoxiong and Li, Annan},\n year = {2024},\n booktitle = {Proceedings of the 32nd ACM International Conference on Multimedia},\n pages = {826--835}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 81 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait recognition has attracted increasing attention from academia and industry as a human recognition technology from a distance in non-intrusive ways without requiring cooperation. Although advanced methods have achieved impressive success in lab scenarios, most of them perform poorly in the wild. Recently, some Convolution Neural Networks (ConvNets) based methods have been proposed to address the issue of gait recognition in the wild. However, the temporal receptive field obtained by convolution operations is limited for long gait sequences. If directly replacing convolution blocks with visual transformer blocks, the model may not enhance a local temporal receptive field, which is important for covering a complete gait cycle. To address this issue, we design a Global-Local Temporal Receptive Field Network (GLGait). GLGait employs a Global-Local Temporal Module (GLTM) to establish a global-local temporal receptive field, which mainly consists of a Pseudo Global Temporal Self-Attention (PGTA) and a temporal convolution operation. Specifically, PGTA is used to obtain a pseudo global temporal receptive field with less memory and computation complexity compared with a multi-head self-attention (MHSA). The temporal convolution operation is used to enhance the local temporal receptive field. Besides, it can also aggregate pseudo global temporal receptive field to a true holistic temporal receptive field. Furthermore, we also propose a Center-Augmented Triplet Loss (CTL) in GLGait to reduce the intra-class distance and expand the positive samples in the training stage. Extensive experiments show that our method obtains state-of-the-art results on in-the-wild datasets, i.e. , Gait3D and GREW. The code is available at https://github.com/bgdpgz/GLGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1145/3664647.3680812", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "han2024gait-acmmm-2024", "citationKey": "han2024gait", "archiveKey": "han2024gait", "method": "HMRNet", "title": "Gait recognition in large-scale free environment via single lidar", "authors": "Han, Xiao and Ren, Yiming and Cong, Peishan and Sun, Yujing and Wang, Jingya and Xu, Lan and Ma, Yuexin", "year": 2024, "venue": "ACM MM", "motivation": "Enhance LiDAR-based gait recognition", "idea": "Fuse LiDAR range and point cloud", "techniques": "Multi-modal feature extraction, cross-representation fusion, attention enhancement", "inputs": "Point", "backbone": "CNN, MLP", "loss": "CE, Triplet", "datasets": "SUSTech1K, FreeGait", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1145/3664647.3681166", "bibtex": "@inproceedings{han2024gait,\n title = {Gait recognition in large-scale free environment via single lidar},\n author = {Han, Xiao and Ren, Yiming and Cong, Peishan and Sun, Yujing and Wang, Jingya and Xu, Lan and Ma, Yuexin},\n year = {2024},\n booktitle = {Proceedings of the 32nd ACM International Conference on Multimedia},\n pages = {380--389}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 83 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Human gait recognition is crucial in multimedia, enabling identification through walking patterns without direct interaction, enhancing the integration across various media forms in real-world applications like smart homes, healthcare and non-intrusive security. LiDAR’s ability to capture depth makes it pivotal for robotic perception and holds promise for real-world gait recognition. In this paper, based on a single LiDAR, we present the Hierarchical Multirepresentation Feature Interaction Network (HMRNet) for robust gait recognition. Prevailing LiDAR-based gait datasets primarily derive from controlled settings with predefined trajectory, remaining a gap with real-world scenarios. To facilitate LiDAR-based gait recognition research, we introduce FreeGait, a comprehensive gait dataset from large-scale, unconstrained settings, enriched with multi-modal and varied 2D/3D data. Notably, our approach achieves state-of-the-art performance on prior dataset (SUSTech1K) and on FreeGait. https://4dvlab.github.io/project_page/FreeGait.html", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1145/3664647.3681166", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "zheng2024takes-acmmm-2024", "citationKey": "zheng2024takes", "archiveKey": "zheng2024takes", "method": "Xgait", "title": "It takes two: Accurate gait recognition in the wild via cross-granularity alignment", "authors": "Zheng, Jinkai and Liu, Xinchen and Zhang, Boyue and Yan, Chenggang and Zhang, Jiyong and Liu, Wu and Zhang, Yongdong", "year": 2024, "venue": "ACM MM", "motivation": "Improve gait recognition robustness", "idea": "Parsing and Silhouette Fusion", "techniques": "Feature extraction, cross-level alignment, occlusion handling", "inputs": "Silhouette, Parsing", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "Gait3D, CCPG", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1145/3664647.3680577", "bibtex": "@inproceedings{zheng2024takes,\n title = {It takes two: Accurate gait recognition in the wild via cross-granularity alignment},\n author = {Zheng, Jinkai and Liu, Xinchen and Zhang, Boyue and Yan, Chenggang and Zhang, Jiyong and Liu, Wu and Zhang, Yongdong},\n year = {2024},\n booktitle = {Proceedings of the 32nd ACM International Conference on Multimedia},\n pages = {8786--8794}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 82 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Handling Real-World Noise" ], "abstract": "Existing studies for gait recognition primarily utilized sequences of either binary silhouette or human parsing to encode the shapes and dynamics of persons during walking. Silhouettes exhibit accurate segmentation quality and robustness to environmental variations, but their low information entropy may result in sub-optimal performance. In contrast, human parsing provides fine-grained part segmentation with higher information entropy, but the segmentation quality may deteriorate due to the complex environments. To discover the advantages of silhouette and parsing and overcome their limitations, this paper proposes a novel cross-granularity alignment gait recognition method, named XGait, to unleash the power of gait representations of different granularity. To achieve this goal, the XGait first contains two branches of backbone encoders to map the silhouette sequences and the parsing sequences into two latent spaces, respectively. Moreover, to explore the complementary knowledge across the features of two representations, we design the Global Cross-granularity Module (GCM) and the Part Cross-granularity Module (PCM) after the two encoders. In particular, the GCM aims to enhance the quality of parsing features by leveraging global features from silhouettes, while the PCM aligns the dynamics of human parts between silhouette and parsing features using the high information entropy in parsing sequences. In addition, to effectively guide the alignment of two representations with different granularity at the part level, an elaborate-designed learnable division mechanism is proposed for the parsing features. Finally, comprehensive experiments on two large-scale gait datasets not only show the superior performance of XGait with the Rank-1 accuracy of 80.5% on Gait3D and 88.3% CCPG but also reflect the robustness of the learned features even under challenging conditions like occlusions and cloth changes.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1145/3664647.3680577", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ye2024biggait-cvpr-2024", "citationKey": "ye2024biggait", "archiveKey": "ye2024biggait", "method": "BigGait", "title": "Biggait: Learning gait representation you want by large vision models", "authors": "Ye, Dingqiang and Fan, Chao and Ma, Jingzhe and Liu, Xiaoming and Yu, Shiqi", "year": 2024, "venue": "CVPR", "motivation": "Handcrafted modality conversion can discard useful gait information.", "idea": "Extract gait representations from pretrained large vision models.", "techniques": "Large-model features; gait representation extractor; soft constraints.", "inputs": "RGB", "backbone": "Transformer", "loss": "CE, Triplet", "datasets": "CASIA-B, CCPG, SUSTech1K", "codeAvailable": "OpenGait", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2024/html/Ye_BigGait_Learning_Gait_Representation_You_Want_by_Large_Vision_Models_CVPR_2024_paper.html", "bibtex": "@inproceedings{ye2024biggait,\n title = {Biggait: Learning gait representation you want by large vision models},\n author = {Ye, Dingqiang and Fan, Chao and Ma, Jingzhe and Liu, Xiaoming and Yu, Shiqi},\n year = {2024},\n booktitle = {Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},\n pages = {200--210}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 30 }, { "source": "Manuscript Table I" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Introduction / Motivation", "Background / Task Definition", "Background / Historical Evolution", "Advances in Benchmarks / State-of-the-Art Performance", "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Gait Learning Frameworks" ], "abstract": "Gait recognition stands as one of the most pivotal remote identification technologies and progressively expands across research and industry communities. However, existing gait recognition methods heavily rely on task-specific upstream driven by supervised learning to provide explicit gait representations like silhouette sequences, which inevitably introduce expensive annotation costs and potential error accumulation. Escaping from this trend, this work explores effective gait representations based on the all-purpose knowledge produced by task-agnostic Large Vision Models (LVMs) and proposes a simple yet efficient gait framework, termed BigGait. Specifically, the Gait Representation Extractor (GRE) within BigGait draws upon design principles from established gait representations, effectively transforming all-purpose knowledge into implicit gait representations without requiring third-party supervision signals. Experiments on CCPG, CAISA-B* and SUSTech1K indicate that BigGait significantly outperforms the previous methods in both within-domain and cross-domain tasks in most cases, and provides a more practical paradigm for learning the next-generation gait representation. Finally, we delve into prospective challenges and promising directions in LVMs-based gait recognition, aiming to inspire future work in this emerging topic. The source code is available at https:// github.com/ ShiqiYu/ OpenGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2024/html/Ye_BigGait_Learning_Gait_Representation_You_Want_by_Large_Vision_Models_CVPR_2024_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ma2024learning-cvpr-2024", "citationKey": "ma2024learning", "archiveKey": "ma2024learning", "method": "VPNet", "title": "Learning visual prompt for gait recognition", "authors": "Ma, Kang and Fu, Ying and Cao, Chunshui and Hou, Saihui and Huang, Yongzhen and Zheng, Dezhi", "year": 2024, "venue": "CVPR", "motivation": "Adaptive visual prompt", "idea": "Robust gait recognition invariances", "techniques": "Prompts with dynamic attention", "inputs": "Silhouette", "backbone": "CNN, Transformer", "loss": "CE, Triplet, Cosine Loss", "datasets": "CASIA-B, OU-MVLP, Gait3D", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2024/html/Ma_Learning_Visual_Prompt_for_Gait_Recognition_CVPR_2024_paper.html", "bibtex": "@inproceedings{ma2024learning,\n title = {Learning visual prompt for gait recognition},\n author = {Ma, Kang and Fu, Ying and Cao, Chunshui and Hou, Saihui and Huang, Yongzhen and Zheng, Dezhi},\n year = {2024},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n pages = {593--603}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 29 }, { "source": "Manuscript Table I" }, { "source": "Verified publication metadata" } ], "summaryStatus": "Manuscript summary", "sections": [ "Background / Historical Evolution" ], "abstract": "Gait, a prevalent and complex form of human motion, plays a significant role in the field of long-range pedestrian retrieval due to the unique characteristics inherent in individual motion patterns. However, gait recognition in real-world scenarios is challenging due to the limitations of capturing comprehensive cross-viewing and crossclothing data. Additionally, distractors such as occlusions, directional changes, and lingering movements further complicate the problem. The widespread application of deep learning techniques has led to the development of various potential gait recognition methods. However, these methods utilize convolutional networks to extract shared information across different views and attire conditions. Once trained, the parameters and non-linear function become constrained to fixed patterns, limiting their adaptability to various distractors in real-world scenarios. In this paper, we present a unified gait recognition framework to extract global motion patterns and develop a novel dynamic transformer to generate representative gait features. Specifically, we develop a trainable part-based prompt pool with numerous key-value pairs that can dynamically select prompt templates to incorporate into the gait sequence, thereby providing task-relevant shared knowledge information. Furthermore, we specifically design dynamic attention to extract robust motion patterns and address the length generalization issue. Extensive experiments on four widely recognized gait datasets, i.e., Gait3D, GREW, OUMVLP , and CASIA-B, reveal that the proposed method yields substantial improvements compared to current state-of-the-art approaches.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2024/html/Ma_Learning_Visual_Prompt_for_Gait_Recognition_CVPR_2024_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-li2024finerehab", "citationKey": "li2024finerehab", "archiveKey": "li2024finerehab", "method": "", "title": "FineRehab: A Multi-modality and Multi-task Dataset for Rehabilitation Analysis", "authors": "Li, Jianwei and Xue, Jun and Cao, Rui and Du, Xiaoxia and Mo, Siyu and Ran, Kehao and Zhang, Zeyan", "year": 2024, "venue": "CVPR Workshops", "motivation": "Rehabilitation analysis requires detailed movement-quality annotations across sensors.", "idea": "Provide a multimodal resource for exercise and quality assessment.", "techniques": "Visual and inertial sensing; exercise recognition; multidimensional quality labels.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2024W/CVsports/html/Li_FineRehab_A_Multi-modality_and_Multi-task_Dataset_for_Rehabilitation_Analysis_CVPRW_2024_paper.html", "bibtex": "@inproceedings{li2024finerehab,\n title = {FineRehab: A Multi-modality and Multi-task Dataset for Rehabilitation Analysis},\n author = {Li, Jianwei and Xue, Jun and Cao, Rui and Du, Xiaoxia and Mo, Siyu and Ran, Kehao and Zhang, Zeyan},\n year = {2024},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},\n pages = {3184--3193},\n url = {https://openaccess.thecvf.com/content/CVPR2024W/CVsports/html/Li_FineRehab_A_Multi-modality_and_Multi-task_Dataset_for_Rehabilitation_Analysis_CVPRW_2024_paper.html}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Challenges and Outlooks / Gait Representation Learning for Healthcare" ], "abstract": "FineRehab supports analysis of rehabilitation movements using synchronized visual, depth and inertial recordings. It contains sixteen actions performed by fifty participants, including patients and healthy volunteers. The paper presents baselines for multidimensional movement-quality assessment and investigates how multimodal measurements can support more quantitative exercise feedback.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2024W/CVsports/html/Li_FineRehab_A_Multi-modality_and_Multi-task_Dataset_for_Rehabilitation_Analysis_CVPRW_2024_paper.html", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "guo2024camera-eccv-2024", "citationKey": "guo2024camera", "archiveKey": "guo2024camera", "method": "CL-Gait", "title": "Camera-lidar cross-modality gait recognition", "authors": "Guo, Wenxuan and Liang, Yingping and Pan, Zhiyu and Xi, Ziheng and Feng, Jianjiang and Zhou, Jie", "year": 2024, "venue": "ECCV", "motivation": "Overcome sensor limitations in varied environments", "idea": "Contrastive depth-based pre-training", "techniques": "Cross-modality camera-LiDAR gait recognition", "inputs": "Point, Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "SUSTech1K, LIP, HITSZ-VCM", "codeAvailable": "Yes", "codeUrl": "", "url": "", "bibtex": "@inproceedings{guo2024camera,\n title = {Camera-lidar cross-modality gait recognition},\n author = {Guo, Wenxuan and Liang, Yingping and Pan, Zhiyu and Xi, Ziheng and Feng, Jianjiang and Zhou, Jie},\n year = {2024},\n booktitle = {European Conference on Computer Vision},\n pages = {439--455}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 42 }, { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Ideal Gait Modalities" ], "abstract": "Gait recognition is a crucial biometric identification technique. Camera-based gait recognition has been widely applied in both research and industrial fields. LiDAR-based gait recognition has also begun to evolve most recently, due to the provision of 3D structural information. However, in certain applications, cameras fail to recognize persons, such as in low-light environments and long-distance recognition scenarios, where LiDARs work well. On the other hand, the deployment cost and complexity of LiDAR systems limit its wider application. Therefore, it is essential to consider cross-modality gait recognition between cameras and LiDARs for a broader range of applications. In this work, we propose the first cross-modality gait recognition framework between Camera and LiDAR, namely CL-Gait. It employs a two-stream network for feature embedding of both modalities. This poses a challenging recognition task due to the inherent matching between 3D and 2D data, exhibiting significant modality discrepancy. To align the feature spaces of the two modalities, i.e., camera silhouettes and LiDAR points, we propose a contrastive pre-training strategy to mitigate modality discrepancy. To make up for the absence of paired camera-LiDAR data for pre-training, we also introduce a strategy for generating data on a large scale. This strategy utilizes monocular depth estimated from single RGB images and virtual cameras to generate pseudo point clouds for contrastive pre-training. Extensive experiments show that the cross-modality gait recognition is very challenging but still contains potential and feasibility with our proposed model and pre-training strategy. To the best of our knowledge, this is the first work to address cross-modality gait recognition. The code and dataset are available at https://github.com/GWxuan/CL-Gait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "aliases": [ "ref-guo2024clgait" ], "citationAliases": [ "guo2024clgait" ], "statisticsEligible": true, "venueInScope": true }, { "id": "xiong2024causality-eccv-2024", "citationKey": "xiong2024causality", "archiveKey": "xiong2024causality", "method": "CLTD", "title": "Causality-inspired discriminative feature learning in triple domains for gait recognition", "authors": "Xiong, Haijun and Feng, Bin and Wang, Xinggang and Liu, Wenyu", "year": 2024, "venue": "ECCV", "motivation": "Reduce identity-confounding noise impact", "idea": "Spatial-temporal-spectral causal disentanglement", "techniques": "Multi-domain confounder mitigation module", "inputs": "Silhouette", "backbone": "CNN, Transformer", "loss": "CE, Triplet, NCE", "datasets": "CASIA-B, OU-MVLP, GREW, Gait3D", "codeAvailable": "No", "codeUrl": "", "url": "", "bibtex": "@inproceedings{xiong2024causality,\n title = {Causality-inspired discriminative feature learning in triple domains for gait recognition},\n author = {Xiong, Haijun and Feng, Bin and Wang, Xinggang and Liu, Wenyu},\n year = {2024},\n booktitle = {European Conference on Computer Vision},\n pages = {251--270}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 39 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Gait Learning Frameworks" ], "abstract": "Gait recognition is a biometric technology that distinguishes individuals by their walking patterns. However, previous methods face challenges when accurately extracting identity features because they often become entangled with non-identity clues. To address this challenge, we propose CLTD, a causality-inspired discriminative feature learning module designed to effectively eliminate the influence of confounders in triple domains, i.e., spatial, temporal, and spectral. Specifically, we utilize the Cross Pixel-wise Attention Generator (CPAG) to generate attention distributions for factual and counterfactual features in spatial and temporal domains. Then, we introduce the Fourier Projection Head (FPH) to project spatial features into the spectral space, which preserves essential information while reducing computational costs. Additionally, we employ an optimization method with contrastive learning to enforce semantic consistency constraints across sequences from the same subject. Our approach has demonstrated significant performance improvements on challenging datasets, proving its effectiveness. Moreover, it can be seamlessly integrated into existing gait recognition methods.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "fu2024cut-eccv-2024", "citationKey": "fu2024cut", "archiveKey": "fu2024cut", "method": "GaitHeat++", "title": "Cut out the middleman: Revisiting pose-based gait recognition", "authors": "Fu, Yang and Hou, Saihui and Meng, Shibei and Hu, Xuecai and Cao, Chunshui and Liu, Xu and Huang, Yongzhen", "year": 2024, "venue": "ECCV", "motivation": "Improved pose-based gait recognition", "idea": "Direct pose heatmap representation", "techniques": "Pose Heatmap-based Global-local Network", "inputs": "RGB", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, CCPG, SUSTech1K", "codeAvailable": "FastPoseGait", "codeUrl": "", "url": "", "bibtex": "@inproceedings{fu2024cut,\n title = {Cut out the middleman: Revisiting pose-based gait recognition},\n author = {Fu, Yang and Hou, Saihui and Meng, Shibei and Hu, Xuecai and Cao, Chunshui and Liu, Xu and Huang, Yongzhen},\n year = {2024},\n booktitle = {European Conference on Computer Vision},\n pages = {112--128}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 41 } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Recent pose-based gait recognition methods, which utilize human skeletons as the model input, have demonstrated significant potential in handling variations in clothing and occlusions. However, methods relying on such skeleton to encode pose are constrained mainly by two problems: (1) poor performance caused by the shape loss, and (2) lack of generalizability. Addressing these limitations, we revisit pose-based gait recognition and develop GaitHeat, a heatmap-based framework that largely enhances performance and robustness by utilizing a new modality to encode pose rather than keypoint coordinates. We make our efforts from two aspects, the pipeline and the extraction of multi-channel heatmap features. Specifically, the process of resizing and centering is performed in the RGB space to largely preserve the integrity of heatmap information. To boost the generalization across various datasets further, we propose a pose-guided heatmap alignment module to eliminate the influence of gait-irrelevant covariates. Furthermore, a global-local network incorporating an efficient fusion branch is designed to improve the extraction of semantic information. Compared to skeleton-based methods, GaitHeat exhibits superior performance in learning gait features and demonstrates effective generalization across different datasets. Experiments on three datasets reveal that our proposed method achieves state-of-the-art results for pose-based gait recognition, comparable to that of silhouette-based approaches. All the source code is available at https://github.com/BNU-IVC/FastPoseGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "huang2024occluded-eccv-2024", "citationKey": "huang2024occluded", "archiveKey": "huang2024occluded", "method": "GaitMoE", "title": "Occluded gait recognition with mixture of experts: an action detection perspective", "authors": "Huang, Panjian and Peng, Yunjie and Hou, Saihui and Cao, Chunshui and Liu, Xu and He, Zhiqiang and Huang, Yongzhen", "year": 2024, "venue": "ECCV", "motivation": "Overcome occlusion in gait recognition", "idea": "Action detection and expert modeling", "techniques": "Action-based Mixture of Experts", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "GREW, Gait3D, OccCASIA-B, OccGait", "codeAvailable": "Yes", "codeUrl": "", "url": "", "bibtex": "@inproceedings{huang2024occluded,\n title = {Occluded gait recognition with mixture of experts: an action detection perspective},\n author = {Huang, Panjian and Peng, Yunjie and Hou, Saihui and Cao, Chunshui and Liu, Xu and He, Zhiqiang and Huang, Yongzhen},\n year = {2024},\n booktitle = {European Conference on Computer Vision},\n pages = {380--397}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 40 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Handling Real-World Noise", "Challenges and Outlooks / Real-World Gait Recognition" ], "abstract": "Extensive occlusions in real-world scenarios pose challenges to gait recognition due to missing and noisy information, as well as body misalignment in position and scale. We argue that rich dynamic contextual information within a gait sequence inherently possesses occlusionsolving traits: 1) Adjacent frames with gait continuity allow holistic body regions to infer occluded body regions; 2) Gait cycles allow information integration between holistic actions and occluded actions. Therefore, we introduce an action detection perspective where a gait sequence is regarded as a composition of actions. To detect accurate actions under complex occlusion scenarios, we propose an Action Detection Based Mixture of Experts (GaitMoE), consisting of Mixture of Temporal Experts (MTE) and Mixture of Action Experts (MAE). MTE adaptively constructs action anchors by temporal experts and MAE adaptively constructs action proposals from action anchors by action experts. Especially, action detection as a proxy task with gait recognition is an endto-end joint training only with ID labels. In addition, due to the lack of a unified occluded benchmark, we construct a pioneering Occluded Gait database (OccGait), containing rich occlusion scenarios and annotations of occlusion types. Extensive experiments on OccGait, OccCASIA-B, Gait3D and GREW demonstrate the superior performance of GaitMoE. OccGait is available at https://github.com/BNU-IVC/OccGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-su2024open", "citationKey": "su2024open", "archiveKey": "su2024open", "method": "", "title": "Open-set biometrics: Beyond good closed-set models", "authors": "Su, Yiyang and Kim, Minchul and Liu, Feng and Jain, Anil and Liu, Xiaoming", "year": 2024, "venue": "ECCV", "motivation": "Strong closed-set classification does not ensure reliable rejection of unknown identities.", "idea": "Evaluate and improve biometric recognition under open-set protocols.", "techniques": "Unknown-identity rejection; score analysis; open-set identification evaluation.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1007/978-3-031-73033-7_14", "bibtex": "@inproceedings{su2024open,\n title = {Open-set biometrics: Beyond good closed-set models},\n author = {Su, Yiyang and Kim, Minchul and Liu, Feng and Jain, Anil and Liu, Xiaoming},\n year = {2024},\n booktitle = {European Conference on Computer Vision},\n pages = {243--261}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Task Definition", "Background / Historical Evolution", "Advances in Benchmarks / State-of-the-Art Performance", "Advances in Methods / Gait Metric Designs", "Challenges and Outlooks / Real-World Gait Recognition" ], "abstract": "This study argues that training objectives for closed-set recognition do not directly optimize rejection of unenrolled people. It simulates open-set decisions within training minibatches and introduces losses for identification-detection performance and negative-score thresholds. Evaluations across face, gait and person re-identification assess improvements to open-set recognition while tracking closed-set performance.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2407.16133", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "wang2024free-eccv-2024", "citationKey": "wang2024free", "archiveKey": "wang2024free", "method": "RelationDescriptor", "title": "Free lunch for gait recognition: A novel relation descriptor", "authors": "Wang, Jilong and Hou, Saihui and Huang, Yan and Cao, Chunshui and Liu, Xu and Huang, Yongzhen and Zhang, Tianzhu and Wang, Liang", "year": 2024, "venue": "ECCV", "motivation": "Enhance gait feature robustness", "idea": "Gait Anchor selection and regularization", "techniques": "Relation Descriptor using classifiers", "inputs": "Silhouette", "backbone": "", "loss": "CE, Triplet, Regularization", "datasets": "CASIA-B, OU-MVLP, GREW, Gait3D, CCPG", "codeAvailable": "No", "codeUrl": "", "url": "", "bibtex": "@inproceedings{wang2024free,\n title = {Free lunch for gait recognition: A novel relation descriptor},\n author = {Wang, Jilong and Hou, Saihui and Huang, Yan and Cao, Chunshui and Liu, Xu and Huang, Yongzhen and Zhang, Tianzhu and Wang, Liang},\n year = {2024},\n booktitle = {European Conference on Computer Vision},\n pages = {39--56}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 43 } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait recognition is to seek correct matches for query individuals by their unique walking patterns. However, current methods focus solely on extracting individual-specific features, overlooking “interpersonal\" relationships. In this paper, we propose a novelRelation Descriptor that captures not only individual features but also relations between test gaits and pre-selected gait anchors. Specifically, we reinterpret classifier weights as gait anchors and compute similarity scores between test features and these anchors, which re-expresses individual gait features into a similarity relation distribution. In essence, the relation descriptor offers a holistic perspective that leverages the collective knowledge stored within the classifier’s weights, emphasizing meaningful patterns and enhancing robustness. Despite its potential, relation descriptor poses dimensionality challenges since its dimension depends on the training set’s identity count. To address this, we propose Farthest gait-Anchor Selection to identify the most discriminative gait anchors and an Orthogonal Regularization Loss to increase diversity within gait anchors. Compared to individual-specific features extracted from the backbone, our relation descriptor can boost the performance nearly without any extra costs. We evaluate the effectiveness of our method on the popular GREW, Gait3D, OU-MVLP, CASIA-B, and CCPG, showing that our method consistently outperforms the baselines and achieves state-of-the-art performance.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-wu2024gstnet", "citationKey": "wu2024gstnet", "archiveKey": "wu2024gstnet", "method": "", "title": "GSTNet: Gait Spatio-Temporal Network for Gait Recognition Using Millimeter-Wave Radar", "authors": "Wu, Qiuxia and Wang, Zicheng and Su, Kunming and Xu, Sangni", "year": 2024, "venue": "ICASSP", "motivation": "Radar gait signals contain coupled spatial and temporal identity cues.", "idea": "Model those dependencies within a gait-specific network.", "techniques": "Millimeter-wave radar; spatiotemporal feature extraction; identity learning.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@inproceedings{wu2024gstnet,\n title = {GSTNet: Gait Spatio-Temporal Network for Gait Recognition Using Millimeter-Wave Radar},\n author = {Wu, Qiuxia and Wang, Zicheng and Su, Kunming and Xu, Sangni},\n year = {2024},\n booktitle = {ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},\n pages = {4855--4859}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Task Definition" ], "abstract": "GSTNet addresses gait recognition from sparse millimeter-wave radar point clouds captured along unconstrained walking routes. A graph module aggregates observations at multiple temporal resolutions, while a dynamic feature module weights temporal evidence. The approach is evaluated on radar gait benchmarks to study how spatial relations and motion can complement each other under sparse observations.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://doi.org/10.1109/ICASSP48485.2024.10447288", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "okimura2024cropping-ijcb-2024", "citationKey": "okimura2024cropping", "archiveKey": "okimura2024cropping", "method": "CVL", "title": "On Cropping for Gait Recognition: Does Constant-velocity Locomotion Assumption Improve Gait Recognition Accuracy?", "authors": "Okimura, Tappei and Li, Xiang and Xu, Chi and Yagi, Yasushi", "year": 2024, "venue": "IJCB", "motivation": "Improve gait recognition cropping consistency", "idea": "Mitigate scale variation in cropping", "techniques": "Constant-velocity modeling, 3D projection", "inputs": "RGB, SMPL, Silhouette", "backbone": "", "loss": "", "datasets": "OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/IJCB62174.2024.10744433", "bibtex": "@inproceedings{okimura2024cropping,\n title = {On Cropping for Gait Recognition: Does Constant-velocity Locomotion Assumption Improve Gait Recognition Accuracy?},\n author = {Okimura, Tappei and Li, Xiang and Xu, Chi and Yagi, Yasushi},\n year = {2024},\n booktitle = {2024 IEEE International Joint Conference on Biometrics (IJCB)},\n pages = {1--12}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 71 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Ideal Gait Modalities" ], "abstract": "Most of gait recognition studies focus on feature extraction and matching steps by using cropped image sequences as inputs. We usually use frame-by-frame tight bounding boxes (TBBs) obtained by pedestrian detection or instance segmentation for cropping. Cropping by the TBB, however, suffers from apparent scale changes within gait period (e.g., apparent height changes between single/double support phases), which may cause a drop in gait recognition accuracy. We therefore propose a method of cropping for gait recognition to better preserve the apparent scale by introducing constant-velocity locomotion (CVL) assumption for a short period (e.g., one second). We derive that a bounding box sequence (BBS) in the 2D image coordinate under CVL assumption in the 3D camera coordinate, is represented by non-linear interpolation between the starting and ending frames without camera calibration parameters. We then estimate BBS parameters (i.e., bounding boxes at the starting and ending frames) by generalized Hough transform with voting from pedestrian region proposals. Experiments with OU-MVLP show that the proposed cropping improves the gait recognition accuracies for both model-based and appearance-based approaches.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/IJCB62174.2024.10744433", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "guo2024distillation-ijcb-2024", "citationKey": "guo2024distillation", "archiveKey": "guo2024distillation", "method": "GADER", "title": "Distillation-guided representation learning for unconstrained gait recognition", "authors": "Guo, Yuxiang and Huang, Siyuan and Prabhakar, Ram and Lau, Chun Pong and Chellappa, Rama and Peng, Cheng", "year": 2024, "venue": "IJCB", "motivation": "Enhance gait recognition in wild", "idea": "Separate motion segments, augment features", "techniques": "Motion detection, cross-modality distillation", "inputs": "RGB, Silhouette", "backbone": "CNN", "loss": "Triplet", "datasets": "CASIA-B, Gait3D, BRIAR", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/IJCB62174.2024.10744527", "bibtex": "@inproceedings{guo2024distillation,\n title = {Distillation-guided representation learning for unconstrained gait recognition},\n author = {Guo, Yuxiang and Huang, Siyuan and Prabhakar, Ram and Lau, Chun Pong and Chellappa, Rama and Peng, Cheng},\n year = {2024},\n booktitle = {2024 IEEE International Joint Conference on Biometrics (IJCB)},\n pages = {1--11}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 70 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait recognition holds the promise of robustly identifying subjects based on walking patterns instead of appearance information. While previous approaches have performed well for curated indoor data, they tend to underperform in unconstrained situations, e.g. in outdoor, long distance scenes, etc. We propose a framework, termed GAit DEtection and Recognition (GADER), for human authentication in challenging outdoor scenarios. Specifically, GADER leverages a Double Helical Signature to detect segments that contain human movement and builds discriminative features through a novel gait recognition method, where only frames containing gait information are used. To further enhance robustness, GADER encodes viewpoint information in its architecture, and distills representation from an auxiliary RGB recognition model, which enables GADER to learn from silhouette and RGB data at training time. At test time, GADER only infers from the silhouette modality. We evaluate our method on multiple State-of-The- Arts(SoTA) gait baselines and demonstrate consistent improvements on indoor and outdoor datasets, especially with a significant 25.2% improvement on unconstrained, remote gait data.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/IJCB62174.2024.10744527", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-hid2024", "citationKey": "hid2024", "archiveKey": "hid2024", "method": "", "title": "Human Identification at a Distance: Challenges, Methods and Results on the Competition HID 2024", "authors": "Yu, Shiqi and Wu, Weiming and Hu, Jiacong and Wang, Zepeng and Wang, Jingjie and Zhang, Meng and Wang, Runsheng and Ni, Yunfei and Huang, Yongzhen and Wang, Liang and Rahman Ahad, Md Atiqur", "year": 2024, "venue": "IJCB", "motivation": "Shared evaluation is needed to compare human identification systems at a distance.", "idea": "Report the HID 2024 competition and its participating methods.", "techniques": "Benchmark protocols; competition evaluation; system comparison.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/IJCB62174.2024.10744507", "bibtex": "@inproceedings{hid2024,\n title = {Human Identification at a Distance: Challenges, Methods and Results on the Competition HID 2024},\n author = {Yu, Shiqi and Wu, Weiming and Hu, Jiacong and Wang, Zepeng and Wang, Jingjie and Zhang, Meng and Wang, Runsheng and Ni, Yunfei and Huang, Yongzhen and Wang, Liang and Rahman Ahad, Md Atiqur},\n year = {2024},\n booktitle = {2024 IEEE International Joint Conference on Biometrics (IJCB)},\n pages = {1--8},\n doi = {10.1109/IJCB62174.2024.10744507}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Advances in Benchmarks / Competitions" ], "abstract": "This report summarizes the fifth HID competition, its evaluation setting and the approaches developed by leading teams. It documents improved recognition under a challenging cross-domain protocol and compares progress across successive competitions. The discussion uses these results to identify remaining obstacles to practical gait recognition.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://doi.org/10.1109/IJCB62174.2024.10744507", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-liu2024gait", "citationKey": "liu2024gait", "archiveKey": "liu2024gait", "method": "", "title": "Gait recognition of camouflaged people based on UAV infrared imaging", "authors": "Liu, Yang and Wang, Cong-qing and Li, Hui-peng and Zhou, Yong-jun", "year": 2024, "venue": "Infrared Physics & Technology", "motivation": "Camouflage and aerial acquisition complicate visible-light identification.", "idea": "Study gait recognition from UAV infrared observations.", "techniques": "Infrared sensing; aerial gait sequences; identity feature extraction.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1016/j.infrared.2024.105262", "bibtex": "@article{liu2024gait,\n title = {Gait recognition of camouflaged people based on UAV infrared imaging},\n author = {Liu, Yang and Wang, Cong-qing and Li, Hui-peng and Zhou, Yong-jun},\n year = {2024},\n journal = {Infrared Physics & Technology},\n volume = {138},\n pages = {105262},\n publisher = {Elsevier}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Task Definition" ], "abstract": "The paper investigates gait recognition from UAV-mounted infrared cameras, including people whose clothing blends into the surroundings. It introduces IR-150 and CTGait, a hierarchical architecture combining convolutional locality with Transformer attention. Experiments compare aerial infrared recognition with established visible-light and infrared benchmarks.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.sciencedirect.com/science/article/pii/S1350449524001464", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-zhou2024gait", "citationKey": "zhou2024gait", "archiveKey": "zhou2024gait", "method": "ScoNet / ScoNet-MT", "title": "Gait patterns as biomarkers: A video-based approach for classifying scoliosis", "authors": "Zhou, Zirui and Liang, Junhao and Peng, Zizhao and Fan, Chao and An, Fengwei and Yu, Shiqi", "year": 2024, "venue": "MICCAI", "motivation": "Conventional scoliosis screening depends on expertise and may involve radiation.", "idea": "Investigate video gait as a noninvasive signal for scoliosis classification.", "techniques": "Scoliosis1K; ScoNet; multi-task learning with ScoNet-MT.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://papers.miccai.org/miccai-2024/348-Paper1763.html", "bibtex": "@inproceedings{zhou2024gait,\n title = {Gait patterns as biomarkers: A video-based approach for classifying scoliosis},\n author = {Zhou, Zirui and Liang, Junhao and Peng, Zizhao and Fan, Chao and An, Fengwei and Yu, Shiqi},\n year = {2024},\n booktitle = {International conference on medical image computing and computer-assisted intervention},\n pages = {284--294}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Introduction", "Background / Broader Applications", "Challenges and Outlooks / Gait Representation Learning for Healthcare" ], "abstract": "This work studies gait video as a non-invasive source of information for scoliosis classification. It introduces Scoliosis1K and compares a baseline ScoNet with a multitask extension, ScoNet-MT. The experiments assess whether learned walking representations can support screening research without relying on radiographs as the model input.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://papers.miccai.org/miccai-2024/348-Paper1763.html", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-liu2024depression", "citationKey": "liu2024depression", "archiveKey": "liu2024depression", "method": "", "title": "Depression risk recognition based on gait: A benchmark", "authors": "Liu, Xiaotong and Li, Qiong and Hou, Saihui and Ren, Min and Hu, Xuecai and Huang, Yongzhen", "year": 2024, "venue": "Neurocomputing", "motivation": "Depression-related walking changes require systematic data and evaluation.", "idea": "Build a gait-based depression-risk recognition benchmark.", "techniques": "Gait representation learning; risk classification; benchmark evaluation.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1016/j.neucom.2024.128045", "bibtex": "@article{liu2024depression,\n title = {Depression risk recognition based on gait: A benchmark},\n author = {Liu, Xiaotong and Li, Qiong and Hou, Saihui and Ren, Min and Hu, Xuecai and Huang, Yongzhen},\n year = {2024},\n journal = {Neurocomputing},\n volume = {596},\n pages = {128045},\n publisher = {Elsevier}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Introduction", "Background / Broader Applications" ], "abstract": "D-Gait provides consented gait recordings for studying associations with depression risk. Its benchmark includes 292 volunteers, multiple views and clothing conditions, with both skeletal and silhouette inputs. Representative learning methods are compared to investigate how gait representations relate to risk labels derived from psychological assessments.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.sciencedirect.com/science/article/pii/S0925231224008166", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "castro2024attengait-pr-2024", "citationKey": "castro2024attengait", "archiveKey": "castro2024attengait", "method": "AttenGait", "title": "AttenGait: Gait recognition with attention and rich modalities", "authors": "Castro, Francisco M and Delgado-Escano, Ruben and Hernandez-Garc\\'\\ia, Ruber and Mar\\'\\in-Jimenez, Manuel J and Guil, Nicolas", "year": 2024, "venue": "PR", "motivation": "Limitations of silhouette-based recognition", "idea": "Attention network with richer modalities", "techniques": "Optical flow input; spatial-temporal attention", "inputs": "RGB, Silhouette, Flow", "backbone": "CNN, Attention", "loss": "CE, Triplet", "datasets": "CASIA-B, GREW", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2023.110171", "bibtex": "@article{castro2024attengait,\n title = {AttenGait: Gait recognition with attention and rich modalities},\n author = {Castro, Francisco M and Delgado-Escano, Ruben and Hernandez-Garc\\'\\ia, Ruber and Mar\\'\\in-Jimenez, Manuel J and Guil, Nicolas},\n year = {2024},\n journal = {Pattern Recognition},\n volume = {148},\n pages = {110171},\n publisher = {Elsevier}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 106 }, { "source": "Author paper summaries", "sheet": "journals", "row": 13 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Ideal Gait Modalities" ], "abstract": "Current gait recognition systems employ different types of manual attention mechanisms, like horizontal cropping of the input data to guide the training process and extract useful gait signatures for people identification. Typically, these techniques are applied using silhouettes as input, which limits the learning capabilities of the models. Thus, due to the limited information provided by silhouettes, state-of-the-art gait recognition approaches must use very simple and manually designed mechanisms, in contrast to approaches proposed for other topics such as action recognition. To tackle this problem, we propose AttenGait, a novel model for gait recognition equipped with trainable attention mechanisms that automatically discover interesting areas of the input data. AttenGait can be used with any kind of informative modalities, such as optical flow, obtaining state-of-the-art results thanks to the richer information contained in those modalities. We evaluate AttenGait on two public datasets for gait recognition: CASIA-B and GREW; improving the previous state-of-the-art results on them, obtaining 95.8% and 70.7% average accuracy, respectively. Code will be available at https://github.com/fmcp/attengait .", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2023.110171", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "sezavar2024dcapsnet-pr-2024", "citationKey": "sezavar2024dcapsnet", "archiveKey": "sezavar2024dcapsnet", "method": "DCapsNet", "title": "DCapsNet: Deep Capsule Network for Human Activity and Gait Recognition with Smartphone Sensors", "authors": "Ahmadreza Sezavar and Randa Atta and Mohammed Ghanbari", "year": 2024, "venue": "PR", "motivation": "Improve mobile sensor gait recognition", "idea": "Deep capsule network for time-series", "techniques": "Multi-scale convolutions; capsule routing", "inputs": "IMU", "backbone": "CNN, CapsNet", "loss": "Margin Loss", "datasets": "UCI-HAR, WISDM, WhuGAIT", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2023.110054", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 107 }, { "source": "Author paper summaries", "sheet": "journals", "row": 14 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "DCapsNet--Deep-capsule-network-for-human-activity-and-gait-_2024_Pattern-Rec.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Recently, deep neural networks are used to recognize human activity/gait through mobile sensors which have attracted a great attention. Although the existing deep neural networks that perform automatic feature extraction have achieved desirable performance, their classification accuracy needs to be improved. In this paper, a deep neural network that combines a set of convolutional layers and capsule network is proposed. The proposed ar - chitecture named DCapsNet is suited to automatically extract the activity or gait features through built in sensors and classify them. The convolutional layers of the DCapsNet are more suitable for processing temporal sequences and provide scalar outputs but not the equivariance. The capsule network (CapsNet) is then trained by a dynamic routing algorithm to capture the equivariance having a magnitude and orientation, which increases the efficiency of the model classification. The performance of the proposed model is evaluated on four public datasets: two HAR datasets (UCI-HAR and WISDM) and two gait datasets (WhuGAIT). The recognition accuracy of the proposed model for the UCI-HAR and WISDM datasets are 97.92 % and 99.30 %, respectively, and for the WhuGAIT Dataset #1 and Dataset #2 are 94.75 % and 97.16 %, respectively. Experimental results show that the proposed model achieves the highest recognition accuracy over the reported results of the state-of-the-art models.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2023.110054", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "yin2024msa-pr-2024", "citationKey": "yin2024msa", "archiveKey": "yin2024msa", "method": "MSA-GCN", "title": "MSA-GCN: Multiscale Adaptive Graph Convolution Network for Gait Emotion Recognition", "authors": "Yunfei Yin and Li Jing and Faliang Huang and Guangchao Yang and Zhuowei Wang", "year": 2024, "venue": "PR", "motivation": "Limitations of existing GCN for emotion", "idea": "Multiscale adaptive graph learning", "techniques": "Selective spatio-temporal convolution; global-local fusion", "inputs": "Skeleton", "backbone": "GCN, Attention", "loss": "CE", "datasets": "EMOGAIT", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2023.110117", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 109 }, { "source": "Author paper summaries", "sheet": "journals", "row": 16 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "MSA-GCN--Multiscale-Adaptive-Graph-Convolution-Network-fo_2024_Pattern-Recog.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait emotion recognition plays a crucial role in the intelligent system. Most existing approaches identify emotions by focusing on local actions over time. However, some valuable observational facts that the effective distances of different emotions in the time domain are different, and the local actions during walking are quite similar, are put aside in those methods. And this ignorance often ends up impairing performance of emotion recognition. To address the issues, a novel model, named MSA-GCN ( MultiScale Adaptive Graph Convolution Network), is proposed to utilize the valuable observational knowledge for improving emotion recognition performance. In the proposed model, an adaptive spatio-temporal graph convolution is designed to dynamically select convolution kernels to learn the spatio-temporal features of different emotions. Moreover, a Cross-Scale Mapping Interaction mechanism (CSMI) is proposed to construct an adaptive adjacency matrix for high-quality aggregation of the multiscale information. Extensive experimental results on public datasets indicate that, compared with the state-of-the-art methods, the proposed approach achieves better performance in terms of emotion recognition accuracy, and shows the proposed approach is promising.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2023.110117", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "chen2024gait-pr-2024", "citationKey": "chen2024gait", "archiveKey": "chen2024gait", "method": "Spatio-Temporal Two-Branch Network (STTBN)", "title": "Gait Feature Learning via Spatio-temporal Two-branch Networks", "authors": "Yifan Chen and Xuelong Li", "year": 2024, "venue": "PR", "motivation": "Improve robustness under view/walking variations", "idea": "Dual-branch spatio-temporal feature learning", "techniques": "Temporal sequence modeling; selective spatial convolution", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2023.110090", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 108 }, { "source": "Author paper summaries", "sheet": "journals", "row": 15 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Gait-feature-learning-via-spatio-temporal-two-branch-n_2024_Pattern-Recognit.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition has become a mainstream technology for identification due to its ability to capture gait features over long distances without subject cooperation and resistance to camouflage. However, current gait recognition methods face challenges as they use a single network to extract both temporal and spatial features from gait sequences. This approach imposes a heavy burden on the network, resulting in reduced extraction efficiency. To solve this problem, we propose a two-branch network to extract the spatio-temporal features of gait sequences. One branch primarily focuses on spatial feature extraction, while the other concentrates on temporal feature extraction. This design can make one branch focus on a specific task, leading to significant performance improvements. For temporal feature extraction, we propose the Global Temporal Information Extraction Network (GTIEN). GTIEN extracts temporal features of gait sequences by sequentially exploring the relationship between adjacent gait silhouettes from pixel and block levels. For spatial feature extraction, we introduce the Selective Horizontal Pyramid Convolution Network (SHPCN). SHPCN explores the multigranularity features of gait silhouettes from global and local perspectives and assigns them appropriate weights according to their importance. By reasonably combining the temporal features extracted from GTIEN and spatial features extracted from SHPCN, we can effectively learn the spatial–temporal information of the gait sequences. Extensive experiments on CASIA-B and OUMVLP demonstrate that our method has better performance than some state-of-the-art methods.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2023.110090", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "yue2024deep-prl-2024", "citationKey": "yue2024deep", "archiveKey": "yue2024deep", "method": "GAN-LSTM", "title": "Deep Motion Estimation through Adversarial Learning for Gait Recognition", "authors": "Yuanhao Yue and Laixiang Shi and Zheng Zheng and Long Chen and Zhongyuan Wang and Qin Zou", "year": 2024, "venue": "PRL", "motivation": "Improve robustness in daily-life scenarios", "idea": "Model-free gait with enhanced silhouettes", "techniques": "Multi-scale segmentation; adversarial temporal modeling; sequence classification", "inputs": "Silhouette, RGB, Depth", "backbone": "CNN, GAN, LSTM", "loss": "CE", "datasets": "OU-LP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1016/j.patrec.2024.06.031", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 116 }, { "source": "Author paper summaries", "sheet": "journals", "row": 23 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Deep-motion-estimation-through-adversarial-learning_2024_Pattern-Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition is a form of identity verification that can be performed over long distances without requiring the subject’s cooperation, making it particularly valuable for applications such as access control, surveillance, and criminal investigation. The essence of gait lies in the motion dynamics of a walking individual. Accurate gait-motion estimation is crucial for high-performance gait recognition. In this paper, we introduce two main designs for gait motion estimation. Firstly, we propose a fully convolutional neural network named W-Net for silhouette segmentation from video sequences. Secondly, we present an adversarial learning-based algorithm for robust gait motion estimation. Together, these designs contribute to a high-performance system for gait recognition and user authentication. In the experiment, two datasets, i.e., OU-IRIS and our own dataset, are used for performance evaluation. Experimental results show that, the W-Net achieves an accuracy of 89.46% in silhouette segmentation, and the proposed user-authentication method achieves over 99.6% and 93.8% accuracy on the two datasets, respectively.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patrec.2024.06.031", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "mehmood2024human-prl-2024", "citationKey": "mehmood2024human", "archiveKey": "mehmood2024human", "method": "Two-Stream CNN-55", "title": "Human Gait Recognition by Using Two Stream Neural Network along with Spatial and Temporal Features", "authors": "Asif Mehmood and Javeria Amin and Muhammad Sharif and Seifedine Kadry", "year": 2024, "venue": "PRL", "motivation": "Overcome covariates in gait recognition", "idea": "Automated CNN-based gait framework", "techniques": "Deep CNN extraction; dual-stream features; evolutionary feature optimization", "inputs": "RGB, Flow", "backbone": "CNN", "loss": "", "datasets": "CASIA-B", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1016/j.patrec.2024.02.010", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 117 }, { "source": "Author paper summaries", "sheet": "journals", "row": 24 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Human-Gait-Recognition-by-using-Two-Stream-Neural-Netwo_2024_Pattern-Recogni.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Human Gait Recognition (HGR) is referred to as a biometric tactic that is broadly used for the recognition of an individual by using the pattern of walking. There are some key factors such as angle variation, clothing variation, foot shadows, and carrying conditions that affect the human gait. In this work, a new approach is proposed for the HGR that contains five major steps. In the first step, the video data is converted into image frames. In the second step, RGB to GRAY conversion is carried out. After that, a two-stream network is designed by using a 55- layer CNN model called CNN-55 trained on CIFAR-100. The CNN-55 is designed from scratch and trained on the CIFAR-100 dataset by selecting hyperparameters. This pre-trained CNN-55 is used to build a two-stream network. In Stream-1 the optical flow frames are obtained by Horn and Schunk algorithm. These frames are fed into a CNN-55 to extract temporal features. In Stream-2 the GRAY frames are fed to the CNN-55 model for extraction of spatial features. After that, both vectors are serially fused. In the fourth step, the fused feature vector is fed into the Genetic Algorithm for optimization. Finally, the feature vector is fed into the One-Versus-All SVM classifier for recognition. The system is tested on all CASIA-B angles such as 00 0 , 18 0 , 36 0 , 54 0 , 72 0 , 90 0 , 108 0 , 126 0 , 144 0 , 162 0 , and 180 0 which provides accuracy of 97.10%, 96.80%, 94.60%, 98.0%, 98.30%, 96.80%, 97.60, 96.90%, 99.60%, 96.80%, and 97.60%, respectively. The proposed method produces better outcomes compared to recent techniques.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patrec.2024.02.010", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-ye2024robotreid", "citationKey": "ye2024robotreid", "archiveKey": "ye2024robotreid", "method": "", "title": "Person Re-Identification for Robot Person Following With Online Continual Learning", "authors": "Ye, Hanjing and Zhao, Jieting and Zhan, Yu and Chen, Weinan and He, Li and Zhang, Hong", "year": 2024, "venue": "RA-L", "motivation": "A following robot encounters changing views and appearances over time.", "idea": "Adapt person re-identification online during robot operation.", "techniques": "Online continual learning; person ReID; robot person following.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/LRA.2024.3438042", "bibtex": "@article{ye2024robotreid,\n title = {Person Re-Identification for Robot Person Following With Online Continual Learning},\n author = {Ye, Hanjing and Zhao, Jieting and Zhan, Yu and Chen, Weinan and He, Li and Zhang, Hong},\n year = {2024},\n journal = {IEEE Robotics and Automation Letters},\n volume = {9},\n number = {11},\n pages = {9151--9158},\n doi = {10.1109/LRA.2024.3438042}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Challenges and Outlooks / Gait Recognition for Agents" ], "abstract": "This work adapts person re-identification during robot following. An online continual-learning system updates its feature extractor using recent observations and a memory of earlier appearances, helping the robot recover its target after occlusion. Experiments examine viewpoint and lighting changes alongside distraction from visually similar people.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2309.11727", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "thejaswin2024exploring-tbiom-2024", "citationKey": "thejaswin2024exploring", "archiveKey": "thejaswin2024exploring", "method": "Adapt-FuseNet", "title": "Exploring Fusion Techniques and Explainable AI on Adapt-FuseNet Context-Adaptive Fusion of Face and Gait for Person Identification", "authors": "Thejaswin S and Ashwin Prakash and Athira Nambiar and Alexandre Bernardino", "year": 2024, "venue": "TBIOM", "motivation": "Robust multimodal biometric recognition", "idea": "Context-adaptive fusion of features", "techniques": "Attention-based explainable feature integration", "inputs": "RGB", "backbone": "CNN", "loss": "CE", "datasets": "CASIA-A, CASIA-B", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.54499/LA/P/0083/2020", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 128 }, { "source": "Author paper summaries", "sheet": "journals", "row": 35 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Exploring_Fusion_Techniques_and_Explainable_AI_on_Adapt-FuseNet_Context-Adaptive_Fusion_of_Face_and_Gait_for_Person_Identification.pdf", "abstract": "Biometrics such as human gait and face play a significant role in vision-based surveillance applications. However, multimodal fusion of biometric features is a challenging task in non-controlled environments due to varying reliability of the features from different modalities in changing contexts, such as viewpoints, illuminations, occlusion, background clutter, and clothing. For instance, in person identification in the wild, facial and gait features play a complementary role, as, in principle, face provides more discriminatory features than gait if the person is frontal to the camera, while gait features are more discriminative in lateral views. Classical fusion techniques typically address this problem by explicitly computing in which context the data is obtained (e.g., frontal or lateral) and designing custom data fusion strategies for each context. However, this requires an initial enumeration of all the possible contexts and the design of context “detectors”, which bring their own challenges. Hence, how to effectively utilize both facial and gait information in arbitrary conditions is still an open problem. In this paper we present a context-adaptive multi-biometric fusion strategy that does not require the prior determination of context features; instead, the context is implicitly encoded in the fusion process by a set of attentional weights that encode the relevance of the different modalities for each particular data sample. The key contributions of the paper are threefold. First, we propose a novel framework for the dynamic fusion of multiple biometrics modalities leveraging attention techniques, denoted ‘Adapt-FuseNet’. Second, we perform an extensive evaluation of the proposed method in comparison to various other fusion techniques such as Bilinear Pooling, Parallel Co-attention, Keyless Attention, Multi-modal Factorized High-order Pooling, and Multimodal Tucker Fusion. Third, an Explainable Artificial Intelligencebased interpretation tool is used to analyse how the attention mechanism of ‘Adapt-FuseNet’ is capturing context implicitly and making the best weighting of the different modalities for the task at hand. This enables the interpretability of results in a more human-compliant way, hence boosting our confidence of the operation of AI systems in the wild. Extensive experiments are carried out on two public gait datasets (CASIA-A and CASIA- B), showing that ‘Adapt-FuseNet’ significantly outperforms the state-of-the-art.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.54499/LA/P/0083/2020", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "zheng2024gaitstr-tbiom-2024", "citationKey": "zheng2024gaitstr", "archiveKey": "zheng2024gaitstr", "method": "GaitSTR", "title": "GaitSTR Gait Recognition With Sequential Two-Stream Refinement", "authors": "Wanrong Zheng and Haidong Zhu and Zhaoheng Zheng and Ram Nevatia", "year": 2024, "venue": "TBIOM", "motivation": "Improve skeleton-silhouette gait recognition", "idea": "Refine skeletons using silhouettes", "techniques": "Cross-modal and self-correcting fusion", "inputs": "Silhouette, Skeleton", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, Gait3D, GREW", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TBIOM.2024.3390626", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 129 }, { "source": "Author paper summaries", "sheet": "journals", "row": 36 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "GaitSTR_Gait_Recognition_With_Sequential_Two-Stream_Refinement.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition aims to identify a person based on their walking sequences, serving as a useful biometric modality as it can be observed from long distances without requiring cooperation from the subject. In representing a person’s walking sequence, silhouettes and skeletons are the two primary modalities used. Silhouette sequences lack detailed part information when overlapping occurs between different body segments and are affected by carried objects and clothing. Skeletons, comprising joints and bones connecting the joints, provide more accurate part information for different segments; however, they are sensitive to occlusions and low-quality images, causing inconsistencies in frame-wise results within a sequence. In this paper, we explore the use of a two-stream representation of skeletons for gait recognition, alongside silhouettes. By fusing the combined data of silhouettes and skeletons, we refine the two-stream skeletons, joints, and bones through self-correction in graph convolution, along with cross-modal correction with temporal consistency from silhouettes. We demonstrate that with refined skeletons, the performance of the gait recognition model can achieve further improvement on public gait recognition datasets compared with state-of-the-art methods without extra annotations.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TBIOM.2024.3390626", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "zou2024multi-tbiom-2024", "citationKey": "zou2024multi", "archiveKey": "zou2024multi", "method": "MSAFF", "title": "A multi-stage adaptive feature fusion neural network for multimodal gait recognition", "authors": "Zou, Shinan and Xiong, Jianbo and Fan, Chao and Shen, Chuanfu and Yu, Shiqi and Tang, Jin", "year": 2024, "venue": "TBIOM", "motivation": "Enhance multimodal gait recognition", "idea": "Fuse silhouette and skeleton features", "techniques": "Multi-stage adaptive spatio-temporal extraction", "inputs": "Silhouette, Skeleton", "backbone": "CNN, GCN", "loss": "Triplet", "datasets": "CASIA-B, Gait3D, GREW", "codeAvailable": "No", "codeUrl": "", "url": "", "bibtex": "@article{zou2024multi,\n title = {A multi-stage adaptive feature fusion neural network for multimodal gait recognition},\n author = {Zou, Shinan and Xiong, Jianbo and Fan, Chao and Shen, Chuanfu and Yu, Shiqi and Tang, Jin},\n year = {2024},\n journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},\n volume = {6},\n number = {4},\n pages = {539--549},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 127 }, { "source": "Author paper summaries", "sheet": "journals", "row": 34 } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait recognition is a biometric technology that has received extensive attention. Most existing gait recognition algorithms are unimodal, and a few multimodal gait recognition algorithms perform multimodal fusion only once. None of these algorithms may fully exploit the complementary advantages of the multiple modalities. In this paper, by considering the temporal and spatial characteristics of gait data, we propose a multi-stage feature fusion strategy (MSFFS), which performs multimodal fusions at different stages in the feature extraction process. Also, we propose an adaptive feature fusion module (AFFM) that considers the semantic association between silhouettes and skeletons. The fusion process fuses different silhouette areas with their more related skeleton joints. Since visual appearance changes and time passage co-occur in a gait period, we propose a multiscale spatial-temporal feature extractor (MSSTFE) to learn the spatial-temporal linkage features thoroughly. Specifically, MSSTFE extracts and aggregates spatial-temporal linkages information at different spatial scales. Combining the strategy and modules mentioned above, we propose a multi-stage adaptive feature fusion (MSAFF) neural network, which shows state-ofthe-art performance in many experiments on three datasets. Besides, MSAFF is equipped with feature dimensional pooling (FD Pooling), which can significantly reduce the dimension of the gait representations without hindering the accuracy. The code can be found here. https://github.com/ShinanZou/MSAFF.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "sun2024trigait-tbiom-2025", "citationKey": "sun2024trigait", "archiveKey": "sun2024trigait", "method": "TriGait", "title": "TriGait: Hybrid Fusion Strategy for Multimodal Alignment and Integration in Gait Recognition", "authors": "Sun, Yan and Feng, Xueling and Liu, Xiaolei and Ma, Liyan and Hu, Long and Nixon, Mark S", "year": 2024, "venue": "TBIOM", "motivation": "Robust multimodal gait recognition", "idea": "Combine silhouette and skeleton features", "techniques": "Hybrid early-late cross-modal fusion", "inputs": "Silhouette, Skeleton", "backbone": "CNN, Transformer", "loss": "CE, Triplet", "datasets": "CASIA-B, Gait3D, OU-MVLP", "codeAvailable": "Yes", "codeUrl": "", "url": "", "bibtex": "@article{sun2024trigait,\n title = {TriGait: Hybrid Fusion Strategy for Multimodal Alignment and Integration in Gait Recognition},\n author = {Sun, Yan and Feng, Xueling and Liu, Xiaolei and Ma, Liyan and Hu, Long and Nixon, Mark S},\n year = {2024},\n journal = {IEEE Transactions on Biometrics, Behavior, and Identity Science},\n volume = {7},\n number = {1},\n pages = {82--94},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 132 }, { "source": "Author paper summaries", "sheet": "journals", "row": 39 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Ideal Gait Modalities" ], "yearNote": "Archive: 2025; bibliography: 2024", "abstract": "Due to the inherent limitations of single modalities, multimodal fusion has become increasingly popular in many computer vision fields, leveraging the complementary advantages of unimodal methods. As an emerging biometric technology with great application potential, gait recognition faces similar challenges. The prevailing silhouette-based and skeleton-based gait recognition methods have their respective limitations: one focuses on appearance information while neglecting structural details, and the other does the opposite. Multimodal gait recognition, which combines silhouette and skeleton, promises more robust predictions. However, it is essential and difficult to explore the implicit interaction between dense pixels and discrete coordinate points. Most existing multimodal gait recognition methods basically concatenated features from silhouette and skeleton and did not fully exploit complementarity between them. This paper presents a hybrid fusion strategy called TriGait, which is a three-branch structural model and thoroughly explores the interaction and complementarity of the two modalities. To solve the problem of data heterogeneity and explore the mutual information of two modalities, we propose the use of a crossmodal token generator (CMTG) within a fusion branch to align and fuse the low-level features of the two modalities. Additionally, TriGait has two extra branches for extracting high-level semantic information from silhouette and skeleton. By combining low-level correlation information and high-level semantic information, TriGait provides a comprehensive and discriminative representation of a subject’s gait. Extensive experimental results on CASIA-B, Gait3D and OUMVLP demonstrate the effectiveness of TriGait. Remarkably, TriGait achieves the rank-1 mean accuracy of 96.6%, 61.4% and 91.1% on CASIA-B, Gait3D and OUMVLP respectively, outperforming the state-of-the-art methods. The source code is available at: https://github.com/YanSun-github/TriGait/.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "hou2024cloth-tcsvt-2024", "citationKey": "hou2024cloth", "archiveKey": "hou2024cloth", "method": "CCH", "title": "Cloth-Imbalanced Gait Recognition via Hallucination", "authors": "Hou, Saihui and Huang, Panjian and Liu, Xu and Cao, Chunshui and Huang, Yongzhen", "year": 2024, "venue": "TCSVT", "motivation": "Address clothing imbalance in gait", "idea": "Generate cross-clothes features efficiently", "techniques": "Feature transfer, variation modeling, non-parametric augmentation", "inputs": "Silhouette", "backbone": "CNN", "loss": "Triplet", "datasets": "CASIA-B, Outdoor-Gait", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TCSVT.2024.3360232", "bibtex": "@article{hou2024cloth,\n title = {Cloth-Imbalanced Gait Recognition via Hallucination},\n author = {Hou, Saihui and Huang, Panjian and Liu, Xu and Cao, Chunshui and Huang, Yongzhen},\n year = {2024},\n journal = {IEEE Transactions on Circuits and Systems for Video Technology},\n volume = {34},\n number = {7},\n pages = {5665--5676},\n doi = {10.1109/TCSVT.2024.3360232}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 144 }, { "source": "Author paper summaries", "sheet": "journals", "row": 50 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Handling Real-World Noise", "Advances in Methods / Gait Metric Designs" ], "abstract": "The study in the gait field has rarely paid attention to the class-imbalanced learning, while the realistic data always exhibits an imbalanced distribution. The main reason lies in the difficulty of collecting the cross-clothes sequences, since the collection is usually aided by person re-identification and it is more likely to obtain the sequences for a subject wearing the same clothes. In this work, we formulate a new problem to tackle the task-specific cloth-imbalanced issue, dubbed as Cloth-Imbalanced Gait Recognition, and the training data consists of two parts denoted as head set and tail set. The sequences for a subject in head set cover the cross-clothes variation which is scarce in tail set to mimic the collection difficulty. Along with the problem formulation, we design a new method to deal with the inherent challenges, called Cross-Clothes Hallucination or CCH for short. Our method is inspired by the observation that certain directions in deep feature space correspond to meaningful semantic transformations, and it tries to generate the cross-clothes sequences for tail set referring to the cloth-changing transformation in head set. To evaluate CCH, we build two cloth-imbalanced benchmarks based on the widely-used CASIA-B and Outdoor-Gait. Extensive experiments demonstrate that CCH brings significant improvements over the baselines.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TCSVT.2024.3360232", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "huang2024gaitdan-tcsvt-2024", "citationKey": "huang2024gaitdan", "archiveKey": "huang2024gaitdan", "method": "GaitDAN", "title": "GaitDAN Cross-View Gait Recognition via Adversarial Domain Adaptation", "authors": "Tianhuan Huang and Xianye Ben and Chen Gong and Wenzheng Xu and Qiang Wu and Hongchao Zhou", "year": 2024, "venue": "TCSVT", "motivation": "Address cross-view gait recognition challenge", "idea": "Treat view change as domain adaptation", "techniques": "Hierarchical feature aggregation, adversarial alignment, metric learning", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-LP, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TCSVT.2024.3384308", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 146 }, { "source": "Author paper summaries", "sheet": "journals", "row": 52 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "GaitDAN_Cross-View_Gait_Recognition_via_Adversarial_Domain_Adaptation.pdf", "authorsSource": "Public scholarly metadata", "abstract": "View change causes significant differences in the gait appearance. Consequently, recognizing gait in cross-view scenarios is highly challenging. Most recent approaches either convert the gait from the original view to the target view before recognition is carried out or extract the gait feature irrelevant to the camera view through either brute force learning or decouple learning. However, these approaches have many constraints, such as the difficulty of handling unknown camera views. This work treats the view-change issue as a domainchange issue and proposes to tackle this problem through adversarial domain adaptation. This way, gait information from different views is regarded as the data from different subdomains. The proposed approach focuses on adapting the gait feature differences caused by such sub-domain change and, at the same time, maintaining sufficient discriminability across the different people. For this purpose, a Hierarchical Feature Aggregation (HFA) strategy is proposed for discriminative feature extraction. By incorporating HFA, the feature extractor can well aggregate the spatial-temporal feature across the various stages of the network and thereby comprehensive gait features can be obtained. Then, an Adversarial View-change Elimination (A VE) module equipped with a set of explicit models for recognizing the different gait viewpoints is proposed. Through the adversarial learning process, A VE would not be able to identify the gait viewpoint in the end, given the gait features generated by the feature extractor. That is, the adversarial domain adaptation mitigates the view change factor, and discriminative gait features that are compatible with all sub-domains are effectively extracted. Extensive experiments on three of the most popular public datasets, CASIA-B, OULP, and OUMVLP richly demonstrate the effectiveness of our approach.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TCSVT.2024.3384308", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "hou2024adaptive-tifs-2024", "citationKey": "hou2024adaptive", "archiveKey": "hou2024adaptive", "method": "Adaptive Knowledge Transfer (AKT)", "title": "Adaptive Knowledge Transfer for Weak-Shot Gait Recognition", "authors": "Hou, Saihui and Hu, Xuecai and Cao, Chunshui and Liu, Xu and Huang, Yongzhen", "year": 2024, "venue": "TIFS", "motivation": "Handle clothing variation in gait", "idea": "Weak-shot learning formulation", "techniques": "Pairwise reliability analysis; supervised knowledge learning; adaptive knowledge transfer", "inputs": "Silhouette", "backbone": "CNN", "loss": "Triplet, Contrastive", "datasets": "CASIA-B, Outdoor-Gait, CASIA-E", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TIFS.2024.3428371", "bibtex": "@article{hou2024adaptive,\n title = {Adaptive Knowledge Transfer for Weak-Shot Gait Recognition},\n author = {Hou, Saihui and Hu, Xuecai and Cao, Chunshui and Liu, Xu and Huang, Yongzhen},\n year = {2024},\n journal = {IEEE Transactions on Information Forensics and Security},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 157 }, { "source": "Author paper summaries", "sheet": "journals", "row": 61 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Gait Learning Frameworks" ], "abstract": "Most works for cloth-changing gait recognition assume that the sequences of different clothes are accessible for each subject in the training set, which, however, is almost impossible for real applications. In practice, the collection of gait sequences is usually aided by person re-identification which is more likely to cluster the cloth-consistent sequences for a subject, and it is laborious to merge the cloth-changing clusters with the same identity. As a result, the training set is usually comprised of two subsets, i.e., a fully-annotated base set where the cloth-changing sequences are available for each subject, and a weakly-annotated wild set where the sequences of different clothes for a subject are assigned diverse labels. In this work, we formulate a problem named Weak-Shot Gait Recognition which seeks to learn discriminative features from the mixture of base set and wild set . Furthermore, we propose an effective method called Adaptive Knowledge Transfer to deal with the weak-shot issue. In particular, we define the knowledge as the ability to judge whether two sequences come from the same subject or not, and we take an adaptive way to mine the useful information from wild set . For the experimental study, we build three weak-shot benchmarks based on CASIA-B, Outdoor-Gait, and CASIA-E respectively. Extensive experiments show that our method can bring consistent improvement. For example, under the cloth-changing condition of on the weak-shot CASIA-B, our method exceeds a naïve baseline by 7 .10%.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIFS.2024.3428371", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "liang2024dcs-tifs-2024", "citationKey": "liang2024dcs", "archiveKey": "liang2024dcs", "method": "DCS-Gait", "title": "DCS-Gait A Class-Level Domain Adaptation Approach for Cross-Scene and Cross-State Gait Recognition Using Wi-Fi CSI", "authors": "Ying Liang and Wenjie Wu and Haobo Li and Xiaojun Chang and Xiaojiang Chen and Jinye Peng and Pengfei Xu", "year": 2024, "venue": "TIFS", "motivation": "Robust Wi-Fi based gait identification", "idea": "Domain-invariant feature learning", "techniques": "Cross-attention alignment; pseudo-label generation; matching-based filtering", "inputs": "Wi-Fi CSI", "backbone": "CNN, Transformer", "loss": "CE", "datasets": "Wi-Fi CSI gait dataset", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TIFS.2024.3356827", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "journals", "row": 62 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "DCS-Gait_A_Class-Level_Domain_Adaptation_Approach_for_Cross-Scene_and_Cross-State_Gait_Recognition_Using_Wi-Fi_CSI.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Wi-Fi CSI-based gait recognition is a non-intrusive passive biometric identification technology that has garnered significant attention in the fields of security and smart furniture due to its user-friendly nature. However, in practical application scenarios, gait recognition systems face the challenge of reliably identifying subjects across different scenes or states. To overcome this challenge, this paper proposes DCS-Gait, a domain adaptation solution for cross-scene and cross-state gait recognition based on Wi-Fi CSI. DCS-Gait leverages a novel data distribution measurement called Cross-Attention Metric to align the class-level data distribution differences, enabling the model to learn invariant features across scenes and states. To address the issue of data annotation, we employ a pre-training method to obtain pseudo labels for the dataset. Additionally, a combined matching filtering technique is utilized to generate high-quality pseudo labels for unrecognized data, which can be further employed for supervised model training. We evaluated the effectiveness of DCS-Gait on a large test set consisting of 34 subjects, 2 scenes, and 3 different states, and the results demonstrate significant improvements over the state-of-the-art baselines in both cross-scene and cross-state gait recognition tasks. DCS-Gait provides a promising and reliable solution for accurate cross-scene and cross-state gait recognition in real-world settings.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIFS.2024.3356827", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "chen2024egst-tifs-2024", "citationKey": "chen2024egst", "archiveKey": "chen2024egst", "method": "EGST", "title": "EGST: An efficient solution for human gaits recognition using neuromorphic vision sensor", "authors": "Chen, Liaogehao and Zhang, Zhenjun and Xiao, Yang and Wang, Yaonan", "year": 2024, "venue": "TIFS", "motivation": "Efficient event-based gait recognition", "idea": "Graph sequence representation learning", "techniques": "Event-stream slicing; spatio-temporal correlation modeling; lightweight graph network", "inputs": "Event-stream", "backbone": "GCN", "loss": "CE", "datasets": "DVS128-Gait-Day, DVS128-Gait-Night, CASIA-B", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1109/TIFS.2024.3409167", "bibtex": "@article{chen2024egst,\n title = {EGST: An efficient solution for human gaits recognition using neuromorphic vision sensor},\n author = {Chen, Liaogehao and Zhang, Zhenjun and Xiao, Yang and Wang, Yaonan},\n year = {2024},\n journal = {IEEE Transactions on Information Forensics and Security},\n volume = {19},\n pages = {6144--6154},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 159 }, { "source": "Author paper summaries", "sheet": "journals", "row": 63 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Ideal Gait Modalities" ], "abstract": "Traditional cameras struggle to perform in challenging scenarios such as low latency, high speed and high dynamic range. In contrast, neuromorphic vision sensors (event cameras) have great potential for robotics and computer vision due to the advantages of high temporal resolution, high dynamic range, and ultra-low resource consumption. Event cameras are novel bio-inspired sensors that monitor the brightness change of each pixel asynchronously and provide a stream of events encoding the time, position and sign of the brightness changes. Hence, traditional computer vision methods cannot be directly applied to the event-stream. Finding event representations that completely maintain event attributes, as well as efficient and accurate learning approaches, is the key to unlocking the potential of event cameras. In this study, we reveal the rigid transfer from event-stream to graph that has been overlooked in previous work and introduce a novel event representation, namely event graph sequence (EGS) considering the local and global temporal clues. Coupled with EGS, we propose a spatio-temporal pattern extracting (STPE) module to capture the spatio-temporal correlation and evolution of EGS. Our novel framework, Event Graph Sequence Transformer (EGST), exploits event properties to provide efficient and accurate recognition. This study focuses on the event-based human gaits recognition task, and EGST is evaluated on three different event-based gait datasets. The evaluation results show better or comparable accuracy than the state-of-theart, while requiring extremely low computation resources. The code will be available at https://github.com/C19h/EGST.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIFS.2024.3409167", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "song2023gait-tifs-2024", "citationKey": "song2023gait", "archiveKey": "song2023gait", "method": "GAR-Net", "title": "Gait Attribute Recognition A New Benchmark for Learning Richer Attributes From Human Gait Patterns", "authors": "Xu Song and Saihui Hou and Yan Huang and Chunshui Cao and Xu Liu and Yongzhen Huang and Caifeng Shan", "year": 2024, "venue": "TIFS", "motivation": "Rich gait attribute understanding", "idea": "Multi-attribute gait recognition framework", "techniques": "Large-scale attribute dataset; adaptive feature selection; multi-scale motion encoding", "inputs": "Silhouette, Skeleton, RGB", "backbone": "CNN", "loss": "", "datasets": "MA-Gait", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TIFS.2023.3318934", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 160 }, { "source": "Author paper summaries", "sheet": "journals", "row": 64 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Gait_Attribute_Recognition_A_New_Benchmark_for_Learning_Richer_Attributes_From_Human_Gait_Patterns.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Compared to gait recognition, Gait Attribute Recognition (GAR) is a seldom-investigated problem. However, since gait attribute recognition can provide richer and finer semantic descriptions, it is an indispensable part of building intelligent gait analysis systems. Nonetheless, the types of attributes considered in the existing datasets are very limited. This paper contributes a new benchmark dataset for gait attribute recognition named Multi-Attribute Gait (MA-Gait). Our MA-Gait contains 95 subjects recorded from 12 camera views, resulting in more than 13000 sequences, with 16 attributes labeled, including six attributes that have never been considered in the literature. Moreover, we propose a Multi-Scale Motion Encoder (MSME) to extract robust motion features, and an Attribute-Guided Feature Selection Module (AGFSM) to adaptively capture the most discriminative attribute features from static appearance features and dynamic motion features for different attributes. Our method achieves the best GAR accuracy on the new dataset. Comprehensive experiments show the effectiveness of the proposed method through both quantitative and qualitative evaluations.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIFS.2023.3318934", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "huang2024integral-tifs-2024", "citationKey": "huang2024integral", "archiveKey": "huang2024integral", "method": "GaitApp", "title": "Integral Pose Learning via Appearance Transfer for Gait Recognition", "authors": "Panjian Huang and Saihui Hou and Chunshui Cao and Xu Liu and Xuecai Hu and Yongzhen Huang", "year": 2024, "venue": "TIFS", "motivation": "Overcome appearance covariate challenges", "idea": "Extract appearance-invariant integral poses", "techniques": "Pose-shape fusion; generative appearance transfer; disentangled representation learning", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet, Contrastive", "datasets": "CASIA-B, FVG", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1109/TIFS.2024.3382606", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 161 }, { "source": "Author paper summaries", "sheet": "journals", "row": 65 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Integral_Pose_Learning_via_Appearance_Transfer_for_Gait_Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition plays an important role in video surveillance and security by identifying humans based on their unique walking patterns. The existing gait recognition methods have achieved competitive accuracy with shape and motion patterns under limited-covariate conditions. However, when extreme appearance changes distort discriminative features, gait recognition yields unsatisfactory results under cross-covariate conditions. In this work, we first indicate that the integral pose in each silhouette maintains an appearance-unrelated discriminative identity. However, the monotonous appearance variables in a gait database cause gait models to have difficulty extracting integral poses. Therefore, we propose an Appearance-transferable Disentangling and Generative Network (GaitApp) to generate gait silhouettes with rich appearances and invariant poses. Specifically, GaitApp leverages multi-branch cooperation to disentangle pose features and appearance features, and transfers the appearance information from one subject to another. By simulating a person constantly changing appearances under limitedcovariate conditions, downstream models enable to extract discriminative integral pose features. Extensive experiments demonstrate that our method allows representative gait models to stand at a new altitude, further promoting the exploration to cross-covariate gait recognition. All the code is available at https://github.com/Hpjhpjhs/GaitApp.git", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIFS.2024.3382606", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "dou2022gaitmpl-tip-2024", "citationKey": "dou2024gaitmpl", "archiveKey": "dou2022gaitmpl", "method": "GaitMPL", "title": "GaitMPL: Gait Recognition With Memory-Augmented Progressive Learning", "authors": "Dou, Huanzhang and Zhang, Pengyi and Zhao, Yuhan and Dong, Lin and Qin, Zequn and Li, Xi", "year": 2024, "venue": "TIP", "motivation": "Address hard sample inconsistency", "idea": "Progressive learning with memory guidance", "techniques": "Dynamic sample reweighting, memory bank, structure-aligned updates, feature compactness", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1109/TIP.2022.3164543", "bibtex": "@article{dou2024gaitmpl,\n title = {GaitMPL: Gait Recognition With Memory-Augmented Progressive Learning},\n author = {Dou, Huanzhang and Zhang, Pengyi and Zhao, Yuhan and Dong, Lin and Qin, Zequn and Li, Xi},\n year = {2024},\n journal = {IEEE Transactions on Image Processing},\n volume = {33},\n pages = {1464--1475},\n doi = {10.1109/TIP.2022.3164543}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 170 }, { "source": "Author paper summaries", "sheet": "journals", "row": 72 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Gait Metric Designs" ], "abstract": "Gait recognition aims at identifying the pedestrians at a long distance by their biometric gait patterns. It is inherently challenging due to the various covariates and the properties of silhouettes (textureless and colorless), which result in two kinds of pair-wise hard samples: the same pedestrian could have distinct silhouettes (intra-class diversity) and different pedestrians could have similar silhouettes (interclass similarity). In this work, we propose to solve the hard sample issue with a Memory-augmented Progressive Learning network (GaitMPL), including Dynamic Reweighting Progressive Learning module (DRPL) and Global Structure-Aligned Memory bank (GSAM). Specifically, DRPL reduces the learning difficulty of hard samples by easy-to-hard progressive learning. GSAM further augments DRPL with a structure-aligned memory mechanism, which maintains and models the feature distribution of each ID. Experiments on two commonly used datasets, CASIA-B and OU-MVLP, demonstrate the effectiveness of GaitMPL. On CASIA-B, we achie ve the state-of-the-art performance, i.e., 88.0% on the most challenging condition (Clothing) and 93.3% on the average condition, which outperforms the other methods by at least 3.8% and 1.4%, respectively. Code will be available at https://github.com/WhiteDOU/GaitMPL https://github.com/WhiteDOU/GaitMPL", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIP.2022.3164543", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "deng2023human-tmm-2024", "citationKey": "deng2023human", "archiveKey": "deng2023human", "method": "Gait Dynamics", "title": "Human Gait Recognition Based on Frontal-View Sequences Using Gait Dynamics and Deep Learning", "authors": "Muqing Deng and Zhuyao Fan and Peng Lin and Xiaoreng Feng", "year": 2024, "venue": "TMM", "motivation": "Accurate frontal-view gait recognition", "idea": "Capture and compare gait dynamics", "techniques": "Kinematic/spatial/area features, dynamic RBF modeling, similarity error computation, deep feature fusion", "inputs": "Silhouette, RGB", "backbone": "CNN", "loss": "CE", "datasets": "CASIA-B, CMU MoBo", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TMM.2023.3262131", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 179 }, { "source": "Author paper summaries", "sheet": "journals", "row": 81 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Human_Gait_Recognition_Based_on_Frontal-View_Sequences_Using_Gait_Dynamics_and_Deep_Learning.pdf", "authorsSource": "Public scholarly metadata", "abstract": "To facilitate human gait recognition, this paper proposes a new frontal-view gait recognition method using gait dynamics and deep learning. Rather than adopting lateralview parameters as gait features in the literature, we employ improved frontal-view features and classification methods to avoid the recognition rate drops due to the complicated surveillance environment. Specifically, we characterize the binary walking silhouettes with three different kinds of frontal-view gait features, including kinematic features, spatial ratio features and area features. In addition, we capture the gait dynamics underlying the time-varying gait features to reflect temporal dynamics information of human walking. Furthermore, we incorporate the deep feature learning information into the recognition procedure to take advantage of the deep learning technique. To obtain the optimal recognition accuracy and robustness performance against walking condition variations, we calculate the similarity between the appearing test gait dynamics and the trained gait dynamics, and propose an error-based feature fusion scheme for gait recognition. To validate the efficacy of the proposed method, we conduct experiments on published gait databases by comparing with other existing gait recognition methods.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TMM.2023.3262131", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "wang2023gait-tcsvt-2024", "citationKey": "wang2023gait", "archiveKey": "wang2023gait", "method": "MSGR", "title": "Gait Recognition With Multi-Level Skeleton-Guided Refinement", "authors": "Runsheng Wang and Yuxuan Shi and Hefei Ling and Zongyi Li and Chengxin Zhao and Bohao Wei and He Li and Ping Li", "year": 2024, "venue": "TMM", "motivation": "Overcome limitations of single modality", "idea": "Refine gait with multi-level skeleton guidance", "techniques": "Frame refinement, cross-modal interaction, part enhancement, feature fusion", "inputs": "Silhouette, Skeleton", "backbone": "CNN, GCN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, GREW", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TMM.2023.3323887", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 145 }, { "source": "Author paper summaries", "sheet": "journals", "row": 51 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Gait_Recognition_With_Multi-Level_Skeleton-Guided_Refinement.pdf", "notes": "The archive venue was corrected from TCSVT to TMM against the published PDF.", "authorsSource": "Public scholarly metadata", "abstract": "Existing methods combining skeleton and silhouette representations demonstrate explicit effectiveness for gait recognition. However, current related methods simply combine the video-level representations of model-based skeleton data and gait silhouettes for retrieval. Therefore, diverse skeleton information is not fully exploited in existing related works: Firstly, the position and movement of bones are not clear from individual silhouettes. This indicates that the frame-level interaction between features of skeletons and silhouettes is critical, which is ignored by previous methods. Secondly, diverse part-level skeleton-guided gait features are not fully captured in existing related approaches. To solve the above issues, we present a novel framework with multi-level skeleton-guided refinement, including frame-level, part-level, and video-level skeleton-guided refinement, for comprehensive skeleton-aided gait representation learning. First, two modules are proposed for frame-level skeleton-guided refinement. Specifically, Visual Skeleton Enhanced Backbone (VSEB) is proposed to visually highlight the global and part-level skeleton regions for the feature of each silhouette frame. Moreover, Cross-Visual-Model Frame-level Interaction (CVMFI) is proposed to further transfer the model-based skeleton information to features of the visual modalities. Secondly, part-level visual and model-based skeleton features are utilized to refine the final gait representation. Concretely, in VSEB, Part Skeleton Enhance Network (PSEN) is proposed to visually enhance the position and movement of part-level skeletons. In addition, Semantic Part Pooling (SPP) is proposed for capturing the model-based skeleton features of different semantic parts. Finally, as the video-level skeleton-guided refinement, multi-modal video-level features are combined to boost the final recognition performance. Extensive experimental results on prevailing datasets demonstrate that our approach outperforms most existing methods, including the skeleton-aided multi-modal methods. With the multi-level refinement guided by the skeleton modalities, the framework is expected to provide a deeper understanding of skeleton-aided gait recognition.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TMM.2023.3323887", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "habib2024watch-wacv-2024", "citationKey": "habib2024watch", "archiveKey": "habib2024watch", "method": "GOUDA", "title": "Watch where you head: A view-biased domain gap in gait recognition and unsupervised adaptation", "authors": "Habib, Gavriel and Barzilay, Noa and Shimshi, Or and Ben-Ari, Rami and Darshan, Nir", "year": 2024, "venue": "WACV", "motivation": "Address view bias in UDA", "idea": "Triplet selection with curriculum learning", "techniques": "Triplet selection, self-supervised learning, curriculum adaptation", "inputs": "Silhouette", "backbone": "", "loss": "Triplet", "datasets": "CASIA-B, OU-MVLP, GREW, Gait3D", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/WACV2024/html/Habib_Watch_Where_You_Head_A_View-Biased_Domain_Gap_in_Gait_WACV_2024_paper.html", "bibtex": "@inproceedings{habib2024watch,\n title = {Watch where you head: A view-biased domain gap in gait recognition and unsupervised adaptation},\n author = {Habib, Gavriel and Barzilay, Noa and Shimshi, Or and Ben-Ari, Rami and Darshan, Nir},\n year = {2024},\n booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},\n pages = {6109--6119}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 90 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait Recognition is a computer vision task aiming to identify people by their walking patterns. Although existing methods often show high performance on specific datasets, they lack the ability to generalize to unseen scenarios. Unsupervised Domain Adaptation (UDA) tries to adapt a model, pre-trained in a supervised manner on a source domain, to an unlabelled target domain. There are only a few works on UDA for gait recognition proposing solutions to limited scenarios. In this paper, we reveal a fundamental phenomenon in adaptation of gait recognition models, caused by the bias in the target domain to viewing angle or walking direction. We then suggest a remedy to reduce this bias with a novel triplet selection strategy combined with curriculum learning. To this end, we present Gait Orientation-based method for Unsupervised Domain Adaptation (GOUDA). We provide extensive experiments on four widely-used gait datasets, CASIA-B, OU-MVLP , GREW, and Gait3D, and on three backbones, GaitSet, Gait- Part, and GaitGL, justifying the view bias and showing the superiority of our proposed method over prior UDA works.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/WACV2024/html/Habib_Watch_Where_You_Head_A_View-Biased_Domain_Gap_in_Gait_WACV_2024_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "gupta2024you-wacv-2024", "citationKey": "gupta2024you", "archiveKey": "gupta2024you", "method": "", "title": "You can run but not hide: Improving gait recognition with intrinsic occlusion type awareness", "authors": "Gupta, Ayush and Chellappa, Rama", "year": 2024, "venue": "WACV", "motivation": "Tackle gait occlusion problem", "idea": "Learn occlusion types for recognition", "techniques": "Occlusion detection, feature guidance, model-agnostic integration", "inputs": "Silhouette", "backbone": "CNN", "loss": "", "datasets": "GREW, BRIAR", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/WACV2024/html/Gupta_You_Can_Run_but_Not_Hide_Improving_Gait_Recognition_With_WACV_2024_paper.html", "bibtex": "@inproceedings{gupta2024you,\n title = {You can run but not hide: Improving gait recognition with intrinsic occlusion type awareness},\n author = {Gupta, Ayush and Chellappa, Rama},\n year = {2024},\n booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},\n pages = {5893--5902}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 89 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "While gait recognition has seen many advances in recent years, the occlusion problem has largely been ignored. This problem is especially important for gait recognition from uncontrolled outdoor sequences at range - since any small obstruction can affect the recognition system. Most current methods assume the availability of complete body information while extracting the gait features. When parts of the body are occluded, these methods may hallucinate and output a corrupted gait signature as they try to look for body parts which are not present in the input at all. To address this, we exploit the learned occlusion type while extracting identity features from videos. Thus, in this work, we propose an occlusion aware gait recognition method which can be used to model intrinsic occlusion awareness into potentially any state-of-the-art gait recognition method. Our experiments on the challenging GREW and BRIAR datasets show that networks enhanced with this occlusion awareness perform better at recognition tasks than their counterparts trained on similar occlusions.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/WACV2024/html/Gupta_You_Can_Run_but_Not_Hide_Improving_Gait_Recognition_With_WACV_2024_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-ma2024gaiteditor", "citationKey": "ma2024gaiteditor", "archiveKey": "ma2024gaiteditor", "method": "GaitEditor", "title": "Pedestrian Attribute Editing for Gait Recognition and Anonymization", "authors": "Ma, Jingzhe and Ye, Dingqiang and Fan, Chao and Yu, Shiqi", "year": 2024, "venue": "arXiv", "motivation": "Gait data editing must balance attribute control, recognition and anonymity.", "idea": "Edit pedestrian attributes while studying their effects on gait identity.", "techniques": "Generative attribute editing; gait-aware analysis; anonymization.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://arxiv.org/abs/2303.05076v2", "bibtex": "@article{ma2024gaiteditor,\n title = {Pedestrian Attribute Editing for Gait Recognition and Anonymization},\n author = {Ma, Jingzhe and Ye, Dingqiang and Fan, Chao and Yu, Shiqi},\n year = {2024},\n journal = {arXiv preprint arXiv:2303.05076},\n url = {https://arxiv.org/abs/2303.05076v2}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Privacy Concerns and Potential Solutions", "Advances in Methods / Synthetic Gait Data and Generative Learning" ], "abstract": "GaitEditor modifies multiple attributes in real walking sequences while preserving plausible visual appearance. The same editing framework serves two objectives: producing training variations for recognition and reducing identifying information for anonymization. Experiments across three gait benchmarks evaluate both uses, examining how controlled attribute changes can enhance or weaken identity matching.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2303.05076", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "wang2023landmarkgait-acmmm-2023", "citationKey": "wang2023landmarkgait", "archiveKey": "wang2023landmarkgait", "method": "LandmarkGait", "title": "LandmarkGait: intrinsic human parsing for gait recognition", "authors": "Wang, Zengbin and Hou, Saihui and Zhang, Man and Liu, Xu and Cao, Chunshui and Huang, Yongzhen and Xu, Shibiao", "year": 2023, "venue": "ACM MM", "motivation": "Improve gait recognition accuracy", "idea": "Unsupervised landmark-based parsing", "techniques": "Landmark discovery, landmark grouping, multi-scale feature extraction", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, FVG, CCVID, CCPG", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1145/3581783.3611840", "bibtex": "@inproceedings{wang2023landmarkgait,\n title = {LandmarkGait: intrinsic human parsing for gait recognition},\n author = {Wang, Zengbin and Hou, Saihui and Zhang, Man and Liu, Xu and Cao, Chunshui and Huang, Yongzhen and Xu, Shibiao},\n year = {2023},\n booktitle = {Proceedings of the 31st ACM International Conference on Multimedia},\n pages = {2305--2314}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 78 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait recognition is an emerging biometric technology for identifying pedestrians based on their unique walking patterns. In past gait recognition, global-based methods are inadequate to meet the growing demand for accuracy, while commonly used part-based methods provided coarse and inaccurate feature representation for specific body parts. Human parsing appears to be a better option for accurately representing specific and complete body parts in gait recognition. However, its practical application in gait recognition is often hindered by missing RGB modality, lack of annotated body parts, and difficulty in balancing parsing quantity and quality. To address this issue, we proposeLandmarkGait, an accessible and alternative parsing-based solution for gait recognition. LandmarkGait introduces an unsupervised landmark discovery network to transform the dense silhouette into a finite set of landmarks with remarkable consistency across various conditions. By grouping landmarks subsets corresponding to distinct body part regions, following a reconstruction task and further refinement from high-quality input silhouettes, we can directly obtain fine-grained parsing results from original binary silhouettes in an unsupervised manner. Moreover, we also develop a multi-scale feature extractor that simultaneously captures global and parsing feature representations based on the integrity and flexibility of specific body parts. Extensive experiments demonstrate that our LandmarkGait can extract more stable features and exhibit significant performance improvement under all conditions, especially in various dressing conditions. Code is available at https://github.com/wzb-bupt/LandmarkGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1145/3581783.3611840", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "zheng2023parsing-acmmm-2023", "citationKey": "zheng2023parsing", "archiveKey": "zheng2023parsing", "method": "ParsingGait", "title": "Parsing is all you need for accurate gait recognition in the wild", "authors": "Zheng, Jinkai and Liu, Xinchen and Wang, Shuai and Wang, Lihao and Yan, Chenggang and Liu, Wu", "year": 2023, "venue": "ACM MM", "motivation": "Improve gait recognition robustness", "idea": "Use fine-grained part-level representation", "techniques": "Human parsing sequences, CNN backbone, cross-part GCN integration", "inputs": "Silhouette, Parsing", "backbone": "CNN, GCN", "loss": "CE, Triplet", "datasets": "Gait3D", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1145/3581783.3612052", "bibtex": "@inproceedings{zheng2023parsing,\n title = {Parsing is all you need for accurate gait recognition in the wild},\n author = {Zheng, Jinkai and Liu, Xinchen and Wang, Shuai and Wang, Lihao and Yan, Chenggang and Liu, Wu},\n year = {2023},\n booktitle = {Proceedings of the 31st ACM International Conference on Multimedia},\n pages = {116--124}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 77 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Ideal Gait Modalities" ], "abstract": "Binary silhouettes and keypoint-based skeletons have dominated human gait recognition studies for decades since they are easy to extract from video frames. Despite their success in gait recognition for in-the-lab environments, they usually fail in real-world scenarios due to their low information entropy for gait representations. To achieve accurate gait recognition in the wild, this paper presents a novel gait representation, named Gait Parsing Sequence (GPS) . GPSs are sequences of fine-grained human segmentation, i.e., human parsing, extracted from video frames, so they have much higher information entropy to encode the shapes and dynamics of finegrained human parts during walking. Moreover, to effectively explore the capability of the GPS representation, we propose a novel human parsing-based gait recognition framework, named ParsingGait. ParsingGait contains a Convolutional Neural Network (CNN)-based backbone and two light-weighted heads. The first head extracts global semantic features from GPSs, while the other one learns mutual information of part-level features through Graph Convolutional Networks to model the detailed dynamics of human walking. Furthermore, due to the lack of suitable datasets, we build the first parsing-based dataset for gait recognition in the wild, named Gait3D-Parsing, by extending the large-scale and challenging Gait3D dataset. Based on Gait3D-Parsing, we comprehensively evaluate our method and existing gait recognition methods. Specifically, ParsingGait achieves a17.5% Rank-1 increase compared with the state-of-the-art silhouette-based method. In addition, by replacing silhouettes with GPSs, current gait recognition methods achieve about 12.5% ∼ 19.2% improvements in Rank-1 accuracy. The experimental results show a significant improvement in accuracy brought by the GPS representation and the superiority of ParsingGait. The code and dataset are available at https://gait3d.github.io/gait3d-parsing-hp.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1145/3581783.3612052", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "wang2023causal-acmmm-2023", "citationKey": "wang2023causal", "archiveKey": "wang2023causal", "method": "SparseView", "title": "Causal intervention for sparse-view gait recognition", "authors": "Wang, Jilong and Hou, Saihui and Huang, Yan and Cao, Chunshui and Liu, Xu and Huang, Yongzhen and Wang, Liang", "year": 2023, "venue": "ACM MM", "motivation": "Address sparse-view gait recognition", "idea": "Causal modeling and view clustering", "techniques": "Confounder clustering, backdoor-adjusted metric learning", "inputs": "Silhouette", "backbone": "", "loss": "Triplet with CI, CE", "datasets": "CASIA-B, OU-MVLP, GREW, Gait3D", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1145/3581783.3612124", "bibtex": "@inproceedings{wang2023causal,\n title = {Causal intervention for sparse-view gait recognition},\n author = {Wang, Jilong and Hou, Saihui and Huang, Yan and Cao, Chunshui and Liu, Xu and Huang, Yongzhen and Wang, Liang},\n year = {2023},\n booktitle = {Proceedings of the 31st ACM International Conference on Multimedia},\n pages = {77--85}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 79 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait recognition aims at identifying individuals by unique walking patterns at a long distance. However, prevailing methods suffer from a large degradation when applied to large-scale surveillance systems. We find a significant cause of this issue is that previous methods heavily rely on full-view person annotations to reduce view differences by pulling closer the anchor to positive samples from different viewpoints. But, subjects under in-the-wild scenarios usually have only a limited number of sequences from different viewpoints. As a result, the available viewpoints of each subject are sparse compared to the whole dataset, and simply minimizing intra-identity differences cannot well reducing the view differences in the whole dataset. In this work, we formulate this overlooked problem as Sparse-View Gait Recognition and provide a comprehensive analysis of it by a Structural Causal Model for causalities among latent features, view distribution, and labels. Based on our analysis, we propose a simple yet effective method that enables networks to learn a more robust representation among different views. Specifically, our method consists of two parts: 1) an effective metric learning algorithmic implementation based on the backdoor adjustment, which improves the consistency of representations among different views; 2) an unsupervised view cluster algorithm to discover and identify the most influential view contexts. We evaluate the effectiveness of our method on popular GREW, Gait3D, CASIA-B, and OU-MVLP, showing that our method consistently outperforms baselines and achieves state-of-the-art performance. The code will be available at https://github.com/wj1tr0y/GaitCSV.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1145/3581783.3612124", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-rollo2023followme", "citationKey": "rollo2023followme", "archiveKey": "rollo2023followme", "method": "", "title": "FollowMe: A Robust Person Following Framework Based on Visual Re-Identification and Gestures", "authors": "Rollo, Federico and Zunino, Andrea and Raiola, Gennaro and Amadio, Fabio and Ajoudani, Arash and Tsagarakis, Nikolaos", "year": 2023, "venue": "ARSO", "motivation": "Robots must maintain a target identity while interpreting user interaction.", "idea": "Integrate visual re-identification and gesture understanding into person following.", "techniques": "Person ReID; gesture recognition; robot navigation.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/ARSO56563.2023.10187536", "bibtex": "@inproceedings{rollo2023followme,\n title = {FollowMe: A Robust Person Following Framework Based on Visual Re-Identification and Gestures},\n author = {Rollo, Federico and Zunino, Andrea and Raiola, Gennaro and Amadio, Fabio and Ajoudani, Arash and Tsagarakis, Nikolaos},\n year = {2023},\n booktitle = {IEEE International Conference on Advanced Robotics and Its Social Impacts},\n pages = {84--89},\n doi = {10.1109/ARSO56563.2023.10187536}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Challenges and Outlooks / Gait Recognition for Agents" ], "abstract": "FollowMe combines visual re-identification, gesture recognition and autonomous navigation so a mobile robot can follow a designated person. The system learns the target's appearance and uses it for subsequent re-identification, while the navigation component avoids obstacles. Laboratory experiments include unfamiliar moving obstacles to test the integrated behavior.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2311.12992", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "li2023depth-cvpr-2023", "citationKey": "li2023depth", "archiveKey": "li2023depth", "method": "AUG-OGBase", "title": "An in-depth exploration of person re-identification and gait recognition in cloth-changing conditions", "authors": "Li, Weijia and Hou, Saihui and Zhang, Chunjie and Cao, Chunshui and Liu, Xu and Huang, Yongzhen and Zhao, Yao", "year": 2023, "venue": "CVPR", "motivation": "Address cloth variation challenge", "idea": "Fair cloth-changing comparison", "techniques": "Cloth-changing benchmark", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CCPG", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2023/html/Li_An_In-Depth_Exploration_of_Person_Re-Identification_and_Gait_Recognition_in_CVPR_2023_paper.html", "bibtex": "@inproceedings{li2023depth,\n title = {An in-depth exploration of person re-identification and gait recognition in cloth-changing conditions},\n author = {Li, Weijia and Hou, Saihui and Zhang, Chunjie and Cao, Chunshui and Liu, Xu and Huang, Yongzhen and Zhao, Yao},\n year = {2023},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n pages = {13824--13833}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 26 }, { "source": "Manuscript Table I" }, { "source": "Verified publication metadata" } ], "summaryStatus": "Manuscript summary", "sections": [ "Background / Historical Evolution", "Background / Broader Applications", "Advances in Benchmarks / Popular Datasets", "Advances in Methods / Handling Real-World Noise" ], "abstract": "The target of person re-identification (ReID) and gait recognition is consistent, that is to match the target pedestrian under surveillance cameras. For the cloth-changing problem, video-based ReID is rarely studied due to the lack of a suitable cloth-changing benchmark, and gait recognition is often researched under controlled conditions. To tackle this problem, we propose a Cloth-Changing benchmark for Person re-identification and Gait recognition (CCPG). It is a cloth-changing dataset, and there are several highlights in CCPG, (1) it provides 200 identities and over 16K sequences are captured indoors and outdoors, (2) each identity has seven different cloth-changing statuses, which is hardly seen in previous datasets, (3) RGB and silhouettes version data are both available for research purposes. Moreover, aiming to investigate the cloth-changing problem systematically, comprehensive experiments are conducted on video-based ReID and gait recognition methods. The experimental results demonstrate the superiority of ReID and gait recognition separately in different cloth-changing conditions and suggest that gait recognition is a potential solution for addressing the cloth-changing problem. Our dataset will be available at https://github.com/BNU-IVC/CCPG.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2023/html/Li_An_In-Depth_Exploration_of_Person_Re-Identification_and_Gait_Recognition_in_CVPR_2023_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ma2023dynamic-cvpr-2023", "citationKey": "ma2023dynamic", "archiveKey": "ma2023dynamic", "method": "DANet", "title": "Dynamic aggregated network for gait recognition", "authors": "Ma, Kang and Fu, Ying and Zheng, Dezhi and Cao, Chunshui and Hu, Xuecai and Huang, Yongzhen", "year": 2023, "venue": "CVPR", "motivation": "Robust gait under variations", "idea": "Local-global motion modeling", "techniques": "Dynamic aggregated motion network", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, Gait3D", "codeAvailable": "Yes", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2023/html/Ma_Dynamic_Aggregated_Network_for_Gait_Recognition_CVPR_2023_paper.html", "bibtex": "@inproceedings{ma2023dynamic,\n title = {Dynamic aggregated network for gait recognition},\n author = {Ma, Kang and Fu, Ying and Zheng, Dezhi and Cao, Chunshui and Hu, Xuecai and Huang, Yongzhen},\n year = {2023},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n pages = {22076--22085}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 27 }, { "source": "Manuscript Table I" }, { "source": "Verified publication metadata" } ], "summaryStatus": "Manuscript summary", "sections": [ "Background / Historical Evolution", "Advances in Methods / Strong Gait Backbones" ], "abstract": "Gait recognition is beneficial for a variety of applications, including video surveillance, crime scene investigation, and social security, to mention a few. However, gait recognition often suffers from multiple exterior factors in real scenes, such as carrying conditions, wearing overcoats, and diverse viewing angles. Recently, various deep learning-based gait recognition methods have achieved promising results, but they tend to extract one of the salient features using fixed-weighted convolutional networks, do not well consider the relationship within gait features in key regions, and ignore the aggregation of complete motion patterns. In this paper, we propose a new perspective that actual gait features include global motion patterns in multiple key regions, and each global motion pattern is composed of a series of local motion patterns. To this end, we propose a Dynamic Aggregation Network (DANet) to learn more discriminative gait features. Specifically, we create a dynamic attention mechanism between the features of neighboring pixels that not only adaptively focuses on key regions but also generates more expressive local motion patterns. In addition, we develop a self-attention mechanism to select representative local motion patterns and further learn robust global motion patterns. Extensive experiments on three popular public gait datasets, i.e., CASIA- B, OUMVLP , and Gait3D, demonstrate that the proposed method can provide substantial improvements over the current state-of-the-art methods.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2023/html/Ma_Dynamic_Aggregated_Network_for_Gait_Recognition_CVPR_2023_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "fan2023opengait-cvpr-2023", "citationKey": "fan2023opengait", "archiveKey": "fan2023opengait", "method": "GaitBase", "title": "Opengait: Revisiting gait recognition towards better practicality", "authors": "Fan, Chao and Liang, Junhao and Shen, Chuanfu and Hou, Saihui and Huang, Yongzhen and Yu, Shiqi", "year": 2023, "venue": "CVPR", "motivation": "Flexible, extensible gait codebase", "idea": "Revisiting toward better practicality", "techniques": "comprehensive benchmark study", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, GREW, Gait3D", "codeAvailable": "OpenGait", "codeUrl": "", "url": "https://doi.org/10.1109/cvpr52729.2023.00936", "bibtex": "@inproceedings{fan2023opengait,\n title = {Opengait: Revisiting gait recognition towards better practicality},\n author = {Fan, Chao and Liang, Junhao and Shen, Chuanfu and Hou, Saihui and Huang, Yongzhen and Yu, Shiqi},\n year = {2023},\n booktitle = {Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},\n pages = {9707--9716}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 25 }, { "source": "Manuscript Table I" }, { "source": "Verified publication metadata" } ], "summaryStatus": "Manuscript summary", "sections": [ "Background / Task Definition", "Background / Historical Evolution", "Advances in Benchmarks / Public Codebases", "Advances in Benchmarks / Research Projects", "Advances in Methods / Strong Gait Backbones", "Advances in Methods / Gait Learning Frameworks" ], "abstract": "Gait recognition is one of the most critical long-distance identification technologies and increasingly gains popularity in both research and industry communities. Despite the significant progress made in indoor datasets, much evidence shows that gait recognition techniques perform poorly in the wild. More importantly, we also find that some conclusions drawn from indoor datasets cannot be generalized to real applications. Therefore, the primary goal of this paper is to present a comprehensive benchmark study for better practicality rather than only a particular model for better performance. To this end, we first develop a flexible and efficient gait recognition codebase named OpenGait. Based on OpenGait, we deeply revisit the recent development of gait recognition by re-conducting the ablative experiments. Encouragingly,we detect some unperfect parts of certain prior woks, as well as new insights. Inspired by these discoveries, we develop a structurally simple, empirically powerful, and practically robust baseline model, Gait- Base. Experimentally, we comprehensively compare Gait- Base with many current gait recognition methods on multiple public datasets, and the results reflect that GaitBase achieves significantly strong performance in most cases regardless of indoor or outdoor situations. Code is available at https://github.com/ShiqiYu/OpenGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/cvpr52729.2023.00936", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "dou2023gaitgci-cvpr-2023", "citationKey": "dou2023gaitgci", "archiveKey": "dou2023gaitgci", "method": "GaitGCI", "title": "Gaitgci: Generative counterfactual intervention for gait recognition", "authors": "Dou, Huanzhang and Zhang, Pengyi and Su, Wei and Yu, Yunlong and Lin, Yining and Li, Xi", "year": 2023, "venue": "CVPR", "motivation": "Alleviate confounders in gait", "idea": "Counterfactual intervention", "techniques": "Causal inference and adaptive conv", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, Gait3D", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2023/html/Dou_GaitGCI_Generative_Counterfactual_Intervention_for_Gait_Recognition_CVPR_2023_paper.html", "bibtex": "@inproceedings{dou2023gaitgci,\n title = {Gaitgci: Generative counterfactual intervention for gait recognition},\n author = {Dou, Huanzhang and Zhang, Pengyi and Su, Wei and Yu, Yunlong and Lin, Yining and Li, Xi},\n year = {2023},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n pages = {5578--5588}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 24 }, { "source": "Manuscript Table I" }, { "source": "Verified publication metadata" } ], "summaryStatus": "Manuscript summary", "sections": [ "Background / Historical Evolution", "Advances in Methods / Gait Learning Frameworks", "Advances in Methods / Handling Real-World Noise" ], "abstract": "Gait is one of the most promising biometrics that aims to identify pedestrians from their walking patterns. However, prevailing methods are susceptible to confounders, resulting in the networks hardly focusing on the regions that reflect effective walking patterns. To address this fundamental problem in gait recognition, we propose a Generative Counterfactual Intervention framework, dubbed GaitGCI, consisting of Counterfactual Intervention Learning (CIL) and Diversity-Constrained Dynamic Convolution (DCDC). CIL eliminates the impacts of confounders by maximizing the likelihood difference between factual/counterfactual attention while DCDC adaptively generates sample-wise factual/counterfactual attention to efficiently perceive the sample-wise properties. With matrix decomposition and diversity constraint, DCDC guarantees the model to be efficient and effective. Extensive experiments indicate that proposed GaitGCI: 1) could effectively focus on the discriminative and interpretable regions that reflect gait pattern; 2) is model-agnostic and could be plugged into existing models to improve performance with nearly no extra cost; 3) efficiently achieves state-of-the-art performance on arbitrary scenarios (in-the-lab and in-the-wild).", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2023/html/Dou_GaitGCI_Generative_Counterfactual_Intervention_for_Gait_Recognition_CVPR_2023_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "shen2023lidargait-cvpr-2023", "citationKey": "shen2023lidargait", "archiveKey": "shen2023lidargait", "method": "LidarGait", "title": "Lidargait: Benchmarking 3d gait recognition with point clouds", "authors": "Shen, Chuanfu and Fan, Chao and Wu, Wei and Wang, Rui and Huang, George Q and Yu, Shiqi", "year": 2023, "venue": "CVPR", "motivation": "Overcome 2D visual ambiguities", "idea": "Large-scale LiDAR benchmarking", "techniques": "Projection-based baseline", "inputs": "Point", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "SUSTech1K", "codeAvailable": "OpenGait", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2023/html/Shen_LidarGait_Benchmarking_3D_Gait_Recognition_With_Point_Clouds_CVPR_2023_paper.html", "bibtex": "@inproceedings{shen2023lidargait,\n title = {Lidargait: Benchmarking 3d gait recognition with point clouds},\n author = {Shen, Chuanfu and Fan, Chao and Wu, Wei and Wang, Rui and Huang, George Q and Yu, Shiqi},\n year = {2023},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n pages = {1054--1063}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 28 }, { "source": "Manuscript Table I" }, { "source": "Verified publication metadata" } ], "summaryStatus": "Manuscript summary", "sections": [ "Background / Task Definition", "Background / Historical Evolution", "Background / Broader Applications", "Advances in Benchmarks / Popular Datasets", "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Handling Real-World Noise" ], "abstract": "Video-based gait recognition has achieved impressive results in constrained scenarios. However, visual cameras neglect human 3D structure information, which limits the feasibility of gait recognition in the 3D wild world. Instead of extracting gait features from images, this work explores precise 3D gait features from point clouds and proposes a simple yet efficient 3D gait recognition framework, termed LidarGait. Our proposed approach projects sparse point clouds into depth maps to learn the representations with 3D geometry information, which outperforms existing point-wise and camera-based methods by a significant margin. Due to the lack of point cloud datasets, we build the first large-scale LiDAR-based gait recognition dataset, SUSTech1K, collected by a LiDAR sensor and an RGB camera. The dataset contains 25,239 sequences from 1,050 subjects and covers many variations, including visibility, views, occlusions, clothing, carrying, and scenes. Extensive experiments show that (1) 3D structure information serves as a significant feature for gait recognition. (2) LidarGait outperforms existing point-based and silhouettebased methods by a significant margin, while it also offers stable cross-view results. (3) The LiDAR sensor is superior to the RGB camera for gait recognition in the outdoor environment. The source code and dataset have been made available at https://lidargait.github.io.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2023/html/Shen_LidarGait_Benchmarking_3D_Gait_Recognition_With_Point_Clouds_CVPR_2023_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "cui2023multi-cvpr-2023", "citationKey": "cui2023multi", "archiveKey": "cui2023multi", "method": "MMGaitFormer", "title": "Multi-modal gait recognition via effective spatial-temporal feature fusion", "authors": "Cui, Yufeng and Kang, Yimei", "year": 2023, "venue": "CVPR", "motivation": "Complement silhouette and skeleton", "idea": "Multi-modal transformer fusion", "techniques": "Spatial-temporal co-attention fusion", "inputs": "Silhouette, Skeleton", "backbone": "CNN, Transformer", "loss": "Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2023/html/Cui_Multi-Modal_Gait_Recognition_via_Effective_Spatial-Temporal_Feature_Fusion_CVPR_2023_paper.html", "bibtex": "@inproceedings{cui2023multi,\n title = {Multi-modal gait recognition via effective spatial-temporal feature fusion},\n author = {Cui, Yufeng and Kang, Yimei},\n year = {2023},\n booktitle = {Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},\n pages = {17949--17957}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 23 }, { "source": "Manuscript Table I" }, { "source": "Verified publication metadata" } ], "summaryStatus": "Manuscript summary", "sections": [ "Background / Historical Evolution", "Advances in Methods / Ideal Gait Modalities" ], "abstract": "Gait recognition is a biometric technology that identifies people by their walking patterns. The silhouettesbased method and the skeletons-based method are the two most popular approaches. However, the silhouette data are easily affected by clothing occlusion, and the skeleton data lack body shape information. To obtain a more robust and comprehensive gait representation for recognition, we propose a transformer-based gait recognition framework called MMGaitFormer, which effectively fuses and aggregates the spatial-temporal information from the skeletons and silhouettes. Specifically, a Spatial Fusion Module (SFM) and a Temporal Fusion Module (TFM) are proposed for effective spatial-level and temporal-level feature fusion, respectively. The SFM performs fine-grained body parts spatial fusion and guides the alignment of each part of the silhouette and each joint of the skeleton through the attention mechanism. The TFM performs temporal modeling through Cycle Position Embedding (CPE) and fuses temporal information of two modalities. Experiments demonstrate that our MMGaitFormer achieves state-of-the-art performance on popular gait datasets. For the most challenging “CL” (i.e., walking in different clothes) condition in CASIA- B, our method achieves a rank-1 accuracy of 94.8%, which outperforms the state-of-the-art single-modal methods by a large margin.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2023/html/Cui_Multi-Modal_Gait_Recognition_via_Effective_Spatial-Temporal_Feature_Fusion_CVPR_2023_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "wang2023dygait-iccv-2023", "citationKey": "wang2023dygait", "archiveKey": "wang2023dygait", "method": "DyGait", "title": "DyGait: Exploiting Dynamic Representations for High-performance Gait Recognition", "authors": "Wang, Ming and Guo, Xianda and Lin, Beibei and Yang, Tian and Zhu, Zheng and Li, Lincheng and Zhang, Shunli and Yu, Xin", "year": 2023, "venue": "ICCV", "motivation": "Reduce static-region interference", "idea": "Focus on dynamic body parts", "techniques": "Motion extraction, temporal aggregation", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, GREW, Gait3D", "codeAvailable": "No", "codeUrl": "", "url": "https://arxiv.org/abs/2303.14953", "bibtex": "@inproceedings{wang2023dygait,\n title = {DyGait: Exploiting Dynamic Representations for High-performance Gait Recognition},\n author = {Wang, Ming and Guo, Xianda and Lin, Beibei and Yang, Tian and Zhu, Zheng and Li, Lincheng and Zhang, Shunli and Yu, Xin},\n year = {2023},\n booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision},\n pages = {13378--13387},\n url = {https://arxiv.org/abs/2303.14953}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 52 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Strong Gait Backbones", "Advances in Methods / Handling Real-World Noise" ], "abstract": "DyGait emphasizes changing body regions rather than weighting static and dynamic information equally. Its dynamic augmentation module constructs features around motion across silhouette frames. Evaluations on GREW, Gait3D, CASIA-B and OU-MVLP investigate whether explicitly emphasizing these regions improves learned identity representations.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2303.14953", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "fu2023gpgait-iccv-2023", "citationKey": "fu2023gpgait", "archiveKey": "fu2023gpgait", "method": "GPGait", "title": "Gpgait: Generalized pose-based gait recognition", "authors": "Fu, Yang and Meng, Shibei and Hou, Saihui and Hu, Xuecai and Huang, Yongzhen", "year": 2023, "venue": "ICCV", "motivation": "improve pose-based cross-domain gait recognition", "idea": "Generalizable pose-based gait recognition", "techniques": "GCN with human-oriented designs", "inputs": "Skeleton", "backbone": "GCN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, GREW, Gait3D", "codeAvailable": "FastPoseGait", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/ICCV2023/html/Fu_GPGait_Generalized_Pose-based_Gait_Recognition_ICCV_2023_paper.html", "bibtex": "@inproceedings{fu2023gpgait,\n title = {Gpgait: Generalized pose-based gait recognition},\n author = {Fu, Yang and Meng, Shibei and Hou, Saihui and Hu, Xuecai and Huang, Yongzhen},\n year = {2023},\n booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision},\n pages = {19595--19604}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 49 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Background / Task Definition", "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Strong Gait Backbones", "Advances in Methods / Handling Real-World Noise" ], "abstract": "Recent works on pose-based gait recognition have demonstrated the potential of using such simple information to achieve results comparable to silhouette-based methods. However, the generalization ability of pose-based methods on different datasets is undesirably inferior to that of silhouette-based ones, which has received little attention but hinders the application of these methods in real-world scenarios. To improve the generalization ability of pose-based methods across datasets, we propose a Generalized Pose-based Gait recognition (GPGait) framework. First, a Human-Oriented Transformation (HOT) and a series of Human-Oriented Descriptors (HOD) are proposed to obtain a unified pose representation with discriminative multi-features. Then, given the slight variations in the unified representation after HOT and HOD, it becomes crucial for the network to extract local-global relationships between the keypoints. To this end, a Part-Aware Graph Convolutional Network (PAGCN) is proposed to enable efficient graph partition and local-global spatial feature extraction. Experiments on four public gait recognition datasets, CASIA-B, OUMVLP-Pose, Gait3D and GREW, show that our model demonstrates better and more stable cross-domain capabilities compared to existing skeletonbased methods, achieving comparable recognition results to silhouette-based ones. Code is available at https://github.com/BNU-IVC/FastPoseGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/ICCV2023/html/Fu_GPGait_Generalized_Pose-based_Gait_Recognition_ICCV_2023_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "wang2023hierarchical-iccv-2023", "citationKey": "wang2023hierarchical", "archiveKey": "wang2023hierarchical", "method": "HSTL", "title": "Hierarchical spatio-temporal representation learning for gait recognition", "authors": "Wang, Lei and Liu, Bo and Liang, Fangfang and Wang, Bincheng", "year": 2023, "venue": "ICCV", "motivation": "Capture body-part motion hierarchy", "idea": "Learn hierarchical motion patterns", "techniques": "Region partitioning, spatio-temporal pooling", "inputs": "Silhouette", "backbone": "CNN, Attention", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, GREW, Gait3D", "codeAvailable": "Yes", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/ICCV2023/html/Wang_Hierarchical_Spatio-Temporal_Representation_Learning_for_Gait_Recognition_ICCV_2023_paper.html", "bibtex": "@inproceedings{wang2023hierarchical,\n title = {Hierarchical spatio-temporal representation learning for gait recognition},\n author = {Wang, Lei and Liu, Bo and Liang, Fangfang and Wang, Bincheng},\n year = {2023},\n booktitle = {2023 IEEE/CVF International Conference on Computer Vision (ICCV)},\n pages = {19582--19592}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 53 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Strong Gait Backbones" ], "abstract": "HSTL organizes gait features from whole-body structure to finer local motion. Hierarchical clustering defines body partitions, adaptive region extractors learn their dynamics, and spatiotemporal pooling combines levels. Temporal downsampling reduces redundant frames. Four benchmark evaluations examine whether this hierarchy improves recognition while keeping model complexity manageable.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://openaccess.thecvf.com/content/ICCV2023/html/Wang_Hierarchical_Spatio-Temporal_Representation_Learning_for_Gait_Recognition_ICCV_2023_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "guo2023physics-iccv-2023", "citationKey": "guo2023physics", "archiveKey": "guo2023physics", "method": "PAA", "title": "Physics-augmented autoencoder for 3d skeleton-based gait recognition", "authors": "Guo, Hongji and Ji, Qiang", "year": 2023, "venue": "ICCV", "motivation": "Robust 3D gait recognition", "idea": "Physics-based 3D gait modeling", "techniques": "Spatial-temporal GCN, Lagrangian dynamics decoder", "inputs": "Skeleton", "backbone": "GCN, RNN", "loss": "CE, Triplet", "datasets": "GREW, Gait3D, KinectGait", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/ICCV2023/html/Guo_Physics-Augmented_Autoencoder_for_3D_Skeleton-Based_Gait_Recognition_ICCV_2023_paper.html", "bibtex": "@inproceedings{guo2023physics,\n title = {Physics-augmented autoencoder for 3d skeleton-based gait recognition},\n author = {Guo, Hongji and Ji, Qiang},\n year = {2023},\n booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision},\n pages = {19627--19638}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 50 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "In this paper, we introduce physics-augmented autoencoder (PAA) framework for 3D skeleton-based human gait recognition. Specifically, we construct the autoencoder with a graph-convolution-based encoder and a physics-based decoder. The encoder takes the skeleton sequence as input and produces the generalized positions and forces of each joint, which are taken by the decoder to reconstruct the input skeleton based on the Lagrangian dynamics. In this way, the intermediate representations are physically plausible and discriminative. During the inference, the decoder is discared and a RNN-based classifier takes the output of the encoder for gait recognition. We evaluated our proposed method on three benchmark datasets including Gait3D, GREW, and KinectGait. Our method achieves state-ofthe-art performance for 3D skeleton-based gait recognition. Furthermore, extensive ablation studies show that our method generalizes better and is more robust with smallscale training data by incorporating the physics knowledge. We also validated the physical plausibility of the intermediate representations by making force predictions on real data with physical annotations.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/ICCV2023/html/Guo_Physics-Augmented_Autoencoder_for_3D_Skeleton-Based_Gait_Recognition_ICCV_2023_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ma2023fine-iccv-2023", "citationKey": "ma2023fine", "archiveKey": "ma2023fine", "method": "UDA", "title": "Fine-grained unsupervised domain adaptation for gait recognition", "authors": "Ma, Kang and Fu, Ying and Zheng, Dezhi and Peng, Yunjie and Cao, Chunshui and Huang, Yongzhen", "year": 2023, "venue": "ICCV", "motivation": "Difficulty labeling long-distance gait data", "idea": "Unsupervised domain adaptation for gait", "techniques": "Cross-view linking, motion modeling", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet, Contrastive", "datasets": "CASIA-B, OU-MVLP, GREW", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/ICCV2023/html/Ma_Fine-grained_Unsupervised_Domain_Adaptation_for_Gait_Recognition_ICCV_2023_paper.html", "bibtex": "@inproceedings{ma2023fine,\n title = {Fine-grained unsupervised domain adaptation for gait recognition},\n author = {Ma, Kang and Fu, Ying and Zheng, Dezhi and Peng, Yunjie and Cao, Chunshui and Huang, Yongzhen},\n year = {2023},\n booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision},\n pages = {11313--11322}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 51 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Gait Learning Frameworks" ], "abstract": "Gait recognition has emerged as a promising technique for the long-range retrieval of pedestrians, providing numerous advantages such as accurate identification in challenging conditions and non-intrusiveness, making it highly desirable for improving public safety and security. However, the high cost of labeling datasets, which is a prerequisite for most existing fully supervised approaches, poses a significant obstacle to the development of gait recognition. Recently, some unsupervised methods for gait recognition have shown promising results. However, these methods mainly rely on a fine-tuning approach that does not sufficiently consider the relationship between source and target domains, leading to the catastrophic forgetting of source domain knowledge. This paper presents a novel perspective that adjacent-view sequences exhibit overlapping views, which can be leveraged by the network to gradually attain cross-view and cross-dressing capabilities without pre-training on the labeled source domain. Specifically, we propose a fine-grained Unsupervised Domain Adaptation (UDA) framework that iteratively alternates between two stages. The initial stage involves offline clustering, which transfers knowledge from the labeled source domain to the unlabeled target domain and adaptively generates pseudo-labels according to the expressiveness of each part. Subsequently, the second stage encompasses online training, which further achieves cross-dressing capabilities by continuously learning to distinguish numerous features of source and target domains. The effectiveness of the proposed method is demonstrated through extensive experiments conducted on widely-used public gait datasets.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/ICCV2023/html/Ma_Fine-grained_Unsupervised_Domain_Adaptation_for_Gait_Recognition_ICCV_2023_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-agreid2023", "citationKey": "agreid2023", "archiveKey": "agreid2023", "method": "", "title": "AG-ReID 2023: Aerial-Ground Person Re-identification Challenge Results", "authors": "Nguyen, Kien and Fookes, Clinton and Sridharan, Sridha and Liu, Feng and Liu, Xiaoming and Ross, Arun and Michalski, Dana and Nguyen, Huy and others", "year": 2023, "venue": "IJCB", "motivation": "Aerial and ground observations create a substantial person-recognition domain gap.", "idea": "Benchmark cross-view person re-identification through a shared challenge.", "techniques": "Aerial-ground protocols; competition evaluation; method comparison.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/IJCB57857.2023.10448780", "bibtex": "@inproceedings{agreid2023,\n title = {AG-ReID 2023: Aerial-Ground Person Re-identification Challenge Results},\n author = {Nguyen, Kien and Fookes, Clinton and Sridharan, Sridha and Liu, Feng and Liu, Xiaoming and Ross, Arun and Michalski, Dana and Nguyen, Huy and others},\n year = {2023},\n booktitle = {2023 IEEE International Joint Conference on Biometrics (IJCB)},\n pages = {1--10},\n doi = {10.1109/IJCB57857.2023.10448780}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Advances in Benchmarks / Competitions" ], "abstract": "The AG-ReID 2023 challenge studies person retrieval between aerial and ground cameras, where viewpoint and image quality create substantial differences. Its benchmark combines identity annotations with fifteen soft attributes. The report describes the evaluation protocols and participating systems, using challenge results to assess progress on cross-platform recognition and remaining difficulties in combining heterogeneous observations.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://cvlab.cse.msu.edu/pdfs/IJCB_AG_ReID2023_Challenge_Summary_Paper.pdf", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "zhu2023gaitref-ijcb-2023", "citationKey": "zhu2023gaitref", "archiveKey": "zhu2023gaitref", "method": "GaitRef", "title": "Gaitref: Gait recognition with refined sequential skeletons", "authors": "Zhu, Haidong and Zheng, Wanrong and Zheng, Zhaoheng and Nevatia, Ram", "year": 2023, "venue": "IJCB", "motivation": "Improve gait recognition accuracy", "idea": "Refine skeletons using silhouettes", "techniques": "Dual-encoder fusion, skeleton refinement", "inputs": "Silhouette, Skeleton", "backbone": "CNN, GCN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, GREW, Gait3D", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/IJCB57857.2023.10448634", "bibtex": "@inproceedings{zhu2023gaitref,\n title = {Gaitref: Gait recognition with refined sequential skeletons},\n author = {Zhu, Haidong and Zheng, Wanrong and Zheng, Zhaoheng and Nevatia, Ram},\n year = {2023},\n booktitle = {2023 IEEE International Joint Conference on Biometrics (IJCB)},\n pages = {1--10}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 68 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Identifying humans with their walking sequences, known as gait recognition, is a useful biometric understanding task as it can be observed from a long distance and does not require cooperation from the subject. Two common modalities used for representing the walking sequence of a person are silhouettes and joint skeletons. Silhouette sequences, which record the boundary of the walking person in each frame, may suffer from the variant appearances from carried-on objects and clothes of the person. Framewise joint detections are noisy and introduce some jitters that are not consistent with sequential detections. In this paper, we combine the silhouettes and skeletons and refine the framewise joint predictions for gait recognition. With temporal information from the silhouette sequences. We show that the refined skeletons can improve gait recognition performance without extra annotations. We compare our methods on four public datasets, CASIA-B, OUMVLP , Gait3D and GREW, and show state-of-the-art performance.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/IJCB57857.2023.10448634", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-hid2023", "citationKey": "hid2023", "archiveKey": "hid2023", "method": "", "title": "Human Identification at a Distance: Challenges, Methods and Results on HID 2023", "authors": "Yu, Shiqi and Wang, Chenye and Zhao, Yuwei and Wang, Li and Wang, Ming and Li, Qing and Li, Wenlong and Wang, Runsheng and Huang, Yongzhen and Wang, Liang and Makihara, Yasushi and Ahad, Md Atiqur Rahman", "year": 2023, "venue": "IJCB", "motivation": "Distance-based human identification needs comparable community baselines.", "idea": "Evaluate submissions under the HID 2023 challenge protocols.", "techniques": "Gait benchmarks; shared evaluation; analysis of submitted methods.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/IJCB57857.2023.10448952", "bibtex": "@inproceedings{hid2023,\n title = {Human Identification at a Distance: Challenges, Methods and Results on HID 2023},\n author = {Yu, Shiqi and Wang, Chenye and Zhao, Yuwei and Wang, Li and Wang, Ming and Li, Qing and Li, Wenlong and Wang, Runsheng and Huang, Yongzhen and Wang, Liang and Makihara, Yasushi and Ahad, Md Atiqur Rahman},\n year = {2023},\n booktitle = {2023 IEEE International Joint Conference on Biometrics (IJCB)},\n pages = {1--8},\n doi = {10.1109/IJCB57857.2023.10448952}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Advances in Benchmarks / Competitions" ], "abstract": "HID 2023 introduces SUSTech-Competition, containing 859 subjects with varied clothing, carrying conditions, occlusions and views. The report describes the new cross-domain setting, competition organization, results and leading submissions. It provides a benchmark-level view of progress beyond earlier, more controlled competition protocols.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://hid.iapr-tc4.org/wp-content/uploads/sites/11/2024/03/HID2023-summary.pdf", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "zou2024multi-ijcb-2023", "citationKey": "zou2024multi", "archiveKey": "zou2024multi", "method": "MSAFF", "title": "A Multi-Stage Adaptive Feature Fusion Neural Network for Multimodal Gait Recognition", "authors": "Shinan Zou and Jianbo Xiong and Chao Fan and Shiqi Yu and Jin Tang", "year": 2023, "venue": "IJCB", "motivation": "Improve multimodal gait recognition", "idea": "Separate spatial-temporal multimodal fusion", "techniques": "Multi-stage fusion, multi-scale extraction, attention", "inputs": "Silhouette, Skeleton", "backbone": "CNN, GCN, Transformer", "loss": "Triplet", "datasets": "CASIA-B, GREW, Gait3D", "codeAvailable": "No", "codeUrl": "https://github.com/ShinanZou/MSAFF", "url": "https://arxiv.org/abs/2312.14410", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 67 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Author repository / IJCB 2023 accepted paper list", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition is a biometric technology that has received extensive attention. Most existing gait recognition algorithms are unimodal, and a few multimodal gait recognition algorithms perform multimodal fusion only once. None of these algorithms may fully exploit the complementary advantages of the multiple modalities. In this paper, by considering the temporal and spatial characteristics of gait data, we propose a multi-stage feature fusion strategy (MSFFS), which performs multimodal fusions at different stages in the feature extraction process. Also, we propose an adaptive feature fusion module ( AFFM) that considers the semantic association between silhouettes and skeletons. The fusion process fuses different silhouette areas with their more related skeleton joints. Since visual appearance changes and time passage co-occur in a gait period, we propose a multiscale spatial-temporal feature extractor (MSSTFE) to learn the spatial-temporal linkage features thoroughly. Specifically, MSSTFE extracts and aggregates spatial-temporal linkages information at different spatial scales. Combining the strategy and modules mentioned above, we propose a multi-stage adaptive feature fusion ( MSAFF) neural network, which shows state-of-theart performance in many experiments on three datasets. Besides, MSAFF is equipped with feature dimensional pooling (FD Pooling), which can significantly reduce the dimension of the gait representations without hindering the accuracy.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://arxiv.org/abs/2312.14410", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "wang2023pointgait-ijcb-2023", "citationKey": "wang2023pointgait", "archiveKey": "wang2023pointgait", "method": "PointGait", "title": "Pointgait: Boosting end-to-end 3d gait recognition with point clouds via spatiotemporal modeling", "authors": "Wang, Rui and Shen, Chuanfu and Fan, Chao and Huang, George Q and Yu, Shiqi", "year": 2023, "venue": "IJCB", "motivation": "Enhance point-based gait recognition", "idea": "Direct point cloud gait modeling", "techniques": "Multi-branch local-global extraction, motion modeling", "inputs": "Point", "backbone": "MLP", "loss": "Triplet", "datasets": "SUSTech1K", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/IJCB57857.2023.10448817", "bibtex": "@inproceedings{wang2023pointgait,\n title = {Pointgait: Boosting end-to-end 3d gait recognition with point clouds via spatiotemporal modeling},\n author = {Wang, Rui and Shen, Chuanfu and Fan, Chao and Huang, George Q and Yu, Shiqi},\n year = {2023},\n booktitle = {2023 IEEE International Joint Conference on Biometrics (IJCB)},\n pages = {1--10}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 69 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Ideal Gait Modalities" ], "abstract": "LiDAR is a new type of sensor used for gait recognition. Previous LiDAR-based state-of-the-art methods mostly exploit gait features from the depth maps generated by projecting point clouds in a 3D-to-2D manner, rather than directly using the raw 3D point data. However, these projection-based methods require an additional preprocessing step, which obstructs the universality of the method among different types of LiDARs. On the other hand, while existing point-based methods have achieved promising results in 3D object recognition, they have underperformed in 3D gait recognition, indicating the presence of a domain gap between coarse-grained 3D object classification and fine-grained 3D pedestrians recognition. By analyzing the success achieved by camera-based methods, we perceive that point-based gait recognition fails mainly because of neglecting to capture local representation. To address this issue, we propose an end-to-end 3D gait recognition framework named PointGait, which can directly capture informative gait features from point cloud data. Specifically, PointGait is a multi-stream model consisting of a Global and Local Gait Feature Extractor to extract holistic and fine-grained spatial features. Besides, a Personalized Motion Extractor is introduced to capture inter-frame motion features. Our experimental results on a LiDAR gait dataset, SUSTech1K, outperform all popular point-based methods, demonstrating the effectiveness and potential of our approach. In conclusion, the proposed PointGait promotes the development of point-based gait recognition by highlighting the importance of incorporating fine-grained spatiotemporal information.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/IJCB57857.2023.10448817", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "delgado2023exploring-pr-2023", "citationKey": "delgado2023exploring", "archiveKey": "delgado2023exploring", "method": "Proposed Transformer", "title": "Exploring Transformers for Behavioural Biometrics: A Case Study in Gait Recognition", "authors": "Paula Delgado-Santos and Ruben Tolosana and Richard Guest and Farzin Deravi and Ruben Vera-Rodriguez", "year": 2023, "venue": "PR", "motivation": "Efficient gait recognition on devices", "idea": "Apply Transformers to sequential gait", "techniques": "Self-attention, temporal modeling, parallelization", "inputs": "Skeleton", "backbone": "CNN, Transformer", "loss": "CE", "datasets": "whuGAIT, OU-ISIR IMU", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2023.109798", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 104 }, { "source": "Author paper summaries", "sheet": "journals", "row": 11 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Exploring-transformers-for-behavioural-biometrics--A-case_2023_Pattern-Recog.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Biometrics on mobile devices has attracted a lot of attention in recent years as it is considered a userfriendly authentication method. This interest has also been motivated by the success of Deep Learning (DL). Architectures based on Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) have established convenience for the task, improving the performance and robustness in comparison to traditional machine learning techniques. However, some aspects must still be revisited and improved. To the best of our knowledge, this is the first article that explores and proposes a novel gait biometric recognition systems based on Transformers, which currently obtain state-of-the-art performance in many applications. Several state-of-the-art architectures (Vanilla, Informer, Autoformer, Block-Recurrent Transformer, and THAT) are considered in the experimental framework. In addition, new Transformer configurations are proposed to further increase the performance. Experiments are carried out using the two popular public databases: whuGAIT and OU-ISIR. The results achieved prove the high ability of the proposed Transformer, outperforming state-of-the-art CNN and RNN architectures.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2023.109798", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "he2023temporal-pr-2023", "citationKey": "he2023temporal", "archiveKey": "he2023temporal", "method": "Temporal Sparse Adversarial Attack", "title": "Temporal sparse adversarial attack on sequence-based gait recognition", "authors": "He, Ziwen and Wang, Wei and Dong, Jing and Tan, Tieniu", "year": 2023, "venue": "PR", "motivation": "Assess gait model adversarial robustness", "idea": "Temporal sparsity and silhouette deformation", "techniques": "Temporal-sparse attacks; deformation-based perturbations", "inputs": "Silhouette", "backbone": "CNN, Transformer", "loss": "CE, Cosine Loss", "datasets": "CASIA-B, CASIA-A", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2022.109028", "bibtex": "@article{he2023temporal,\n title = {Temporal sparse adversarial attack on sequence-based gait recognition},\n author = {He, Ziwen and Wang, Wei and Dong, Jing and Tan, Tieniu},\n year = {2023},\n journal = {Pattern Recognition},\n volume = {133},\n pages = {109028},\n publisher = {Elsevier}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 105 }, { "source": "Author paper summaries", "sheet": "journals", "row": 12 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Privacy Concerns and Potential Solutions" ], "abstract": "Gait recognition is widely used in social security applications due to its advantages in long-distance human identification. Recently, sequence-based methods have achieved high accuracy by learning abundant temporal and spatial information. However, their robustness under adversarial attacks in an open world has not been clearly explored. In this paper, we demonstrate that the state-of-the-art gait recognition model is vulnerable to such attacks. To this end, we propose a novel temporal sparse adversarial attack method. Different from previous additive noise models which add perturbations on original samples, we employ a generative adversarial network based architecture to semantically generate adversarial highquality gait silhouettes or video frames. Moreover, by sparsely substituting or inserting a few adversarial gait silhouettes, the proposed method ensures its imperceptibility and achieves a strong attack ability. The experimental results show that if only one-fortieth of the frames are attacked, the accuracy of the target model drops dramatically.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2022.109028", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "parashar2023data-prl-2023", "citationKey": "parashar2023data", "archiveKey": "parashar2023data", "method": "", "title": "Data Preprocessing and Feature Selection Techniques in Gait Recognition: A Comparative Study of Machine Learning and Deep Learning Approaches", "authors": "Anubha Parashar and Apoorva Parashar and Weiping Ding and Mohammad Shabaz and Imad Rida", "year": 2023, "venue": "PRL", "motivation": "Non-intrusive reliable biometric identification", "idea": "Comprehensive review of gait recognition", "techniques": "Data acquisition methods; learning-based models", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1016/j.patrec.2023.05.021", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 115 }, { "source": "Author paper summaries", "sheet": "journals", "row": 22 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Data-preprocessing-and-feature-selection-techniques-in-gait_2023_Pattern-Rec.pdf", "authorsSource": "Public scholarly metadata", "abstract": "The study of gait recognition, a biometric application that identifies individuals based on their unique walking patterns, is an evolving field. In this paper, we conduct a literature review to compare the performance of machine learning and deep learning approaches in covariate conditions, focusing on the specific aspects of deep learning pipelines in gait recognition. We highlight commonly used strategies and open problems in identification based on behavioral traits and propose future perspectives for researchers in this field. Through our investigation, we aim to provide insights that will aid researchers in developing informed decisions when to take which data preprocessing technique in designing gait recognition systems. Our paper provides a comprehensive exposition of machine learning versus deep learning architectures and pipelines for biometric applications using human gait and serves as a valuable resource for researchers in this area.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patrec.2023.05.021", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "wang2023gait-tbiom-2023", "citationKey": "wang2023gait", "archiveKey": "wang2023gait", "method": "GPS / IGPS", "title": "Gait Recognition via Gait Period Set", "authors": "Runsheng Wang and Yuxuan Shi and Hefei Ling and Zongyi Li and Ping Li and Boyuan Liu and Hanqing Zheng and Qian Wang", "year": 2023, "venue": "TBIOM", "motivation": "Improve video-based gait recognition", "idea": "Leverage gait period for features", "techniques": "Phase-aware feature refinement aggregation", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TBIOM.2023.3244206", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 125 }, { "source": "Author paper summaries", "sheet": "journals", "row": 32 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Gait_Recognition_via_Gait_Period_Set.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition has promising application prospects in surveillance applications, with the recently proposed videobased gait recognition methods affording huge progress. However, due to the poor image quality of some gait frames, the original frame-level features extracted from gait silhouettes are not discriminative enough to be aggregated as gait features utilized during the final recognition. Besides, as a type of periodic biometric behavior, periodic gait information is considered efficacious for capturing typical human walking patterns and refining frame-level gait features. Therefore, this paper proposes a novel approach that exploits periodic gait information, named Gait Period Set (GPS), which divides the gait period into several phases and ensembles the gait phase features to refine frame-level features. Then, features from different phases are aggregated into a video-level feature. Moreover, the refined frame-level features are aggregated as the refined gait phase features with higher quality, which can be used to re-refine the frame-level features. Hence, we upgrade the GPS into the Iterative Gait Period Set (IGPS) to iteratively refine the frame-level features. The results of extensive experiments on prevailing gait recognition datasets validate the effectiveness of the GPS and IGPS modules and demonstrate that the proposed method achieves state-of-the-art performance.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TBIOM.2023.3244206", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "hou2022comprehensive-tbiom-2023", "citationKey": "hou2022comprehensive", "archiveKey": "hou2022comprehensive", "method": "Multi-Height Gait (MHG)", "title": "A Comprehensive Study on the Evaluation of Silhouette-Based Gait Recognition", "authors": "Saihui Hou and Chao Fan and Chunshui Cao and Xu Liu and Yongzhen Huang", "year": 2023, "venue": "TBIOM", "motivation": "Improving real-world gait recognition", "idea": "Reevaluate evaluation and datasets", "techniques": "Robust evaluation and multi-condition dataset", "inputs": "Silhouette", "backbone": "CNN", "loss": "", "datasets": "CASIA-B, OU-MVLP, MHG", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1109/TBIOM.2022.3216857", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 124 }, { "source": "Author paper summaries", "sheet": "journals", "row": 31 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "A_Comprehensive_Study_on_the_Evaluation_of_Silhouette-Based_Gait_Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Recently the methods based on silhouettes achieve significant improvement for gait recognition. The performance, e.g., 96.4% on the largest OUMVLP, indicates that a promising gait system is around the corner. However, we argue that the observation is not true. Firstly, we find that there exists a non-negligible gap of gait evaluation between academic research and practical applications. To validate the assumption, we conduct a comprehensive study on the evaluation for silhouette-based gait recognition and provide new insights into the limitations of the current methods. Our key findings include: (a) The current evaluation protocol is excessively simplified and ignores a lot of hard cases. (b) The current methods are sensitive to the noise caused by rotation and occlusion. Secondly, we observe that the data scarcity largely hinders the development of gait recognition and some crucial covariates ( e.g., camera heights) are not thoroughly investigated. To address the issue, we propose a new dataset called Multi-Height Gait (MHG). It collects 200 subjects of normal walking, walking with bags and walking in different clothes. Particularly, it collects the sequences recorded by the cameras at different heights. We hope this work would inspire more advanced research for gait recognition. The project page is available at https://hshustc.github.io/TBIOM-Metric-Gait/.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TBIOM.2022.3216857", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "huang2022star-tbiom-2023", "citationKey": "huang2022star", "archiveKey": "huang2022star", "method": "STAR", "title": "STAR Spatio-Temporal Augmented Relation Network for Gait Recognition", "authors": "Xiaohu Huang and Xinggang Wang and Botao He and Shan He and Wenyu Liu and Bin Feng", "year": 2023, "venue": "TBIOM", "motivation": "Enhance fine-grained gait recognition", "idea": "Capture multi-scale spatio-temporal features", "techniques": "Region-based feature interaction modeling", "inputs": "Silhouette", "backbone": "CNN", "loss": "Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TBIOM.2022.3211843", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 126 }, { "source": "Author paper summaries", "sheet": "journals", "row": 33 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "STAR_Spatio-Temporal_Augmented_Relation_Network_for_Gait_Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition is a promising long-distance biological recognition technology that is widely used in public security and video surveillance. In the most recent literature, researchers use fixed horizontal partitions on input images or intermediate features for feature extraction, which are neither adaptive to the variations in camera viewpoint and individual appearances nor sufficient for spatio-temporal fine-grained feature capturing. Besides, the existing methods either ignore temporal modeling or model insufficient/less-discriminative temporal information for motion capturing. Based on the findings above, it is considered that gait recognition requires both finegrained feature expression and higher spatio-temporal modeling capabilities. Therefore, this study proposes a spatio-temporal augmented relation network (STAR), which adaptively generates multiple salient features in non-overlapped diverse regions for fine-grained feature mining and extracts spatio-temporal augmented features with rich temporal scales based on intra- and inter-relation joint learning among diverse regions. Extensive experiments are conducted on the CASIA-B and OU-MVLP datasets to demonstrate the effectiveness of the proposed method. On the CASIA-B dataset, rank-1 accuracy of 97.4%, 95.6%, and 87.1% is achieved under normal-walking, bag-carrying, and coatwearing conditions, respectively, and 89.7% on the OU-MVLP dataset, which are the new state-of-the-art results.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TBIOM.2022.3211843", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "zhao2022gait-tcsvt-2023", "citationKey": "zhao2022gait", "archiveKey": "zhao2022gait", "method": "AGNet", "title": "Gait-Assisted Video Person Retrieval", "authors": "Yang Zhao and Xinlong Wang and Xiaohan Yu and Chunlei Liu and Yongsheng Gao", "year": 2023, "venue": "TCSVT", "motivation": "Enhance video person retrieval", "idea": "Integrate appearance and gait features", "techniques": "Two-stream network, attention aggregation, knowledge distillation", "inputs": "Silhouette, RGB", "backbone": "CNN, Attention, MLP", "loss": "CE, Triplet", "datasets": "MARS, DukeMTMC-VideoReID", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1109/TCSVT.2022.3202531", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 142 }, { "source": "Author paper summaries", "sheet": "journals", "row": 48 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Gait-Assisted_Video_Person_Retrieval.pdf", "abstract": "Video person retrieval aims at matching video clips of the same person across non-overlapping camera views, where video sequences contain more comprehensive information, e.g., temporal cues. How to extract useful temporal cues is the key to the success of a video person retrieval system. Gait, as a unique biometric modality indicating the way people walk, contains informative temporal information. To date, it is not clear how to fully utilize gait to boost the performance of video person retrieval. In this paper, to validate whether gait could help retrieve person in videos, we build a two-stream architecture, named appearance-gait network (AGNet), to jointly learn the appearance features and gait features from RGB video clips and silhouette video clips. We further explore how to fully utilize gait features to enhance the video feature representation. Specifically, we propose an appearance-gait attention module (AGA) to fuse a discriminative feature representation for the person retrieval task. Furthermore, to eliminate the requirement of silhouette video clips during inference, we propose a simple yet effective appearance-gait distillation module (AGD) which transfers the gait knowledge to appearance stream. As such, we are able to perform the enhanced video person retrieval without silhouette video clips, which makes the inference more flexible and practical. To the best of our knowledge, our work is the first to successfully introduce such appearance-gait knowledge distillation design for video person retrieval. We verify the effectiveness of the proposed methods on two large-scale challenging benchmarks of MARS and DukeMTMC-VideoReID. Extensive experiments demonstrate superior or comparable performance compared to the state-of-the-art methods while being much simpler. Source code is publicly available at https://github.com/yangyangkiki/Gait- Assisted-Video-Reid.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TCSVT.2022.3202531", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "pan2023toward-tifs-2023", "citationKey": "pan2023toward", "archiveKey": "pan2023toward", "method": "LUGAN", "title": "Toward Complete-View and High-Level Pose-Based Gait Recognition", "authors": "Honghu Pan and Yongyong Chen and Tingyang Xu and Yunqi He and Zhenyu He", "year": 2023, "venue": "TIFS", "motivation": "Overcome cross-view gait variance", "idea": "Geometry-guided pose generation", "techniques": "Linear transformation learning; adversarial training; multi-scale hypergraph convolution", "inputs": "Skeleton", "backbone": "CNN, GCN", "loss": "Contrastive", "datasets": "CASIA-B, OU-MVLP-Pose, NLPR", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TIFS.2023.3254449", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 156 }, { "source": "Author paper summaries", "sheet": "journals", "row": 60 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Toward_Complete-View_and_High-Level_Pose-Based_Gait_Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Model-based gait recognition methods usually adopt the pedestrian walking postures to identify human beings. However, existing methods did not explicitly resolve the large intra-class variance of human pose due to changes in camera view. In this paper, we propose a lower-upper generative adversarial network (LUGAN) to generate multi-view pose sequences for each single-view sample to reduce the cross-view variance. Based on the prior of camera imaging, we prove that the spatial coordinates between cross-view poses satisfy a linear transformation of a full-rank matrix. Hence, LUGAN employs the adversarial training to learn full-rank transformation matrices from the source pose and target views to obtain the target pose sequences. The generator of LUGAN is composed of graph convolutional (GCN) layers, fully connected (FC) layers and two-branch convolutional (CNN) layers: GCN layers and FC layers encode the source pose sequence and target view, then CNN layers take as input the encoded features to learn a lower triangular matrix and an upper one, finally the transformation matrix is formulated by multiplying the lower and upper triangular matrices. For the purpose of adversarial training, we develop a conditional discriminator that distinguishes whether the pose sequence is true or generated. Furthermore, to facilitate the high-level correlation learning, we propose a plug-andplay module, named multi-scale hypergraph convolution (HGC), to replace the spatial graph convolutional layer in baseline, which can simultaneously model the joint-level, part-level and body-level correlations. Extensive experiments on three large gait recognition datasets (i.e., CASIA-B, OUMVLP-Pose and NLPR) demonstrate that our method outperforms the baseline model by a large margin.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIFS.2023.3254449", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "xu2023occlusion-tifs-2023", "citationKey": "xu2023occlusion", "archiveKey": "xu2023occlusion", "method": "Occlusion-Aware Mesh", "title": "Occlusion-aware human mesh model-based gait recognition", "authors": "Xu, Chi and Makihara, Yasushi and Li, Xiang and Yagi, Yasushi", "year": 2023, "venue": "TIFS", "motivation": "Overcome gait occlusion challenges", "idea": "Human mesh based recognition", "techniques": "End-to-end model fitting; temporal sequence consistency; occlusion attenuation module", "inputs": "RGB", "backbone": "CNN", "loss": "Triplet", "datasets": "OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TIFS.2023.3236181", "bibtex": "@article{xu2023occlusion,\n title = {Occlusion-aware human mesh model-based gait recognition},\n author = {Xu, Chi and Makihara, Yasushi and Li, Xiang and Yagi, Yasushi},\n year = {2023},\n journal = {IEEE transactions on information forensics and security},\n volume = {18},\n pages = {1309--1321},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 154 }, { "source": "Author paper summaries", "sheet": "journals", "row": 58 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Handling Real-World Noise" ], "abstract": "Partial occlusion of the human body caused by obstacles or a limited camera field of view often occurs in surveillance videos, which affects the performance of gait recognition in practice. Existing methods for gait recognition against occlusion require a bounding box or the height of a full human body as a prerequisite, which is unobserved in occlusion scenarios. In this paper, we propose an occlusion-aware model-based gait recognition method that works directly on gait videos under occlusion without the above-mentioned prerequisite. Specifically, given a gait sequence that only contains non-occluded body parts in the images, we directly fit a skinned multi-person linear (SMPL)-based human mesh model to the input images without any pre-normalization or registration of the human body. We further use the pose and shape features extracted from the estimated SMPL model for recognition purposes, and use the extracted camera parameters in the occlusion attenuation module to reduce intra-subject variation in human model fitting caused by occlusion pattern differences. Experiments on occlusion samples simulated from the OU-MVLP dataset demonstrated the effectiveness of the proposed method, which outperformed state-of-the-art gait recognition methods by about 15% rank-1 identification rate and 2% equal error rate in the identification and verification scenarios, respectively.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIFS.2023.3236181", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "wang2023progressive-tifs-2023", "citationKey": "wang2023progressive", "archiveKey": "wang2023progressive", "method": "PSIM", "title": "Progressive sub-domain information mining for single-source generalizable gait recognition", "authors": "Wang, Yang and Huang, Yan and Shan, Caifeng and Wang, Liang", "year": 2023, "venue": "TIFS", "motivation": "Overcome domain shift challenge", "idea": "Single-source generalizable recognition", "techniques": "Unsupervised clustering; sub-domain partitioning; domain-insensitive loss functions", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP, GREW", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TIFS.2023.3298518", "bibtex": "@article{wang2023progressive,\n title = {Progressive sub-domain information mining for single-source generalizable gait recognition},\n author = {Wang, Yang and Huang, Yan and Shan, Caifeng and Wang, Liang},\n year = {2023},\n journal = {IEEE Transactions on Information Forensics and Security},\n volume = {18},\n pages = {4787--4799},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 155 }, { "source": "Author paper summaries", "sheet": "journals", "row": 59 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Gait Learning Frameworks" ], "abstract": "Recent years have witnessed the deployment of fully supervised gait recognition. However, due to domain diversity, gait recognition models designed under the fully supervised condition suffer from poor generalization in unseen domains. How to improve the generalization ability of gait recognition models and enhance their performance on unseen domains is unexplored in existing gait recognition approaches. This paper investigates the generalizable gait recognition problem and proposes a Progressive Sub-domain Information Mining (PSIM) framework for single-source generalizable gait recognition. During training, the PSIM can mine sub-domain information from a single large-scale source domain by differentiating gait features extracted from different people through unsupervised clustering. Then, domain information mitigation loss and domain homogenization loss are introduced to regularize those gait features to be domaininsensitive. The above procedures are conducted iteratively until the model converges. The PSIM framework is model-agnostic, which can directly improve the generalization ability of state-ofthe-art gait recognition models without significantly increasing the complexity of the model. In experiments, our model-agnostic PSIM framework is adopted on several gait recognition models to show its effectiveness in boosting gait recognition performance for the single-source generalizable gait recognition task.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIFS.2023.3298518", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "huang2023condition-tip-2023", "citationKey": "huang2023condition", "archiveKey": "huang2023condition", "method": "CAG", "title": "Condition-adaptive graph convolution learning for skeleton-based gait recognition", "authors": "Huang, Xiaohu and Wang, Xinggang and Jin, Zhidianqiu and Yang, Bo and He, Botao and Feng, Bin and Liu, Wenyu", "year": 2023, "venue": "TIP", "motivation": "Robust skeleton-based gait recognition", "idea": "Adapt graphs to view and joints", "techniques": "Joint-specific filters, view-adaptive topology, spatial-temporal feature extraction", "inputs": "Skeleton", "backbone": "GCN", "loss": "Triplet, View Loss, Circle Loss", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TIP.2023.3305822", "bibtex": "@article{huang2023condition,\n title = {Condition-adaptive graph convolution learning for skeleton-based gait recognition},\n author = {Huang, Xiaohu and Wang, Xinggang and Jin, Zhidianqiu and Yang, Bo and He, Botao and Feng, Bin and Liu, Wenyu},\n year = {2023},\n journal = {IEEE Transactions on Image Processing},\n volume = {32},\n pages = {4773--4784},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 169 }, { "source": "Author paper summaries", "sheet": "journals", "row": 71 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Strong Gait Backbones" ], "abstract": "Graph convolutional networks have been widely applied in skeleton-based gait recognition. A key challenge in this task is to distinguish the individual walking styles of different subjects across various views. Existing state-of-the-art methods employ uniform convolutions to extract features from diverse sequences and ignore the effects of viewpoint changes. To overcome these limitations, we propose a condition-adaptive graph (CAG) convolution network that can dynamically adapt to the specific attributes of each skeleton sequence and the corresponding view angle. In contrast to using fixed weights for all joints and sequences, we introduce a joint-specific filter learning (JSFL) module in the CAG method, which produces sequence-adaptive filters at the joint level. The adaptive filters capture fine-grained patterns that are unique to each joint, enabling the extraction of diverse spatial-temporal information about body parts. Additionally, we design a view-adaptive topology learning (V ATL) module that generates adaptive graph topologies. These graph topologies are used to correlate the joints adaptively according to the specific view conditions. Thus, CAG can simultaneously adjust to various walking styles and viewpoints. Experiments on the two most widely used datasets (i.e., CASIA-B and OU-MVLP) show that CAG surpasses all previous skeleton-based methods. Moreover, the recognition performance can be enhanced by simply combining CAG with appearancebased methods, demonstrating the ability of CAG to provide useful complementary information.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIP.2023.3305822", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "li2022strong-tmm-2023", "citationKey": "li2022strong", "archiveKey": "li2022strong", "method": "CycleGait", "title": "A Strong and Robust Skeleton-Based Gait Recognition Method with Gait Periodicity Priors", "authors": "Na Li and Xinbo Zhao", "year": 2023, "venue": "TMM", "motivation": "Robust gait recognition under variations", "idea": "Leverage gait periodicity for features", "techniques": "Temporal feature pyramid, graph convolution, irregular pace augmentation, cycle-based extraction", "inputs": "Skeleton", "backbone": "CNN, GCN", "loss": "CE, Center Loss, SupCon Loss", "datasets": "CASIA-B, OG RGB+D", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TMM.2022.3154609", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 175 }, { "source": "Author paper summaries", "sheet": "journals", "row": 77 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "A_Strong_and_Robust_Skeleton-Based_Gait_Recognition_Method_with_Gait_Periodicity_Priors.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition aims to identify people by their walking patterns. Normal human walking is a periodic movement, however, existing gait recognition methods rarely make use of gait periodicity. In this paper, we propose the gait Periodicity-inspired Temporal feature Pyramid aggregator (PTP), which introduces gait periodicity priors into gait feature extraction, resulting in a strong and robust skeleton-based gait recognition method called CycleGait. Specifically, inspired by gait periodicity, PTP adopts multiple parallel temporal convolution operators with pyramid temporal kernel sizes to extract temporal gait features. Then, PTP cooperates with the spatial Graph Convolutional Network (GCN) to form the GCN-PTP network. CycleGait uses this network to extract spatio-temporal gait features from a sequence of skeleton coordinates. In addition, to improve CycleGait’s robustness and performance, we feed more gait samples with various gait cycles into CycleGait with the plug-and-play Irregular Pace Converter (IPC), which can automatically convert normal pace into irregular and reasonable paces. Extensive experiments conducted on the CASIA-B dataset and OG RGB+D dataset show that CycleGait has excellent performance in various complex scenarios, namely, crossview and cross-walking conditions, and becomes one of the best SOTA methods, which not only outperforms the best preexisting gait recognition methods by a large margin but also exhibits a significant level of robustness.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TMM.2022.3154609", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "wang2023gaitparsing-tmm-2024", "citationKey": "wang2023gaitparsing", "archiveKey": "wang2023gaitparsing", "method": "GaitParsing", "title": "GaitParsing: Human semantic parsing for gait recognition", "authors": "Wang, Zengbin and Hou, Saihui and Zhang, Man and Liu, Xu and Cao, Chunshui and Huang, Yongzhen", "year": 2023, "venue": "TMM", "motivation": "Enhance gait recognition with details", "idea": "Use human parsing for local features", "techniques": "Semantic parsing, dual-branch extraction, self-occlusion assessment, frame selection", "inputs": "Silhouette, Parsing", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-BP, FVG-P", "codeAvailable": "No", "codeUrl": "", "url": "", "bibtex": "@article{wang2023gaitparsing,\n title = {GaitParsing: Human semantic parsing for gait recognition},\n author = {Wang, Zengbin and Hou, Saihui and Zhang, Man and Liu, Xu and Cao, Chunshui and Huang, Yongzhen},\n year = {2023},\n journal = {IEEE Transactions on Multimedia},\n volume = {26},\n pages = {4736--4748},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 178 }, { "source": "Author paper summaries", "sheet": "journals", "row": 80 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Ideal Gait Modalities" ], "yearNote": "Archive: 2024; bibliography: 2023", "abstract": "Gait recognition is a soft biotechnology to identify pedestrians observed from different camera views based on specific walking patterns. However, various dressing and wearing conditions bring great challenges to realistic gait recognition. Most existing methods take holistic gait silhouette as input and focus on local areas through horizontal strip division or attention map. We consider that this processing may contain mixed or incomplete information about multiple body parts so that gait information is misused or underutilized. In this paper, we propose a parsingguided framework for gait recognition, named GaitParsing, which explores human semantic parsing to dissect human body into a set of specific and complete body parts. Correspondingly, a simple yet effective dual-branch feature extraction network is adopted to process holistic gait and distinct body parts. To maximize the use of highly discriminated gait frames, we propose a self-occlusion frame assessment to measure the self-occlusion in a gait sequence. Since there is no human parsing modality in current gait datasets, we further develop a general human parsing pipeline specifically tailored for gait datasets. This single training enables widespread application across various gait datasets. Extensive experiments with ablation analyses demonstrate competitive performance even in the most challenging conditions, e.g., Cloth-Changing (CC+5.9%). Especially, It is gratifying to see that our model can be easily applied to existing methods and significantly outperform the original architecture, even without much modification.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "yao2022improving-tcsvt-2023", "citationKey": "yao2022improving", "archiveKey": "yao2022improving", "method": "Group-Supervised Disentangled Representation Learning (GSDRL)", "title": "Improving Disentangled Representation Learning for Gait Recognition Using Group Supervision", "authors": "Lingxiang Yao and Worapan Kusakunniran and Peng Zhang and Qiang Wu and Jian Zhang", "year": 2023, "venue": "TMM", "motivation": "Robust gait recognition under variations", "idea": "Disentangle pose, gait, appearance features", "techniques": "Encoder-decoder, feature swap, feature aggregation", "inputs": "Silhouette, RGB", "backbone": "CNN, Attention, MLP", "loss": "Triplet, MSE", "datasets": "CASIA-B, OU-ISIR Cloth", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/tmm.2022.3171961", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 143 }, { "source": "Author paper summaries", "sheet": "journals", "row": 49 }, { "source": "Verified publication metadata" }, { "source": "Author paper summaries", "sheet": "main-table", "row": 176 }, { "source": "Author paper summaries", "sheet": "journals", "row": 78 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Gait Learning Frameworks", "Advances in Methods / Handling Real-World Noise" ], "authorsSource": "Public scholarly metadata", "abstract": "For decades, gait has been gathering extensive interest due to the advantage that it can be measured from a distance without physical contact. However, for image/video-based gait recognition, its performance can be remarkably influenced by exterior factors, such as viewing angles and clothing changes. Thus, in this paper, a group-supervised disentangled representation learning network is proposed for gait recognition to extract features invariant to these factors. First, sequences are explicitly disentangled into pose, gait, appearance, and view features through a generic encoder-decoder framework. To ensure feature adaptability and independency, a disentanglement swap module is specifically adopted during our encoder-decoder process through a series of swap operations based on the feature attributes. Following the feature disentanglement, a disentanglement aggregation module is also specially proposed for pose, gait, and appearance features to enhance their effectiveness. Finally, the enhanced three features are concatenated together for gait recognition. Relevant experiments certify that compared with other disentangled representation learning-based gait recognition methods, our proposed method enables a more excellent recognition result, despite fewer gait frames being utilized.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/tmm.2022.3171961", "complete": true, "aliases": [ "yao2022improving-tmm-2023" ], "citationAliases": [ "yao2022improving" ], "statisticsEligible": true, "venueInScope": true }, { "id": "li2023gait-tmm-2024", "citationKey": "li2023gait", "archiveKey": "li2023gait", "method": "V-Distill", "title": "Gait recognition with drones: A benchmark", "authors": "Li, Aoqi and Hou, Saihui and Cai, Qingyuan and Fu, Yang and Huang, Yongzhen", "year": 2023, "venue": "TMM", "motivation": "Robust gait recognition at high views", "idea": "Align high-view features to moderate views", "techniques": "Feature distillation, dual-branch learning, cross-view alignment, multi-modal data utilization", "inputs": "Silhouette, 2D Pose, 3D Pose, 3D Mesh, Optical", "backbone": "CNN, GCN", "loss": "CE, Triplet, Contrastive", "datasets": "DroneGait", "codeAvailable": "No", "codeUrl": "", "url": "", "bibtex": "@article{li2023gait,\n title = {Gait recognition with drones: A benchmark},\n author = {Li, Aoqi and Hou, Saihui and Cai, Qingyuan and Fu, Yang and Huang, Yongzhen},\n year = {2023},\n journal = {IEEE Transactions on Multimedia},\n volume = {26},\n pages = {3530--3540},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 177 }, { "source": "Author paper summaries", "sheet": "journals", "row": 79 } ], "summaryStatus": "Author collection", "sections": [ "Background / Broader Applications" ], "yearNote": "Archive: 2024; bibliography: 2023", "abstract": "Gait recognition aims to obtain people’s identity through body shape and walking posture. Existing gait recognition studies focus on low vertical view recognition, in which the person and the camera are nearly at the same height. Differently, in this work, we focus on gait recognition at high vertical views. To facilitate the research, we propose a new dataset named DroneGait, where the drones are used to collect the gait data. This dataset contains 22 k sequences of 96 subjects taken at different vertical views, varying from about 0 ◦ to 80 ◦. Furthermore, we evaluate the effectiveness of several state-of-the-art appearancebased and skeleton-based models using our dataset and establish comprehensive baselines. Our results demonstrate that the dataset is challenging and presents significant opportunities to improve existing gait recognition methods. Moreover, we propose a new method called V ertical Distillation, which is based on the feature distillation across different vertical views. Our proposed method substantially outperforms the state-of-the-art models on DroneGait at high vertical views. Cross-vertical-view and cross-domain experiments are also made to explain the importance of gait recognition at high vertical views. Furthermore, we analyze the differences between gait recognition at different vertical views using heatmap visualization techniques. We will make our dataset and code publicly available upon acceptance.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "hou2022gait-tnnls-2023", "citationKey": "hou2022gait", "archiveKey": "hou2022gait", "method": "GQAN", "title": "Gait quality aware network: toward the interpretability of silhouette-based gait recognition", "authors": "Hou, Saihui and Liu, Xu and Cao, Chunshui and Huang, Yongzhen", "year": 2023, "venue": "TNNLS", "motivation": "Improve interpretability in gait recognition", "idea": "Assess silhouette and part quality", "techniques": "Frame quality scoring, part weighting, set-based feature learning, sequence-level supervision", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TNNLS.2022.3154723", "bibtex": "@article{hou2022gait,\n title = {Gait quality aware network: toward the interpretability of silhouette-based gait recognition},\n author = {Hou, Saihui and Liu, Xu and Cao, Chunshui and Huang, Yongzhen},\n year = {2023},\n journal = {IEEE Transactions on Neural Networks and Learning Systems},\n volume = {34},\n number = {11},\n pages = {8978--8988},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 183 }, { "source": "Author paper summaries", "sheet": "journals", "row": 83 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Handling Real-World Noise" ], "abstract": "Gait recognition receives increasing attention since it can be conducted at a long distance in a nonintrusive way and applied to the condition of changing clothes. Most existing methods take the silhouettes of gait sequences as the input and learn a unified representation from multiple silhouettes to match probe and gallery. However, these models are all faced with the lack of interpretability, e.g., it is not clear which silhouette in a gait sequence and which part in the human body are relatively more important for recognition. In this work, we propose a gait quality aware network (GQAN) for gait recognition which explicitly assesses the quality of each silhouette and each part via two blocks: frame quality block (FQBlock) and part quality block (PQBlock). Specifically, FQBlock works in a squeeze-andexcitation style to recalibrate the features for each silhouette, and the scores of all the channels are added as frame quality indicator. PQBlock predicts a score for each part which is used to compute the weighted distance between the probe and gallery. Particularly, we propose a part quality loss (PQLoss) which enables GQAN to be trained in an end-to-end manner with only sequence-level identity annotations. This work is meaningful by moving toward the interpretability of silhouette-based gait recognition, and our method also achieves very competitive performance on CASIA-B and OUMVLP.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TNNLS.2022.3154723", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "sepas2022deep-tpami-2023", "citationKey": "sepas2022deep", "archiveKey": "sepas2022deep", "method": "", "title": "Deep gait recognition: A survey", "authors": "Sepas-Moghaddam, Alireza and Etemad, Ali", "year": 2023, "venue": "TPAMI", "motivation": "Deep learning gait survey", "idea": "Organize, analyze, compare methods", "techniques": "Taxonomy creation, dataset review, benchmarking", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TPAMI.2022.3151865", "bibtex": "@article{sepas2022deep,\n title = {Deep gait recognition: A survey},\n author = {Sepas-Moghaddam, Alireza and Etemad, Ali},\n year = {2023},\n journal = {IEEE transactions on pattern analysis and machine intelligence},\n volume = {45},\n number = {1},\n pages = {264--284},\n publisher = {IEEE},\n doi = {10.1109/TPAMI.2022.3151865}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 191 }, { "source": "Author paper summaries", "sheet": "journals", "row": 91 } ], "summaryStatus": "Author collection", "sections": [ "Introduction / Context" ], "abstract": "Gait recognition is an appealing biometric modality which aims to identify individuals based on the way they walk. Deep learning has reshaped the research landscape in this area since 2015 through the ability to automatically learn discriminative representations. Gait recognition methods based on deep learning now dominate the state-of-the-art in the field and have fostered realworld applications. In this paper , we present a comprehensive overview of breakthroughs and recent developments in gait recognition with deep learning, and cover broad topics including datasets, test protocols, state-of-the-art solutions, challenges, and future research directions. We first review the commonly used gait datasets along with the principles designed for evaluating them. We then propose a novel taxonomy made up of four separate dimensions namely body representation, temporal representation, feature representation, and neural architecture, to help characterize and organize the research landscape and literature in this area. Following our proposed taxonomy , a comprehensive survey of gait recognition methods using deep learning is presented with discussions on their performances, characteristics, advantages, and limitations. We conclude this survey with a discussion on current challenges and mention a number of promising directions for future research in gait recognition.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TPAMI.2022.3151865", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "fan2023learning-tpami-2023", "citationKey": "fan2023learning", "archiveKey": "fan2023learning", "method": "GaitSSB", "title": "Learning gait representation from massive unlabelled walking videos: A benchmark", "authors": "Fan, Chao and Hou, Saihui and Wang, Jilong and Huang, Yongzhen and Yu, Shiqi", "year": 2023, "venue": "TPAMI", "motivation": "Learn general gait representation", "idea": "Self-supervised learning from videos", "techniques": "Large-scale dataset, contrastive learning, silhouette augmentation", "inputs": "Silhouette", "backbone": "CNN", "loss": "Contrastive, InfoNCE Loss", "datasets": "CASIA-B, OU-MVLP, GREW, Gait3D", "codeAvailable": "Yes", "codeUrl": "", "url": "", "bibtex": "@article{fan2023learning,\n title = {Learning gait representation from massive unlabelled walking videos: A benchmark},\n author = {Fan, Chao and Hou, Saihui and Wang, Jilong and Huang, Yongzhen and Yu, Shiqi},\n year = {2023},\n journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},\n volume = {45},\n number = {12},\n pages = {14920--14937},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 192 }, { "source": "Author paper summaries", "sheet": "journals", "row": 92 } ], "summaryStatus": "Author collection", "sections": [ "Background / Historical Evolution", "Advances in Methods / Gait Learning Frameworks" ], "abstract": "Gait depicts individuals’ unique and distinguishing walking patterns and has become one of the most promising biometric features for human identification. As a fine-grained recognition task, gait recognition is easily affected by many factors and usually requires a large amount of completely annotated data that is costly and insatiable. This paper proposes a large-scale self-supervised benchmark for gait recognition with contrastive learning, aiming to learn the general gait representation from massive unlabelled walking videos for practical applications via offering informative walking priors and diverse real-world variations. Specifically, we collect a large-scale unlabelled gait dataset GaitLU-1M consisting of 1.02M walking sequences and propose a conceptually simple yet empirically powerful baseline model GaitSSB. Experimentally, we evaluate the pre-trained model on four widely-used gait benchmarks, CASIA-B, OU-MVLP, GREW and Gait3D with or without transfer learning. The unsupervised results are comparable to or even better than the early model-based and GEI-based methods. After transfer learning, GaitSSB outperforms existing methods by a large margin in most cases, and also showcases the superior generalization capacity. Further experiments indicate that the pre-training can save about 50% and 80% annotation costs of GREW and Gait3D. Theoretically, we discuss the critical issues for gait-specific contrastive framework and present some insights for further study. As far as we know, GaitLU-1M is the first largescale unlabelled gait dataset, and GaitSSB is the first method that achieves remarkable unsupervised results on the aforementioned benchmarks.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-cornett2023expanding", "citationKey": "cornett2023expanding", "archiveKey": "cornett2023expanding", "method": "", "title": "Expanding accurate person recognition to new altitudes and ranges: The briar dataset", "authors": "Cornett, David and Brogan, Joel and Barber, Nell and Aykac, Deniz and Baird, Seth and Burchfield, Nicholas and Dukes, Carl and Duncan, Andrew and Ferrell, Regina and Goddard, Jim and others", "year": 2023, "venue": "WACV", "motivation": "Existing biometric data underrepresent long distances and elevated views.", "idea": "Collect face, body and gait observations under demanding acquisition conditions.", "techniques": "BRIAR dataset; whole-body biometrics; altitude and range evaluation.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@inproceedings{cornett2023expanding,\n title = {Expanding accurate person recognition to new altitudes and ranges: The briar dataset},\n author = {Cornett, David and Brogan, Joel and Barber, Nell and Aykac, Deniz and Baird, Seth and Burchfield, Nicholas and Dukes, Carl and Duncan, Andrew and Ferrell, Regina and Goddard, Jim and others},\n year = {2023},\n booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},\n pages = {593--602}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution", "Background / Broader Applications", "Advances in Benchmarks / Popular Datasets" ], "abstract": "Face recognition technology has advanced significantly in recent years due largely to the availability of large and increasingly complex training datasets for use in deep learning models. These datasets, however, typically comprise images scraped from news sites or social media platforms and, therefore, have limited utility in more advanced security, forensics, and military applications. These applications require lower resolution, longer ranges, and elevated viewpoints. To meet these critical needs, we collected and curated the first and second subsets of a large multimodal biometric dataset designed for use in the research and development (R&D) of biometric recognition technologies under extremely challenging conditions. Thus far, the dataset includes more than 350,000 still images and over 1,300 hours of video footage of approximately 1,000 subjects. To collect this data, we used Nikon DSLR cameras, a variety of commercial surveillance cameras, specialized long-rage R&D cameras, and Group 1 and Group 2 UAV platforms. The goal is to support the development of algorithms capable of accurately recognizing people at ranges up to 1,000 m and from high angles of elevation. These advances will include improvements to the state of the art in face recognition and will support new research in the area of whole-body recognition using methods based on gait and anthropometry. This paper describes methods used to collect and curate the dataset, and the dataset’s characteristics at the current stage.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "zhu2023gait-wacv-2023", "citationKey": "zhu2023gait", "archiveKey": "zhu2023gait", "method": "HBS", "title": "Gait recognition using 3-d human body shape inference", "authors": "Zhu, Haidong and Zheng, Zhaoheng and Nevatia, Ram", "year": 2023, "venue": "WACV", "motivation": "Handle gait appearance variance", "idea": "3D shape inference from silhouettes", "techniques": "Knowledge distillation, temporal aggregation, feature fusion", "inputs": "Silhouette", "backbone": "CNN", "loss": "", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/WACV2023/html/Zhu_Gait_Recognition_Using_3-D_Human_Body_Shape_Inference_WACV_2023_paper.html", "bibtex": "@inproceedings{zhu2023gait,\n title = {Gait recognition using 3-d human body shape inference},\n author = {Zhu, Haidong and Zheng, Zhaoheng and Nevatia, Ram},\n year = {2023},\n booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},\n pages = {909--918}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 88 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait recognition, which identifies individuals based on their walking patterns, is an important biometric technique since it can be observed from a distance and does not require the subject’s cooperation. Recognizing a person’s gait is difficult because of the appearance variants in human silhouette sequences produced by varying viewing angles, carrying objects, and clothing. Recent research has produced a number of ways for coping with these variants. In this paper, we present the usage of inferring 3-D body shapes distilled from limited images, which are, in principle, invariant to the specified variants. Inference of 3-D shape is a difficult task, especially when only silhouettes are provided in a dataset. We provide a method for learning 3-D body inference from silhouettes by transferring knowledge from 3-D shape prior from RGB photos. We use our method on multiple existing state-of-the-art gait baselines and obtain consistent improvements for gait identification on two public datasets, CASIA-B and OUMVLP , on several variants and settings, including a new setting of novel views not seen during training.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/WACV2023/html/Zhu_Gait_Recognition_Using_3-D_Human_Body_Shape_Inference_WACV_2023_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-fan2023exploring", "citationKey": "fan2023exploring", "archiveKey": "fan2023exploring", "method": "DeepGaitV2 / SwinGait", "title": "Exploring deep models for practical gait recognition", "authors": "Fan, Chao and Hou, Saihui and Huang, Yongzhen and Yu, Shiqi", "year": 2023, "venue": "arXiv", "motivation": "Backbone capacity and modeling choices need testing under practical gait conditions.", "idea": "Systematically explore deep convolutional and Transformer gait models.", "techniques": "DeepGaitV2; SwinGait; architecture and training analysis.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://arxiv.org/abs/2303.03301", "bibtex": "@article{fan2023exploring,\n title = {Exploring deep models for practical gait recognition},\n author = {Fan, Chao and Hou, Saihui and Huang, Yongzhen and Yu, Shiqi},\n year = {2023},\n journal = {arXiv preprint arXiv:2303.03301}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Advances in Methods / Strong Gait Backbones" ], "abstract": "Gait recognition is a rapidly advancing vision technique for person identification from a distance. Prior studies predominantly employed relatively shallow networks to extract subtle gait features, achieving impressive successes in constrained settings. Nevertheless, experiments revealed that existing methods mostly produce unsatisfactory results when applied to newly released real-world gait datasets. This paper presents a unified perspective to explore how to construct deep models for state-of-the-art outdoor gait recognition, including the classical CNN-based and emerging Transformer-based architectures. Specifically, we challenge the stereotype of shallow gait models and demonstrate the superiority of explicit temporal modeling and deep transformer structure for discriminative gait representation learning. Consequently, the proposed CNN-based DeepGaitV2 series and Transformerbased SwinGait series exhibit significant performance improvements on Gait3D and GREW. As for the constrained gait datasets, the DeepGaitV2 series also reaches a new state-of-the-art in most cases, convincingly showing its practicality and generality. The source code is available at https://github.com/ShiqiYu/OpenGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://arxiv.org/abs/2303.03301", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "wang2022gaitstrip-accv-2022", "citationKey": "wang2022gaitstrip", "archiveKey": "wang2022gaitstrip", "method": "GaitStrip", "title": "Gaitstrip: Gait recognition via effective strip-based feature representations and multi-level framework", "authors": "Wang, Ming and Lin, Beibei and Guo, Xianda and Li, Lincheng and Zhu, Zheng and Sun, Jiande and Zhang, Shunli and Liu, Yu and Yu, Xin", "year": 2022, "venue": "ACCV", "motivation": "Improve generalization beyond part-based models", "idea": "Use strips with enhanced convolution module", "techniques": "Multi-level part-based feature extraction", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "Yes", "codeUrl": "", "url": "", "bibtex": "@inproceedings{wang2022gaitstrip,\n title = {Gaitstrip: Gait recognition via effective strip-based feature representations and multi-level framework},\n author = {Wang, Ming and Lin, Beibei and Guo, Xianda and Li, Lincheng and Zhu, Zheng and Sun, Jiande and Zhang, Shunli and Liu, Yu and Yu, Xin},\n year = {2022},\n booktitle = {Proceedings of the Asian conference on computer vision},\n pages = {536--551}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 13 } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Many gait recognition methods first partition the human gait into N-parts and then combine them to establish part-based feature representations. Their gait recognition performance is often affected by partitioning strategies, which are empirically chosen in different datasets. However, we observe that strips as the basic component of parts are agnostic against different partitioning strategies. Motivated by this observation, we present a strip-based multi-level gait recognition network, named GaitStrip, to extract comprehensive gait information at different levels. To be specific, our high-level branch explores the context of gait sequences and our low-level one focuses on detailed posture changes. We introduce a novel StriP-Based feature extractor (SPB) to learn the stripbased feature representations by directly taking each strip of the human body as the basic unit. Moreover, we propose a novel multi-branch structure, called Enhanced Convolution Module (ECM), to extract different representations of gaits. ECM consists of the Spatial-Temporal feature extractor (ST), the Frame-Level feature extractor (FL) and SPB, and has two obvious advantages: First, each branch focuses on a specific representation, which can be used to improve the robustness of the network. Specifically, ST aims to extract spatial-temporal features of gait sequences, while FL is used to generate the feature representation of each frame. Second, the parameters of the ECM can be reduced in test by introducing a structural re-parameterization technique. Extensive experimental results demonstrate that our GaitStrip achieves state-of-the-art performance in both normal walking and complex conditions. The source code is published at https://github.com/M-Candy77/GaitStrip.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "yu2022generalized-acmmm-2022", "citationKey": "yu2022generalized", "archiveKey": "yu2022generalized", "method": "", "title": "Generalized inter-class loss for gait recognition", "authors": "Yu, Weichen and Yu, Hongyuan and Huang, Yan and Wang, Liang", "year": 2022, "venue": "ACM MM", "motivation": "Increase inter-class gait discrimination", "idea": "Dynamic and uniform inter-class constraints", "techniques": "Dynamic pair weighting, uniform feature distribution, adaptive margin adjustment", "inputs": "Silhouette", "backbone": "", "loss": "CE, Triplet, SimCE loss", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1145/3503161.3548311", "bibtex": "@inproceedings{yu2022generalized,\n title = {Generalized inter-class loss for gait recognition},\n author = {Yu, Weichen and Yu, Hongyuan and Huang, Yan and Wang, Liang},\n year = {2022},\n booktitle = {Proceedings of the 30th ACM International Conference on Multimedia},\n pages = {141--150}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 75 }, { "source": "Verified publication metadata" }, { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Gait Metric Designs" ], "abstract": "Gait recognition is a unique biometric technique that can be performed at a long distance non-cooperatively and has broad applications in public safety and intelligent traffic systems. Previous gait works focus more on minimizing the intra-class variance while ignoring the significance in constraining inter-class variance. To this end, we propose a generalized inter-class loss which resolves the inter-class variance from both sample-level feature distribution and class-level feature distribution. Instead of equal penalty strength on pair scores, the proposed loss optimizes sample-level inter-class feature distribution by dynamically adjusting the pairwise weight. Further, in class-level distribution, generalized interclass loss adds a constraint on the uniformity of inter-class feature distribution, which forces the feature representations to approximate a hypersphere and keep maximal inter-class variance. In addition, the proposed method automatically adjusts the margin between classes which enables the inter-class feature distribution to be more flexible. The proposed method can be generalized to different gait recognition networks and achieves significant improvements. We conduct a series of experiments on CASIA-B and OUMVLP, and the experimental results show that the proposed loss can significantly improve the performance and achieves the state-of-the-art performances.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1145/3503161.3548311", "complete": true, "aliases": [ "ref-yu2022interclass" ], "citationAliases": [ "yu2022interclass" ], "statisticsEligible": true, "venueInScope": true }, { "id": "zheng2022mtsgait-acmmm-2022", "citationKey": "zheng2022mstgait", "archiveKey": "zheng2022mtsgait", "method": "MTSGait", "title": "Gait recognition in the wild with multi-hop temporal switch", "authors": "Zheng, Jinkai and Liu, Xinchen and Gu, Xiaoyan and Sun, Yaoqi and Gan, Chuang and Zhang, Jiyong and Liu, Wu and Yan, Chenggang", "year": 2022, "venue": "ACM MM", "motivation": "Improve gait recognition in wild", "idea": "Temporal modeling via channel switching", "techniques": "Channel frame switching, dual-branch design, non-cyclic sampling", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "GREW, Gait3D", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1145/3503161.3547897", "bibtex": "@inproceedings{zheng2022mstgait,\n title = {Gait recognition in the wild with multi-hop temporal switch},\n author = {Zheng, Jinkai and Liu, Xinchen and Gu, Xiaoyan and Sun, Yaoqi and Gan, Chuang and Zhang, Jiyong and Liu, Wu and Yan, Chenggang},\n year = {2022},\n booktitle = {Proceedings of the 30th ACM International Conference on Multimedia},\n pages = {6136--6145}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 74 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Strong Gait Backbones" ], "abstract": "Existing studies for gait recognition are dominated by in-the-lab scenarios. Since people live in real-world senses, gait recognition in the wild is a more practical problem that has recently attracted the attention of the community of multimedia and computer vision. Current methods that obtain state-of-the-art performance on inthe-lab benchmarks achieve much worse accuracy on the recently proposed in-the-wild datasets because these methods can hardly model the varied temporal dynamics of gait sequences in unconstrained scenes. Therefore, this paper presents a novel multi-hop temporal switch method to achieve effective temporal modeling of gait patterns in real-world scenes. Concretely, we design a novel gait recognition network, named Multi-hop Temporal Switch Network (MTSGait), to learn spatial features and multi-scale temporal features simultaneously. Different from existing methods that use 3D convolutions for temporal modeling, our MTSGait models the temporal dynamics of gait sequences by 2D convolutions. By this means, it achieves high efficiency with fewer model parameters and reduces the difficulty in optimization compared with 3D convolution-based models. Based on the specific design of the 2D convolution kernels, our method can eliminate the misalignment of features among adjacent frames. In addition, a new sampling strategy, i.e., non-cyclic continuous sampling, is proposed to make the model learn more robust temporal features. Finally, the proposed method achieves superior performance on two public gait in-the-wild datasets, i.e., GREW and Gait3D, compared with state-of-the-art methods.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1145/3503161.3547897", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "xiang2022multi-acmmm-2022", "citationKey": "xiang2022multi", "archiveKey": "xiang2022multi", "method": "", "title": "Multi-view gait video synthesis", "authors": "Xiang, Weilai and Yang, Hongyu and Huang, Di and Wang, Yunhong", "year": 2022, "venue": "ACM MM", "motivation": "Enhance gait recognition robustness", "idea": "Multi-view gait video synthesis", "techniques": "Two-branch decoupling, silhouette-guided feature transform, identity-aware loss functions", "inputs": "RGB, Silhouette", "backbone": "CNN", "loss": "CE", "datasets": "CASIA-B", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1145/3503161.3547941", "bibtex": "@inproceedings{xiang2022multi,\n title = {Multi-view gait video synthesis},\n author = {Xiang, Weilai and Yang, Hongyu and Huang, Di and Wang, Yunhong},\n year = {2022},\n booktitle = {Proceedings of the 30th ACM International Conference on Multimedia},\n pages = {6783--6791}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 76 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "This paper investigates a new fine-grained video generation task, namely Multi-view Gait Video Synthesis, where the generation model works on a video of a walking human of arbitrary viewpoint and creates multi-view renderings of the subject. This task is particularly challenging, as it requires synthesizing visually plausible results, while simultaneously preserving discriminative gait cues subject to identification. To tackle the challenge caused by the entanglement of viewpoint, texture, and body structure, we present a network with two collaborative branches to decouple the novel view rendering process into two streams for human appearances (texture) and silhouettes (structure), respectively. Additionally, the prior knowledge of person re-identification and gait recognition is incorporated into the training loss for more adequate and accurate dynamic details. Experimental results show that the presented method is able to achieve promising success rates when attacking state-of-the-art gait recognition models. Furthermore, the method can improve gait recognition systems by effective data augmentation. To the best of our knowledge, this is the first task to manipulate views for human videos with person-specific behavioral constraints.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1145/3503161.3547941", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-filipi2022gait", "citationKey": "filipi2022gait", "archiveKey": "filipi2022gait", "method": "", "title": "Gait recognition based on deep learning: a survey", "authors": "Filipi Goncalves dos Santos, Claudio and Oliveira, Diego De Souza and A. Passos, Leandro and Goncalves Pires, Rafael and Felipe Silva Santos, Daniel and Pascotti Valem, Lucas and P. Moreira, Thierry and Cleison S. Santana, Marcos and Roder, Mateus and Paulo Papa, Jo and others", "year": 2022, "venue": "CSUR", "motivation": "Rapid deep-learning progress has fragmented the gait recognition literature.", "idea": "Organize methods, datasets and evaluation challenges in a unified review.", "techniques": "Literature synthesis; deep-learning taxonomy; benchmark comparison.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@article{filipi2022gait,\n title = {Gait recognition based on deep learning: a survey},\n author = {Filipi Goncalves dos Santos, Claudio and Oliveira, Diego De Souza and A. Passos, Leandro and Goncalves Pires, Rafael and Felipe Silva Santos, Daniel and Pascotti Valem, Lucas and P. Moreira, Thierry and Cleison S. Santana, Marcos and Roder, Mateus and Paulo Papa, Jo and others},\n year = {2022},\n journal = {ACM Computing Surveys (CSUR)},\n volume = {55},\n number = {2},\n pages = {1--34},\n publisher = {Acm New York, NY}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Introduction / Context" ], "abstract": "This survey reviews deep-learning approaches to gait-based identification. It categorizes datasets, network architectures and recognition strategies, discussing their benefits and limitations under appearance and acquisition variability. The paper provides an overview of how learned representations address weaknesses of earlier feature extraction and classification methods.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2201.03323", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-gu2022clothes", "citationKey": "gu2022clothes", "archiveKey": "gu2022clothes", "method": "", "title": "Clothes-changing person re-identification with rgb modality only", "authors": "Gu, Xinqian and Chang, Hong and Ma, Bingpeng and Bai, Shutao and Shan, Shiguang and Chen, Xilin", "year": 2022, "venue": "CVPR", "motivation": "Appearance matching is sensitive to clothing changes in RGB observations.", "idea": "Suppress clothing information while retaining identity cues.", "techniques": "Clothes-based adversarial loss; RGB representation learning; cloth-changing evaluation.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2022/html/Gu_Clothes-Changing_Person_Re-Identification_With_RGB_Modality_Only_CVPR_2022_paper.html", "bibtex": "@inproceedings{gu2022clothes,\n title = {Clothes-changing person re-identification with rgb modality only},\n author = {Gu, Xinqian and Chang, Hong and Ma, Bingpeng and Bai, Shutao and Shan, Shiguang and Chen, Xilin},\n year = {2022},\n booktitle = {Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},\n pages = {1060--1069}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Broader Applications" ], "abstract": "The key to address clothes-changing person reidentification (re-id) is to extract clothes-irrelevant features, e.g., face, hairstyle, body shape, and gait. Most current works mainly focus on modeling body shape from multimodality information (e.g., silhouettes and sketches), but do not make full use of the clothes-irrelevant information in the original RGB images. In this paper, we propose a Clothes-based Adversarial Loss (CAL) to mine clothesirrelevant features from the original RGB images by penalizing the predictive power of re-id model w.r.t. clothes. Extensive experiments demonstrate that using RGB images only, CAL outperforms all state-of-the-art methods on widely-used clothes-changing person re-id benchmarks. Besides, compared with images, videos contain richer appearance and additional temporal information, which can be used to model proper spatiotemporal patterns to assist clothes-changing re-id. Since there is no publicly available clothes-changing video re-id dataset, we contribute a new dataset named CCVID and show that there exists much room for improvement in modeling spatiotemporal information. The code and new dataset are available at: https://github.com/guxinqian/Simple-CCReID.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2022/html/Gu_Clothes-Changing_Person_Re-Identification_With_RGB_Modality_Only_CVPR_2022_paper.html", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "chai2022lagrange-cvpr-2022", "citationKey": "chai2022lagrange", "archiveKey": "chai2022lagrange", "method": "LagrangeGait", "title": "Lagrange motion analysis and view embeddings for improved gait recognition", "authors": "Chai, Tianrui and Li, Annan and Zhang, Shaoxiong and Li, Zilong and Wang, Yunhong", "year": 2022, "venue": "CVPR", "motivation": "Overcome cross-view variations", "idea": "Multi-branch motion-aware model", "techniques": "Combine view, appearance, motion", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet, View Loss", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2022/html/Chai_Lagrange_Motion_Analysis_and_View_Embeddings_for_Improved_Gait_Recognition_CVPR_2022_paper.html", "bibtex": "@inproceedings{chai2022lagrange,\n title = {Lagrange motion analysis and view embeddings for improved gait recognition},\n author = {Chai, Tianrui and Li, Annan and Zhang, Shaoxiong and Li, Zilong and Wang, Yunhong},\n year = {2022},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n pages = {20249--20258}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 21 }, { "source": "Manuscript Table I" }, { "source": "Verified publication metadata" } ], "summaryStatus": "Manuscript summary", "sections": [ "Background / Historical Evolution", "Advances in Methods / Handling Real-World Noise" ], "abstract": "Gait is considered the walking pattern of human body, which includes both shape and motion cues. However, the main-stream appearance-based methods for gait recognition rely on the shape of silhouette. It is unclear whether motion can be explicitly represented in the gait sequence modeling. In this paper, we analyzed human walking using the Lagrange’s equation and come to the conclusion that second-order information in the temporal dimension is necessary for identification. We designed a second-order motion extraction module based on the conclusions drawn. Also, a light weight view-embedding module is designed by analyzing the problem that current methods to cross-view task do not take view itself into consideration explicitly. Experiments on CASIA-B and OU-MVLP datasets show the effectiveness of our method and some visualization for extracted motion are done to show the interpretability of our motion extraction module.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2022/html/Chai_Lagrange_Motion_Analysis_and_View_Embeddings_for_Improved_Gait_Recognition_CVPR_2022_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "zheng2022gait-cvpr-2022", "citationKey": "zheng2022gait", "archiveKey": "zheng2022gait", "method": "SMPLGait", "title": "Gait recognition in the wild with dense 3d representations and a benchmark", "authors": "Zheng, Jinkai and Liu, Xinchen and Liu, Wu and He, Lingxiao and Yan, Chenggang and Mei, Tao", "year": 2022, "venue": "CVPR", "motivation": "Address view and shape ambiguity", "idea": "Dual-branch 3D-aware network", "techniques": "Integrate 3D SMPL with silhouettes", "inputs": "Silhouette, SMPL", "backbone": "CNN", "loss": "CE, Triplet, View Loss", "datasets": "Gait3D", "codeAvailable": "Yes", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2022/html/Zheng_Gait_Recognition_in_the_Wild_With_Dense_3D_Representations_and_CVPR_2022_paper.html", "bibtex": "@inproceedings{zheng2022gait,\n title = {Gait recognition in the wild with dense 3d representations and a benchmark},\n author = {Zheng, Jinkai and Liu, Xinchen and Liu, Wu and He, Lingxiao and Yan, Chenggang and Mei, Tao},\n year = {2022},\n booktitle = {Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},\n pages = {20228--20237}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 22 }, { "source": "Manuscript Table I" }, { "source": "Verified publication metadata" } ], "summaryStatus": "Manuscript summary", "sections": [ "Introduction / Motivation", "Background / Historical Evolution", "Background / Broader Applications", "Advances in Benchmarks / Popular Datasets", "Advances in Benchmarks / Public Codebases", "Advances in Benchmarks / Research Projects", "Advances in Methods / Ideal Gait Modalities" ], "abstract": "Existing studies for gait recognition are dominated by 2D representations like the silhouette or skeleton of the human body in constrained scenes. However, humans live and walk in the unconstrained 3D space, so projecting the 3D human body onto the 2D plane will discard a lot of crucial information like the viewpoint, shape, and dynamics for gait recognition. Therefore, this paper aims to explore dense 3D representations for gait recognition in the wild, which is a practical yet neglected problem. In particular, we propose a novel framework to explore the 3D Skinned Multi-Person Linear (SMPL) model of the human body for gait recognition, named SMPLGait. Our framework has two elaborately-designed branches of which one extracts appearance features from silhouettes, the other learns knowledge of 3D viewpoints and shapes from the 3D SMPL model. In addition, due to the lack of suitable datasets, we build the first large-scale 3D representationbased gait recognition dataset, named Gait3D. It contains 4,000 subjects and over 25,000 sequences extracted from 39 cameras in an unconstrained indoor scene. More importantly, it provides 3D SMPL models recovered from video frames which can provide dense 3D information of body shape, viewpoint, and dynamics. Based on Gait3D, we comprehensively compare our method with existing gait recognition approaches, which reflects the superior performance of our framework and the potential of 3D representations for gait recognition in the wild. The code and dataset are available at: https://gait3d.github.io.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2022/html/Zheng_Gait_Recognition_in_the_Wild_With_Dense_3D_Representations_and_CVPR_2022_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-zhang2022bytetrack", "citationKey": "zhang2022bytetrack", "archiveKey": "zhang2022bytetrack", "method": "", "title": "Bytetrack: Multi-object tracking by associating every detection box", "authors": "Zhang, Yifu and Sun, Peize and Jiang, Yi and Yu, Dongdong and Weng, Fucheng and Yuan, Zehuan and Luo, Ping and Liu, Wenyu and Wang, Xinggang", "year": 2022, "venue": "ECCV", "motivation": "Discarding low-confidence detections fragments tracks under occlusion.", "idea": "Associate detections at multiple confidence levels.", "techniques": "Two-stage data association; low-score box recovery; multi-object tracking.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1007/978-3-031-20047-2_1", "bibtex": "@inproceedings{zhang2022bytetrack,\n title = {Bytetrack: Multi-object tracking by associating every detection box},\n author = {Zhang, Yifu and Sun, Peize and Jiang, Yi and Yu, Dongdong and Weng, Fucheng and Yuan, Zehuan and Luo, Ping and Liu, Wenyu and Wang, Xinggang},\n year = {2022},\n booktitle = {European conference on computer vision},\n pages = {1--21}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Task Definition" ], "abstract": "ByteTrack revisits the practice of discarding low-confidence detections in multi-object tracking. It matches these detections against existing trajectories so that partially occluded objects can remain associated without admitting every background response. Experiments integrate this association strategy into several trackers and evaluate a simple implementation on established tracking benchmarks.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/2110.06864", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "liang2022gaitedge-eccv-2022", "citationKey": "liang2022gaitedge", "archiveKey": "liang2022gaitedge", "method": "GaitEdge", "title": "Gaitedge: Beyond plain end-to-end gait recognition for better practicality", "authors": "Liang, Junhao and Fan, Chao and Hou, Saihui and Shen, Chuanfu and Huang, Yongzhen and Yu, Shiqi", "year": 2022, "venue": "ECCV", "motivation": "Eliminate RGB-induced recognition noise", "idea": "Edge-based synthesis and alignment", "techniques": "Edge-enhanced silhouette feature encoder", "inputs": "RGB, Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, TTG-200", "codeAvailable": "OpenGait", "codeUrl": "", "url": "https://arxiv.org/abs/2203.03972", "bibtex": "@inproceedings{liang2022gaitedge,\n title = {Gaitedge: Beyond plain end-to-end gait recognition for better practicality},\n author = {Liang, Junhao and Fan, Chao and Hou, Saihui and Shen, Chuanfu and Huang, Yongzhen and Yu, Shiqi},\n year = {2022},\n booktitle = {European conference on computer vision},\n pages = {375--390}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 37 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Background / Task Definition", "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Handling Real-World Noise" ], "abstract": "Gait is one of the most promising biometrics to identify individuals at a long distance. Although most previous methods have focused on recognizing the silhouettes, several end-to-end methods that extract gait features directly from RGB images perform better. However, we demonstrate that these end-to-end methods may inevitably suffer from the gait-irrelevant noises, i.e., low-level texture and colorful information. Experimentally, we design the cross-domain evaluation to support this view. In this work, we propose a novel end-to-end framework named GaitEdge which can effectively block gait-irrelevant information and release end-to-end training potential. Specifically, GaitEdge synthesizes the output of the pedestrian segmentation network and then feeds it to the subsequent recognition network, where the synthetic silhouettes consist of trainable edges of bodies and fixed interiors to limit the information that the recognition network receives. Besides, GaitAlign for aligning silhouettes is embedded into the GaitEdge without losing differentiability. Experimental results on CASIA-B and our newly built TTG-200 indicate that GaitEdge significantly outperforms the previous methods and provides a more practical end-to-end paradigm. All the source code are available at https://github.com/ShiqiYu/OpenGait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://arxiv.org/abs/2203.03972", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "dou2022metagait-eccv-2022", "citationKey": "dou2022metagait", "archiveKey": "dou2022metagait", "method": "MetaGait", "title": "Metagait: Learning to learn an omni sample adaptive representation for gait recognition", "authors": "Dou, Huanzhang and Zhang, Pengyi and Su, Wei and Yu, Yunlong and Li, Xi", "year": 2022, "venue": "ECCV", "motivation": "Enhance attention adaptiveness for gait", "idea": "Omni-scale, -dimension, meta-attention", "techniques": "Meta-learned adaptive attention mechanism", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://arxiv.org/abs/2306.03445", "bibtex": "@inproceedings{dou2022metagait,\n title = {Metagait: Learning to learn an omni sample adaptive representation for gait recognition},\n author = {Dou, Huanzhang and Zhang, Pengyi and Su, Wei and Yu, Yunlong and Li, Xi},\n year = {2022},\n booktitle = {European conference on computer vision},\n pages = {357--374}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 38 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait recognition, which aims at identifying individuals by their walking patterns, has recently drawn increasing research attention. However, gait recognition still suffers from the conflicts between the limited binary visual clues of the silhouette and numerous covariates with diverse scales, which brings challenges to the model’s adaptiveness. In this paper, we address this conflict by developing a novel MetaGait that learns to learn an omni sample adaptive representation. Towards this goal, MetaGait injects meta-knowledge, which could guide the model to perceive sample-specific properties, into the calibration network of the attention mechanism to improve the adaptiveness from the omni-scale, omni-dimension, and omni-process perspectives. Specifically, we leverage the meta-knowledge across the entire process, where Meta Triple Attention and Meta Temporal Pooling are presented respectively to adaptively capture omni-scale dependency from spatial/channel/temporal dimensions simultaneously and to adaptively aggregate temporal information through integrating the merits of three complementary temporal aggregation methods. Extensive experiments demonstrate the state-of-the-art performance of the proposed MetaGait. On CASIA-B, we achieve rank-1 accuracy of 98.7%, 96.0%, and 89.3% under three conditions, respectively. On OU-MVLP, we achieve rank-1 accuracy of 92.4%.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://arxiv.org/abs/2306.03445", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-hid2022", "citationKey": "hid2022", "archiveKey": "hid2022", "method": "", "title": "HID 2022: The 3rd International Competition on Human Identification at a Distance", "authors": "Yu, Shiqi and Huang, Yongzhen and Wang, Liang and Makihara, Yasushi and Wang, Shengjin and Ahad, Md Atiqur Rahman and Nixon, Mark", "year": 2022, "venue": "IJCB", "motivation": "Gait recognition needs common evaluation for identification at a distance.", "idea": "Organize the third HID competition and compare participating systems.", "techniques": "Shared benchmark protocols; system evaluation; competition analysis.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/IJCB54206.2022.10007993", "bibtex": "@inproceedings{hid2022,\n title = {HID 2022: The 3rd International Competition on Human Identification at a Distance},\n author = {Yu, Shiqi and Huang, Yongzhen and Wang, Liang and Makihara, Yasushi and Wang, Shengjin and Ahad, Md Atiqur Rahman and Nixon, Mark},\n year = {2022},\n booktitle = {2022 IEEE International Joint Conference on Biometrics (IJCB)},\n doi = {10.1109/IJCB54206.2022.10007993}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Advances in Benchmarks / Competitions" ], "abstract": "The HID 2022 report documents the third competition on identifying people at a distance. It describes the dataset, evaluation rules, two-stage submission process and top-performing methods. The analysis compares the resulting accuracies with earlier editions and discusses what the improvements imply for gait recognition research.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://doi.org/10.1109/IJCB54206.2022.10007993", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "han2022unified-pr-2022", "citationKey": "han2022unified", "archiveKey": "han2022unified", "method": "A-Softmax + Triplet", "title": "A unified perspective of classification-based loss and distance-based loss for cross-view gait recognition", "authors": "Han, Feng and Li, Xuejian and Zhao, Jian and Shen, Furao", "year": 2022, "venue": "PR", "motivation": "Cross-view gait recognition improvement", "idea": "Enhance feature separability and discrimination", "techniques": "Angular softmax loss, triplet loss, batch normalization", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, TUM GAID", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2021.108519", "bibtex": "@article{han2022unified,\n title = {A unified perspective of classification-based loss and distance-based loss for cross-view gait recognition},\n author = {Han, Feng and Li, Xuejian and Zhao, Jian and Shen, Furao},\n year = {2022},\n journal = {Pattern Recognition},\n volume = {125},\n pages = {108519},\n publisher = {Elsevier}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 101 }, { "source": "Author paper summaries", "sheet": "journals", "row": 8 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Gait Learning Frameworks", "Advances in Methods / Gait Metric Designs" ], "abstract": "Gait can be used to recognize people in an uncooperative and noninvasive manner and it is hard to imitate or counterfeit, which makes it suitable for video surveillance. The current solutions for gait recognition are still not robust to handle the conditions when the view angles of the gallery and query are different. We improve the performance of cross-view gait recognition from the perspective of metric learning. Specifically, we propose to use angular softmax loss to impose an angular margin for extracting separable features. At the same time, we use triplet loss to make the extracted features more discriminative. Additionally, we add a batch-normalization layer after extracting gait features to effectively optimize two different losses. We evaluate our approach on two widely-used gait dataset: CASIA-B dataset and TUM GAID dataset. The experiment results show that our approach outperforms the prior state-of-the-art approaches, which shows the effectiveness of our approach.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2021.108519", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "li2022gaitslice-pr-2022", "citationKey": "li2022gaitslice", "archiveKey": "li2022gaitslice", "method": "GaitSlice", "title": "GaitSlice: A Gait Recognition Model Based on Spatio-temporal Slice Features", "authors": "Huakang Li and Yidan Qiu and Huimin Zhao and Jin Zhan and Rongjun Chen and Tuanjie Wei and Zhihui Huang", "year": 2022, "venue": "PR", "motivation": "Improve fine-grained gait recognition", "idea": "Highlight key frames and body correlations", "techniques": "Body part association, frame attention, feature slicing", "inputs": "Silhouette", "backbone": "CNN", "loss": "Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2021.108453", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 102 }, { "source": "Author paper summaries", "sheet": "journals", "row": 9 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "GaitSlice--A-gait-recognition-model-based-on-spatio-temp_2022_Pattern-Recogn.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Improving the performance of gait recognition under multiple camera views (i.e., cross-view gait recognition) and various conditions is urgent. From observation, we find that adjacent body parts are interrelated while walking, and each frame in a gait sequence possesses different degrees of semantic information. In this paper, we propose a novel model, GaitSlice, to analyze the human gait based on spatiotemporal slice features. Spatially, we design Slice Extraction Device (SED) to form top-down inter-related slice features. Temporally, we introduce Residual Frame Attention Mechanism (RFAM) to acquire and highlight the key frames. To better simulate reality, GaitSlice combines parallel RFAMs with inter-related slice features to focus on the features’ spatio-temporal information. We evaluate our model on CASIA-B and OU-MVLP gait datasets and compare it with six typical gait recognition models by using rank-1 accuracy. The results show that GaitSlice achieves high accuracy in gait recognition under cross-view and various walking conditions.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2021.108453", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "liu2022symmetry-pr-2022", "citationKey": "liu2022symmetry", "archiveKey": "liu2022symmetry", "method": "SDHF-GCN", "title": "Symmetry-driven hyper feature GCN for skeleton-based gait recognition", "authors": "Liu, Xiaokai and You, Zhaoyang and He, Yuxiang and Bi, Sheng and Wang, Jie", "year": 2022, "venue": "PR", "motivation": "Enhance skeleton-based gait recognition", "idea": "Exploit symmetry and multi-level features", "techniques": "Graph convolution, symmetry adjacency, hyperfeatures", "inputs": "Skeleton", "backbone": "CNN", "loss": "CE", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2022.108520", "bibtex": "@article{liu2022symmetry,\n title = {Symmetry-driven hyper feature GCN for skeleton-based gait recognition},\n author = {Liu, Xiaokai and You, Zhaoyang and He, Yuxiang and Bi, Sheng and Wang, Jie},\n year = {2022},\n journal = {Pattern Recognition},\n volume = {125},\n pages = {108520},\n publisher = {Elsevier}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 103 }, { "source": "Author paper summaries", "sheet": "journals", "row": 10 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Strong Gait Backbones" ], "abstract": "Gait recognition, as an attractive task in biometrics, remains challenging due to significant intra-class changes of clothing and pose variations across different cameras. Recent approaches mainly focus on silhouette-based gait mode, which is easy to model in Convolutional Neural Networks (CNNs). Compared with silhouettes, the dynamics of skeletons essentially convey more robust information, which is invariant to view and clothing changes. Conventional approaches for modeling skeletons usually rely on handcrafted features or traversal rules, thus resulting in limited expressive power and difficulties of generalization. In this work, we address the skeleton-based gait recognition task with a novel Symmetry-Driven Hyper Feature Graph Convolutional Network (SDHF-GCN), which goes beyond the limitations of previous approaches by automatically learning multiple dynamic patterns and hierarchical semantic features in a unified Graph Convolutional Network (GCN). This model involves three dynamic patterns: natural connection, temporal correlation and symmetric interaction, which enriches the description of dynamic patterns by exploiting symmetry perceptual principles. Furthermore, a hyper feature network is proposed to aggregate the hierarchical semantic features, including dynamic features at the high level, structured features at the intermediate level, and static features at the low level, which complement each other to enhance the discriminative ability. By integrating different patterns in the hierarchical structure, the model is able to generate versatile and discriminative representations, thus improving the recognition rate. On the CASIA-B and OUMVLP-Pose datasets, the proposed SDHF-GCN renders substantial improvements over mainstream methods, especially in the coat-wearing scenario, with superior robustness to covariate factors.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2022.108520", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "song2022distilled-prl-2022", "citationKey": "song2022distilled", "archiveKey": "song2022distilled", "method": "DLGS", "title": "Distilled Light GaitSet: Towards Scalable Gait Recognition", "authors": "Xu Song and Yan Huang and Yan Huang and Caifeng Shan and Jilong Wang and Yu Chen", "year": 2022, "venue": "PRL", "motivation": "Enable gait recognition on limited devices", "idea": "Lightweight model with distilled knowledge", "techniques": "Network simplification; multi-level knowledge transfer", "inputs": "Silhouette", "backbone": "CNN", "loss": "Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1016/j.patrec.2022.03.019", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 114 }, { "source": "Author paper summaries", "sheet": "journals", "row": 21 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Distilled-light-GaitSet--Towards-scalable-gait-r_2022_Pattern-Recognition-Le.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition has made significant progress recently. However, most of existing methods utilize complicated neural networks, which lead to high computation cost. In this paper, a lightweight model named Distilled Light GaitSet (DLGS) is proposed for efficient gait recognition. More specifically, a lightweight CNN is designed for efficient computation, and a Joint Knowledge Distillation algorithm is proposed to boost the accuracy of the simplified model. Extensive experiments on the CASIA-B dataset and the OU- MVLP dataset show that the proposed DLGS can reduce the number of parameters and computation cost significantly while achieving the state-of-the-art performance.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patrec.2022.03.019", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "chen2022gait-tbiom-2022", "citationKey": "chen2022gait", "archiveKey": "chen2022gait", "method": "GPAN", "title": "Gait Pyramid Attention Network Toward Silhouette Semantic Relation Learning for Gait Recognition", "authors": "Jianyu Chen and Zhongyuan Wang and Peng Yi and Kangli Zeng and Zheng He and Qin Zou", "year": 2022, "venue": "TBIOM", "motivation": "Robust gait recognition under covariates", "idea": "Multi-scale attention-enhanced feature learning", "techniques": "Pyramidal convolutions; local–global attention fusion", "inputs": "Silhouette", "backbone": "CNN, Attention", "loss": "Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TBIOM.2022.3213545", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 122 }, { "source": "Author paper summaries", "sheet": "journals", "row": 29 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Gait_Pyramid_Attention_Network_Toward_Silhouette_Semantic_Relation_Learning_for_Gait_Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition refers to video-based biometric techniques for recognizing subjects by walking patterns under a long-distance situation. Despite the progress of existing gait recognition methods, the recognition ability of subjects in carrying situations is still limited (e.g., carrying bags or wearing coats or jackets). To address this issue, this paper proposes to extract human gait information from different ranges of receptive fields, providing richer internal features of the deep network. Moreover, we propose two attention mechanisms, Local Pyramid Attention and Global Attention Fusion Learning, to focus on the key features in human gait from different perspectives. Depending on the different attention mechanisms employed in the network, three network variants are derived, where the Gait Pyramid Attention Network (GPAN) contains two attention mechanisms, while GPAN-P and GPAN-L contain a single attention mechanism. We evaluated our method on two large datasets, CASIA-B and OUMVLP. Experiments show that the proposed network gives an average rank-1 accuracy of 97.8% on CASIA-B under normal walking conditions. We also achieve 94.2% and 81.8% accuracy on CASIA-B dataset under the complex bag-walking and coat-walking scenarios, which are dramatically superior to the state-of-the-art methods.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TBIOM.2022.3213545", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "li2022multi-tbiom-2022", "citationKey": "li2022multi", "archiveKey": "li2022multi", "method": "OU-MVLP-Mesh", "title": "Multi-View Large Population Gait Database With Human Meshes and Its Performance Evaluation", "authors": "Xiang Li and Yasushi Makihara and Chi Xu and Yasushi Yagi", "year": 2022, "venue": "TBIOM", "motivation": "Robust gait recognition under covariates", "idea": "3D mesh-based multi-view modeling", "techniques": "Human mesh estimation; asynchronous multi-view fusion; dual-stream training framework", "inputs": "Silhouette, RGB", "backbone": "CNN", "loss": "", "datasets": "OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TBIOM.2022.3174559", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 123 }, { "source": "Author paper summaries", "sheet": "journals", "row": 30 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Multi-View_Large_Population_Gait_Database_With_Human_Meshes_and_Its_Performance_Evaluation.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Existing model-based gait databases provide the 2D poses (i.e., joint locations) extracted by general pose estimators as the human model. However, these 2D poses suffer from information loss and are of relatively low quality. In this paper, we consider a more informative 3D human mesh model with parametric pose and shape features, and propose a multi-view training framework for accurate mesh estimation. Unlike existing methods, which estimate a mesh from a single view and suffer from the ill-posed estimation problem in 3D space, the proposed framework takes asynchronous multi-view gait sequences as input and uses both multi-view and single-view streams to learn consistent and accurate mesh models for both multi-view and single-view sequences. After applying the proposed framework to the existing OU-MVLP database, we establish a large-scale gait database with human meshes (i.e., OUMVLP-Mesh), containing over 10,000 subjects and up to 14 view angles. Experimental results show that the proposed framework estimates human mesh models more accurately than similar methods, providing models of sufficient quality to improve the recognition performance of a baseline model-based gait recognition approach.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TBIOM.2022.3174559", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "chen2022understanding-tcsvt-2022", "citationKey": "chen2022understanding", "archiveKey": "chen2022understanding", "method": "CSTAN", "title": "Understanding Dynamic Associations Gait Recognition via Cross-view Spatiotemporal Aggregation Network", "authors": "Jianyu Chen and Zhongyuan Wang and Caixia Zheng and Kangli Zeng and Qin Zou and Zixiang Xiong", "year": 2022, "venue": "TCSVT", "motivation": "Robust gait recognition under variations", "idea": "Multi-scale temporal feature aggregation", "techniques": "Partial feature extraction, temporal interaction, view selection", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TCSVT.2022.3185642", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 141 }, { "source": "Author paper summaries", "sheet": "journals", "row": 47 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Understanding_Dynamic_Associations_Gait_Recognition_via_Cross-view_Spatiotemporal_Aggregation_Network.pdf", "publicationStatus": "Withdrawn", "notes": "The publisher notice states that this early-access article will not be published in its final form and should not be cited. Retained for archive transparency; excluded from the word cloud.", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition is a technique to recognize subjects under long-distance conditions. Existing gait recognition methods generally rely on extracting different motion representations of body parts to capture the gait information. However, the differences in appearance silhouettes between subjects are not discriminative enough to identify the distinct gaits. Therefore, exploring the representations of unique motion differences of subjects from the temporal domain is an effective way to boost gait recognition performance. In this paper, we propose a cross-view spatiotemporal aggregation network (CSTAN) to learn the most discriminative motion representations of subjects from multi-scale features. Particularly, CSTAN uses temporal contextual feature aggregation (TCFA) block to interact multiscale features, increasing the diversity of temporal representations in feature sequences and producing suitable motion representations. Besides, a potential spatial feature assessment (PSFA) block is proposed to address the problem of missing silhouettes due to occlusion. Considering the multi-view context, we propose a cross-view prediction learning (CVPL) block to mitigate the impact of the view itself. By aggregating features in the spatiotemporal and view domains, CSTAN can learn a more appropriate representation of human motion differences. Extensive experiments on two public gait recognition datasets demonstrate the state-of-the-art performance of our proposed CSTAN. It is worth mentioning that the performance of CSTAN achieves a breakthrough in some complex environments, e.g., the subjects wearing different coats or carrying a bag.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TCSVT.2022.3185642", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "yao2021collaborative-tcsvt-2022", "citationKey": "yao2021collaborative", "archiveKey": "yao2021collaborative", "method": "Collaborative Feature Learning", "title": "Collaborative Feature Learning for Gait Recognition Under Cloth Changes", "authors": "Lingxiang Yao and Worapan Kusakunniran and Qiang Wu and Jingsong Xu and Jian Zhang", "year": 2022, "venue": "TCSVT", "motivation": "Robust gait recognition under clothing", "idea": "Combine part-based and skeleton features", "techniques": "Local temporal pooling with transformer", "inputs": "Silhouette, Skeleton, RGB", "backbone": "CNN, Transformer", "loss": "Triplet", "datasets": "CASIA-B", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TCSVT.2021.3112564", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 138 }, { "source": "Author paper summaries", "sheet": "journals", "row": 44 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Collaborative_Feature_Learning_for_Gait_Recognition_Under_Cloth_Changes.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Since gait can be utilized to identify individuals from a far distance without their interaction and coordination, recently many gait recognition methods have been proposed. However, due to a real-world scenario of clothing changes, a degradation occurs for most of these methods. Thus in this paper, a more efficient gait recognition method is proposed to address the problem of clothing variances. First, part-based gait features are formulated from two different perspectives, i.e., the separated body parts that are more robust to clothing changes and the estimated human skeleton key-point regions. It is reasonable to formulate such features for cloth-changing gait recognition, because these two perspectives are both less vulnerable to clothing changes. Given that each feature has its own advantages and disadvantages, a more efficient gait feature is generated in this paper by assembling these tw o features together. Moreover, since local features are more discriminative than global features, in this paper more attention is focused on the local short-range features. Also, unlike most methods, in our method we treat the estimated key-point features as a set of word embeddings, and a transformer encoder is specifically used to learn the dependence of each correlative key-points. The robustness and effectiveness of our proposed method are certified by experiments on CASIA Gait Dataset B, and it has achieved the state-of-the-art performance on this dataset.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TCSVT.2021.3112564", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "huang2022enhanced-tcsvt-2022", "citationKey": "huang2022enhanced", "archiveKey": "huang2022enhanced", "method": "ESNet", "title": "Enhanced Spatial-Temporal Salience for Cross-View Gait Recognition", "authors": "Tianhuan Huang and Xianye Ben and Chen Gong and Baochang Zhang and Rui Yan and Qiang Wu", "year": 2022, "venue": "TCSVT", "motivation": "Robust cross-view gait recognition", "idea": "Integrate spatial and temporal features", "techniques": "Parallel convolution with attention, context-aware pooling", "inputs": "Silhouette", "backbone": "CNN, Attention", "loss": "Triplet", "datasets": "CASIA-B, OU-LP, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TCSVT.2022.3175959", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 139 }, { "source": "Author paper summaries", "sheet": "journals", "row": 45 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Enhanced_Spatial-Temporal_Salience_for_Cross-View_Gait_Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition can be used in person identification and re-identification by itself or in conjunction with other bio-2 metrics. Although gait has both spatial and temporal attributes, and it has been observed that decoupling spatial feature and temporal feature can better exploit the gait feature on the fine-5 grained level. However, the spatial-temporal correlations of gait video signals are also lost in the decoupling process. Direct 3D7 convolution approaches can retain such correlations, but they also introduce unnecessary interferences. Instead of common 3D convolution solutions, this paper proposes an integration of decoupling process into a 3D convolution framework for cross-11 view gait recognition. In particular, a novel block consisting of a Parallel-insight Convolution layer integrated with a Spatial-13 Temporal Dual-Attention (STDA) unit is proposed as the basic block for global spatial-temporal information extraction. Under the guidance of the STDA unit, this block can well integrate spatial-temporal information extracted by two decoupled mod-17 els and at the same time retain the spatial-temporal correla-18 tions. In addition, a Multi-Scale Salient Feature Extractor is proposed to further exploit the fine-grained features through context awareness extension of part-based features and adaptively aggregating the spatial features. Extensive experiments on three popular gait datasets, namely CASIA-B, OULP and OUMVLP, demonstrate that the proposed method outperforms state-of-the-24 art methods.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TCSVT.2022.3175959", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "qin2021rpnet-tcsvt-2022", "citationKey": "qin2021rpnet", "archiveKey": "qin2021rpnet", "method": "RPNet", "title": "RPNet Gait Recognition With Relationships Between Each Body-Parts", "authors": "Hao Qin and Zhenxue Chen and Qingqiang Guo and Q. M. Jonathan Wu and Mengxu Lu", "year": 2022, "venue": "TCSVT", "motivation": "Robust gait recognition under occlusion", "idea": "Model multi-scale body part relations", "techniques": "Partial feature extraction, relation encoding, aggregation", "inputs": "Silhouette", "backbone": "CNN", "loss": "Triplet", "datasets": "CASIA-B, OU-LP, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TCSVT.2021.3095290", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 140 }, { "source": "Author paper summaries", "sheet": "journals", "row": 46 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "RPNet_Gait_Recognition_With_Relationships_Between_Each_Body-Parts.pdf", "authorsSource": "Public scholarly metadata", "abstract": "At present, many studies have shown that partitioning the gait sequence and its feature map can improve the accuracy of gait recognition. However, most models just cut the feature map at a fixed single scale, which loses the dependence between various parts. So, our paper proposes a structure called Part Feature Relationship Extractor (PFRE) to discover all of the relationships between each parts for gait recognition. The paper uses PFRE and a Convolutional Neural Network (CNN) to form the RPNet. PFRE is divided into two parts. One part that we call the Total-Partial Feature Extractor (TPFE) is used to extract the features of different scale blocks, and the other part, called the Adjacent Feature Relation Extractor (AFRE), is used to find the relationships between each block. At the same time, the paper adjusts the number of input frames during training to perform quantitative experiments and finds the rule between the number of input frames and the performance of the model. Our model is tested on three public gait datasets, CASIA-B, OU-LP and OU-MVLP. It exhibits a significant level of robustness to occlusion situations, and achieves accuracies of 92.82% and 80.26% on CASIA-B under BG # and CL # conditions, respectively. The results show that our method reaches the top level among state-of-the-art methods.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TCSVT.2021.3095290", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "zhao2021associated-tmm-2022", "citationKey": "zhao2021associated", "archiveKey": "zhao2021associated", "method": "ASTCapsNet", "title": "Associated Spatio-Temporal Capsule Network for Gait Recognition", "authors": "Aite Zhao and Junyu Dong and Jianbo Li and Lin Qi and Huiyu Zhou", "year": 2022, "venue": "TMM", "motivation": "Extract rich multimodal gait features", "idea": "Integrate spatial, temporal, structural information", "techniques": "Low/high-level feature extraction, memory module, capsule relationship layer, recurrent unit, multi-class evaluation", "inputs": "Skeleton, RGB", "backbone": "CNN, LSTM", "loss": "CE, Margin Loss", "datasets": "CASIA-A, UNITO, SDUgait, NDDs, PD", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TMM.2021.3060280", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 173 }, { "source": "Author paper summaries", "sheet": "journals", "row": 75 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Associated_Spatio-Temporal_Capsule_Network_for_Gait_Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "It is a challenging task to identify a person based on her/his gait patterns. State-of-the-art approaches rely on the analysis of temporal or spatial characteristics of gait, and gait recognition is usually performed on single modality data (such as images, skeleton joint coordinates, or force signals). Evidence has shown that using multi-modality data is more conducive to gait research. Therefore, we here establish an automated learning system, with an associated spatio-temporal capsule network (ASTCapsNet) trained on multi-sensor datasets, to analyze multimodal information for gait recognition. Specifically, we first design a low-level feature extractor and a high-level feature extractor for spatio-temporal feature extraction of gait with a novel recurrent memory unit and a relationship layer . Subsequently, a Bayesian model is employed for the decision-making of class labels. Extensive experiments on several public datasets (normal and abnormal gait) validate the effectiveness of the proposed ASTCapsNet, compared against several state-of-the-art methods.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TMM.2021.3060280", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "xu2021gait-tmm-2022", "citationKey": "xu2021gait", "archiveKey": "xu2021gait", "method": "LGSD-PSN", "title": "Gait Recognition Based on Local Graphical Skeleton Descriptor With Pairwise Similarity Network", "authors": "Ke Xu and Xinghao Jiang and Tanfeng Sun", "year": 2022, "venue": "TMM", "motivation": "Robust gait recognition under variations", "idea": "Capture local skeleton patterns", "techniques": "Local geometric descriptors, pairwise similarity network, feature encoding, fused comparison", "inputs": "Skeleton, RGB", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, NLPR, CASIA-C", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/tmm.2021.3095809", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 174 }, { "source": "Author paper summaries", "sheet": "journals", "row": 76 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition aims to identify a human through a walking sequence. It is a challenging task in computer vision since monocular camera loses most of the 3D information. Previous works described gait features with the contours of shape or the global geometrical characters of skeleton. So little work is researched on the local patterns of gait skeleton. In this paper, to resist the dress changes and speed changes, a Local Graphical Skeleton Descriptor (LGSD) is proposed to describe both the inner and intra local graphical patterns of a human gait skeleton. The gait features from the same or different identities are paired up and a Pairwise Similarity Network (PSN) is proposed to maximize the similarity of True matched pairs and minimize the similarity of False matched pairs. The contributions of our method are: 1) LGSD is proposed to describe human gait by computing four novel local geometrical patterns of skeleton sequences, which makes use of the intuitive cognition of gait based on the prior knowledge of mankind. 2) PSN is implemented by a two-stream CNN structure to build the gait model, which fused two popular gait recognition strategies. 3) The robustness of our method to dress changes and speed changes is proved on the public datasets. We have also achieved some state-of-the-art results on these datasets. The proposed method is examined on three public gait datasets which have RGB or infrared frames for evaluation: the CASIA-B dataset, the NLPR gait database, and the CASIA-C dataset. The performers in these datasets are walking under different views, speeds or dresses. The results are further compared with previous approaches to confirm the effectiveness and the advantages of our method.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/tmm.2021.3095809", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "song2022casia-tpami-2023", "citationKey": "song2022casia", "archiveKey": "song2022casia", "method": "CASIA-E", "title": "CASIA-E: A large comprehensive dataset for gait recognition", "authors": "Song, Chunfeng and Huang, Yongzhen and Wang, Weining and Wang, Liang", "year": 2022, "venue": "TPAMI", "motivation": "Large-scale outdoor gait dataset", "idea": "Capture diverse gait attributes", "techniques": "Multi-view video collection, statistical analysis, benchmarking", "inputs": "Silhouette", "backbone": "CNN", "loss": "", "datasets": "CASIA-E", "codeAvailable": "Yes", "codeUrl": "", "url": "", "bibtex": "@article{song2022casia,\n title = {CASIA-E: A large comprehensive dataset for gait recognition},\n author = {Song, Chunfeng and Huang, Yongzhen and Wang, Weining and Wang, Liang},\n year = {2022},\n journal = {IEEE transactions on pattern analysis and machine intelligence},\n volume = {45},\n number = {3},\n pages = {2801--2815},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 190 }, { "source": "Author paper summaries", "sheet": "journals", "row": 90 } ], "summaryStatus": "Author collection", "sections": [ "Background / Task Definition", "Background / Broader Applications" ], "yearNote": "Archive: 2023; bibliography: 2022", "abstract": "Gait recognition plays a special role in visual surveillance due to its unique advantage,e.g., long-distance, cross-view and non-cooperative recognition. However , it has not yet been widely applied. One reason for this awkwardness is the lack of a truly big dataset captured in practical outdoor scenarios. Here, the “big” at least means: (1) huge amount of gait videos; (2) sufficient subjects; (3) rich attributes; and (4) spatial and temporal variations. Moreover , most existing large-scale gait datasets are collected indoors, which have few challenges from real scenes, such as the dynamic and complex background clutters, illumination variations, vertical view variations,etc.I n this article, we introduce a newly built big outdoor gait dataset, called CASIA-E. It contains more than one thousand people distributed over near one million videos. Each person involves 26 view angles and varied appearances caused by changes of bag carrying, dressing and walking styles. The videos are captured across five months and across three kinds of outdoor scenes. Soft biometric features are also recorded for all subjects including age, gender , height, weight, and nationality . Besides, we report an experimental benchmark and examine some meaningful problems that have not been well studied previously ,e.g., the influence of million-level training videos, vertical view angles, walking styles, and the thermal infrared modality . We believe that such a big outdoor dataset and the experimental benchmark will promote the development of gait recognition in both academic research and industrial applications.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "chao2021gaitset-tpami-2022", "citationKey": "chao2021gaitset", "archiveKey": "chao2021gaitset", "method": "GaitSet", "title": "GaitSet Cross-View Gait Recognition Through Utilizing Gait As a Deep Set", "authors": "Hanqing Chao and Kun Wang and Yiwei He and Junping Zhang and Jianfeng Feng", "year": 2022, "venue": "TPAMI", "motivation": "Robust gait recognition under variations", "idea": "Treat gait as unordered set", "techniques": "Local feature extraction, set pooling, horizontal pyramid mapping", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1109/TPAMI.2021.3057879", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 188 }, { "source": "Author paper summaries", "sheet": "journals", "row": 88 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "GaitSet_Cross-View_Gait_Recognition_Through_Utilizing_Gait_As_a_Deep_Set.pdf", "abstract": "Gait is a unique biometric feature that can be recognized at a distance; thus, it has broad applications in crime prevention, forensic identification, and social security. To portray a gait, existing gait recognition methods utilize either a gait template which makes it difficult to preserve temporal information, or a gait sequence that maintains unnecessary sequential constraints and thus loses the flexibility of gait recognition. In this paper, we present a novel perspective that utilizes gait as a deep set, which means that a set of gait frames are integrated by a global-local fused deep network inspired by the way our left- and right-hemisphere processes information to learn information that can be used in identification. Based on this deep set perspective, our method is immune to frame permutations, and can naturally integrate frames from different videos that have been acquired under different scenarios, such as diverse viewing angles, different clothes, or different item-carrying conditions. Experiments show that under normal walking conditions, our singlemodel method achieves an average rank-1 accuracy of 96.1 percent on the CASIA-B gait dataset and an accuracy of 87.9 percent on the OU-MVLP gait dataset. Under various complex scenarios, our model also exhibits a high level of robustness. It achieves accuracies of 90.8 and 70.3 percent on CASIA-B under bag-carrying and coat-wearing walking conditions respectively, significantly outperforming the best existing methods. Moreover, the proposed method maintains a satisfactory accuracy even when only small numbers of frames are available in the test samples; for example, it achieves 85.0 percent on CASIA-B even when using only 7 frames. The source code has been released at https://github.com/AbnerHqC/GaitSet", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TPAMI.2021.3057879", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "lin2021gaitmask-bmvc-2021", "citationKey": "lin2021gaitmask", "archiveKey": "lin2021gaitmask", "method": "GaitMask", "title": "GaitMask: Mask-based Model for Gait Recognition.", "authors": "Lin, Beibei and Liu, Yu and Zhang, Shunli", "year": 2021, "venue": "BMVC", "motivation": "Enhance local-global feature completeness", "idea": "Dual-branch complementary mask learning", "techniques": "Mask-based local-global feature fusion", "inputs": "Silhouette", "backbone": "CNN", "loss": "Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@inproceedings{lin2021gaitmask,\n title = {GaitMask: Mask-based Model for Gait Recognition.},\n author = {Lin, Beibei and Liu, Yu and Zhang, Shunli},\n year = {2021},\n booktitle = {BMVC},\n pages = {363}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 16 } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait recognition is an important biometric technology that identifies a person by using walking posture. Recently, most gait recognition methods either take the human gait as a whole to generate Global Feature Representation (GFR) or equivalently divide the human gait into multiple local regions to establish Local Feature Representation (LFR). However, we observe that LFR or GFR does not adequately represent the human gait because that LFR only focuses on the detailed information of each local region and GFR pays more attention to the global context information. On the other hand, the partition manner of the local regions is fixed, which only focuses on the local information of several specific regions. Motivated by this observation, we propose a novel mask-based network, named GaitMask, for gait recognition. GaitMask is built based on the Maskbased Local Augmentation (MLA), which is used to learn more comprehensive feature representations. MLA is a dual-branch structure consisting of a GFR extraction module as the trunk and a mask-based LFR extraction module as the branch. Specifically, the mask-based LFR extraction consists of a pair of complementary masks, where one mask randomly drops a region of the input feature maps and the other one only preserves this region. The complementary mask can be used to generate more comprehensive LFR and enhances the robustness of feature representations of the trunk. Experiments on two popular datasets demonstrate that our method achieves state-of-the-art results. Specifically, the proposed method significantly increases the performance in complex environments.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "zhang2021cross-cvpr-2021", "citationKey": "zhang2021cross", "archiveKey": "zhang2021cross", "method": "KoopmanGait", "title": "Cross-view gait recognition with deep universal linear embeddings", "authors": "Zhang, Shaoxiong and Wang, Yunhong and Li, Annan", "year": 2021, "venue": "CVPR", "motivation": "Address cross-view variations", "idea": "Introduce Koopman operator", "techniques": "AE with linear dynamics constraints", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/CVPR2021/html/Zhang_Cross-View_Gait_Recognition_With_Deep_Universal_Linear_Embeddings_CVPR_2021_paper.html", "bibtex": "@inproceedings{zhang2021cross,\n title = {Cross-view gait recognition with deep universal linear embeddings},\n author = {Zhang, Shaoxiong and Wang, Yunhong and Li, Annan},\n year = {2021},\n booktitle = {Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},\n pages = {9095--9104}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 20 }, { "source": "Manuscript Table I" }, { "source": "Verified publication metadata" } ], "summaryStatus": "Manuscript summary", "sections": [ "Background / Historical Evolution", "Advances in Methods / Handling Real-World Noise" ], "abstract": "Gait is considered an attractive biometric identifier for its non-invasive and non-cooperative features compared with other biometric identifiers such as fingerprint and iris. At present, cross-view gait recognition methods always establish representations from various deep convolutional networks for recognition and ignore the potential dynamical information of the gait sequences. If assuming that pedestrians have different walking patterns, gait recognition can be performed by calculating their dynamical features from each view. This paper introduces the Koopman operator theory to gait recognition, which can find an embedding space for a global linear approximation of a nonlinear dynamical system. Furthermore, a novel framework based on convolutional variational autoencoder and deep Koopman embedding is proposed to approximate the Koopman operators, which is used as dynamical features from the linearized embedding space for cross-view gait recognition. It gives solid physical interpretability for a gait recognition system. Experiments on a large public dataset, OU-MVLP , prove the effectiveness of the proposed method.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/CVPR2021/html/Zhang_Cross-View_Gait_Recognition_With_Deep_Universal_Linear_Embeddings_CVPR_2021_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "huang20213d-iccv-2021", "citationKey": "huang20213d", "archiveKey": "huang20213d", "method": "3DLocal", "title": "3d local convolutional neural networks for gait recognition", "authors": "Huang, Zhen and Xue, Dixiu and Shen, Xu and Tian, Xinmei and Li, Houqiang and Huang, Jianqiang and Hua, Xian-Sheng", "year": 2021, "venue": "ICCV", "motivation": "Address spatial-temporal gait variation", "idea": "Part-specific adaptive volume extraction", "techniques": "Adaptive 3D local convolution", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "Yes", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/ICCV2021/html/Huang_3D_Local_Convolutional_Neural_Networks_for_Gait_Recognition_ICCV_2021_paper.html", "bibtex": "@inproceedings{huang20213d,\n title = {3d local convolutional neural networks for gait recognition},\n author = {Huang, Zhen and Xue, Dixiu and Shen, Xu and Tian, Xinmei and Li, Houqiang and Huang, Jianqiang and Hua, Xian-Sheng},\n year = {2021},\n booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision},\n pages = {14920--14929}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 45 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "The goal of gait recognition is to learn the unique spatiotemporal pattern about the human body shape from its temporal changing characteristics. As different body parts behave differently during walking, it is intuitive to model the spatio-temporal patterns of each part separately. However, existing part-based methods equally divide the feature maps of each frame into fixed horizontal stripes to get local parts. It is obvious that these stripe partition-based methods cannot accurately locate the body parts. First, different body parts can appear at the same stripe (e.g., arms and the torso), and one part can appear at different stripes in different frames (e.g., hands). Second, different body parts possess different scales, and even the same part in different frames can appear at different locations and scales. Third, different parts also exhibit distinct movement patterns (e.g., at which frame the movement starts, the position change frequency, how long it lasts). To overcome these issues, we propose novel 3D local operations as a generic family of building blocks for 3D gait recognition backbones. The proposed 3D local operations support the extraction of local 3D volumes of body parts in a sequence with adaptive spatial and temporal scales, locations and lengths. In this way, the spatio-temporal patterns of the body parts are well learned from the 3D local neighborhood in partspecific scales, locations, frequencies and lengths. Experiments demonstrate that our 3D local convolutional neural networks achieve state-of-the-art performance on popular gait datasets. Code is available at: https://github. com/yellowtownhz/3DLocalCNN.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/ICCV2021/html/Huang_3D_Local_Convolutional_Neural_Networks_for_Gait_Recognition_ICCV_2021_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "huang2021context-iccv-2021", "citationKey": "huang2021context", "archiveKey": "huang2021context", "method": "CSTL", "title": "Context-sensitive temporal feature learning for gait recognition", "authors": "Huang, Xiaohu and Zhu, Duowang and Wang, Hao and Wang, Xinggang and Yang, Bo and He, Botao and Liu, Wenyu and Feng, Bin", "year": 2021, "venue": "ICCV", "motivation": "Adaptive multi-scale motion modeling", "idea": "Joint temporal attention and spatial selection", "techniques": "Context-sensitive temporal-spatial learning", "inputs": "Silhouette", "backbone": "CNN, Attention", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "Yes", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/ICCV2021/html/Huang_Context-Sensitive_Temporal_Feature_Learning_for_Gait_Recognition_ICCV_2021_paper.html", "bibtex": "@inproceedings{huang2021context,\n title = {Context-sensitive temporal feature learning for gait recognition},\n author = {Huang, Xiaohu and Zhu, Duowang and Wang, Hao and Wang, Xinggang and Yang, Bo and He, Botao and Liu, Wenyu and Feng, Bin},\n year = {2021},\n booktitle = {Proceedings of the IEEE/CVF international conference on computer vision},\n pages = {12909--12918}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 46 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Although gait recognition has drawn increasing research attention recently, it remains challenging to learn discriminative temporal representation since the silhouette differences are quite subtle in spatial domain. Inspired by the observation that humans can distinguish gaits of different subjects by adaptively focusing on temporal sequences with different time scales, we propose a context-sensitive temporal feature learning ( CSTL) network in this paper, which aggregates temporal features in three scales to obtain motion representation according to the temporal contextual information. Specifically, CSTL introduces relation modeling among multi-scale features to evaluate feature importances, based on which network adaptively enhances more important scale and suppresses less important scale. Besides that, we propose a salient spatial feature learning (SSFL) module to tackle the misalignment problem caused by temporal operation, e.g., temporal convolution. SSFL recombines a frame of salient spatial features by extracting the most discriminative parts across the whole sequence. In this way, we achieve adaptive temporal learning and salient spatial mining simultaneously. Extensive experiments conducted on two datasets demonstrate the state-of-the-art performance. On CASIA-B dataset, we achieve rank-1 accuracies of 98.0%, 95.4% and 87.0% under normal walking, bag-carrying and coat-wearing conditions. On OU-MVLP dataset, we achieve rank-1 accuracy of 90.2%. The source code will be published at https://github.com/OliverHxh/CSTL.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/ICCV2021/html/Huang_Context-Sensitive_Temporal_Feature_Learning_for_Gait_Recognition_ICCV_2021_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "zhu2021gait-iccv-2021", "citationKey": "zhu2021gait", "archiveKey": "zhu2021gait", "method": "", "title": "Gait recognition in the wild: A benchmark", "authors": "Zhu, Zheng and Guo, Xianda and Yang, Tian and Huang, Junjie and Deng, Jiankang and Huang, Guan and Du, Dalong and Lu, Jiwen and Zhou, Jie", "year": 2021, "venue": "ICCV", "motivation": "In-the-wild Gait Recognition Benchmark", "idea": "Create GREW wild dataset", "techniques": "Large-scale annotated wild data", "inputs": "", "backbone": "", "loss": "", "datasets": "GREW", "codeAvailable": "-", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/ICCV2021/html/Zhu_Gait_Recognition_in_the_Wild_A_Benchmark_ICCV_2021_paper.html", "bibtex": "@inproceedings{zhu2021gait,\n title = {Gait recognition in the wild: A benchmark},\n author = {Zhu, Zheng and Guo, Xianda and Yang, Tian and Huang, Junjie and Deng, Jiankang and Huang, Guan and Du, Dalong and Lu, Jiwen and Zhou, Jie},\n year = {2021},\n booktitle = {Proceedings of the IEEE/CVF international conference on computer vision},\n pages = {14789--14799}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 48 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Introduction / Motivation", "Background / Historical Evolution", "Background / Broader Applications", "Advances in Benchmarks / Popular Datasets", "Advances in Benchmarks / Research Projects" ], "abstract": "Gait benchmarks empower the research community to train and evaluate high-performance gait recognition systems. Even though growing efforts have been devoted to cross-view recognition, academia is restricted by current existing databases captured in the controlled environment. In this paper, we contribute a new benchmark for Gait REcognition in the Wild (GREW). The GREW dataset is constructed from natural videos, which contains hundreds of cameras and thousands of hours streams in open systems. With tremendous manual annotations, the GREW consists of 26K identities and 128K sequences with rich attributes for unconstrained gait recognition. Moreover, we add a distractor set of over 233K sequences, making it more suitable for real-world applications. Compared with prevailing predefined cross-view datasets, the GREW has diverse and practical view variations, as well as more natural challenging factors. To the best of our knowledge, this is the first large-scale dataset for gait recognition in the wild. Equipped with this benchmark, we dissect the unconstrained gait recognition problem. Representative appearance-based and model-based methods are explored, and comprehensive baselines are established. Experimental results show (1) The proposed GREW benchmark is necessary for training and evaluating gait recognizer in the wild. (2) For state-of-the-art gait recognition approaches, there is a lot of room for improvement. (3) The GREW benchmark can be used as effective pretraining for controlled gait recognition. Benchmark website is https://www.grew-benchmark.org/.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/ICCV2021/html/Zhu_Gait_Recognition_in_the_Wild_A_Benchmark_ICCV_2021_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "lin2021gait-iccv-2021", "citationKey": "lin2021gait", "archiveKey": "lin2021gait", "method": "GaitGL", "title": "Gait recognition via effective global-local feature representation and local temporal aggregation", "authors": "Lin, Beibei and Zhang, Shunli and Yu, Xin", "year": 2021, "venue": "ICCV", "motivation": "Improve discriminative gait representation", "idea": "Adaptive global-local feature fusion", "techniques": "Global-local feature extraction", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "OpenGait", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/ICCV2021/html/Lin_Gait_Recognition_via_Effective_Global-Local_Feature_Representation_and_Local_Temporal_ICCV_2021_paper.html", "bibtex": "@inproceedings{lin2021gait,\n title = {Gait recognition via effective global-local feature representation and local temporal aggregation},\n author = {Lin, Beibei and Zhang, Shunli and Yu, Xin},\n year = {2021},\n booktitle = {Proceedings of the IEEE/CVF international conference on computer vision},\n pages = {14648--14656}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 47 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Strong Gait Backbones" ], "abstract": "Gait recognition is one of the most important biometric technologies and has been applied in many fields. Recent gait recognition frameworks represent each gait frame by descriptors extracted from either global appearances or local regions of humans. However, the representations based on global information often neglect the details of the gait frame, while local region based descriptors cannot capture the relations among neighboring regions, thus reducing their discriminativeness. In this paper, we propose a novel feature extraction and fusion framework to achieve discriminative feature representations for gait recognition. Towards this goal, we take advantage of both global visual information and local region details and develop a Global and Local Feature Extractor (GLFE). Specifically, our GLFE module is composed of our newly designed multiple global and local convolutional layers (GLConv) to ensemble global and local features in a principle manner. Furthermore, we present a novel operation, namely Local Temporal Aggregation (LTA), to further preserve the spatial information by reducing the temporal resolution to obtain higher spatial resolution. With the help of our GLFE and LTA, our method significantly improves the discriminativeness of our visual features, thus improving the gait recognition performance. Extensive experiments demonstrate that our proposed method outperforms state-of-theart gait recognition methods on two popular datasets.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content/ICCV2021/html/Lin_Gait_Recognition_via_Effective_Global-Local_Feature_Representation_and_Local_Temporal_ICCV_2021_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "mu2021resgait-ijcb-2021", "citationKey": "mu2021resgait", "archiveKey": "mu2021resgait", "method": "", "title": "ReSGait: The real-scene gait dataset", "authors": "Mu, Zihao and Castro, Francisco M and Marin-Jimenez, Manuel J and Guil, Nicolas and Li, Yan-Ran and Yu, Shiqi", "year": 2021, "venue": "IJCB", "motivation": "Evaluate gait recognition in reality", "idea": "Build real-world unconstrained dataset", "techniques": "Collect natural gait data", "inputs": "", "backbone": "", "loss": "", "datasets": "ReSGait", "codeAvailable": "Yes", "codeUrl": "", "url": "https://doi.org/10.1109/IJCB52358.2021.9484347", "bibtex": "@inproceedings{mu2021resgait,\n title = {ReSGait: The real-scene gait dataset},\n author = {Mu, Zihao and Castro, Francisco M and Marin-Jimenez, Manuel J and Guil, Nicolas and Li, Yan-Ran and Yu, Shiqi},\n year = {2021},\n booktitle = {2021 IEEE international joint conference on biometrics (IJCB)},\n pages = {1--8}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 57 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Many studies have shown that gait recognition can be used to identify humans at a long distance, with promising results on current datasets. However, those datasets are collected under controlled situations and predefined conditions, which limits the extrapolation of the results to unconstrained situations in which the subjects walk freely in scenes. To cover this gap, we release a novel realscene gait dataset (ReSGait), which is the first dataset collected in unconstrained scenarios with freely moving subjects and not controlled environmental parameters. Overall, our dataset is composed of 172 subjects and 870 video sequences, recorded over 15 months. Video sequences are labeled with gender, clothing, carrying conditions, taken walking route, and whether mobile phones were used or not. Therefore, the main characteristics of our dataset that differentiate it from other datasets are as follows: (i) uncontrolled real-life scenes and (ii) long recording time. Finally, we empirically assess the difficulty of the proposed dataset by evaluating state-of-the-art gait approaches for silhouette and pose modalities. The results reveal an accuracy of less than 35%, showing the inherent level of difficulty of our dataset compared to other current datasets, in which accuracies are higher than 90%. Thus, our proposed dataset establishes a new level of difficulty in the gait recognition problem, much closer to real life.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/IJCB52358.2021.9484347", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "sheng2021multi-pr-2021", "citationKey": "sheng2021multi", "archiveKey": "sheng2021multi", "method": "AT-GCN", "title": "Multi-task learning for gait-based identity recognition and emotion recognition using attention enhanced temporal graph convolutional network", "authors": "Sheng, Weijie and Li, Xinde", "year": 2021, "venue": "PR", "motivation": "Joint gait identity and emotion recognition", "idea": "Multi-task learning with skeleton features", "techniques": "Graph convolution, attention mechanism, temporal modeling", "inputs": "Skeleton", "backbone": "GCN", "loss": "CE", "datasets": "EMOGAIT, TUM GAID", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2021.107868", "bibtex": "@article{sheng2021multi,\n title = {Multi-task learning for gait-based identity recognition and emotion recognition using attention enhanced temporal graph convolutional network},\n author = {Sheng, Weijie and Li, Xinde},\n year = {2021},\n journal = {Pattern Recognition},\n volume = {114},\n pages = {107868},\n publisher = {Elsevier}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 100 }, { "source": "Author paper summaries", "sheet": "journals", "row": 7 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Background / Broader Applications" ], "abstract": "Human gait conveys significant information that can be used for identity recognition and emotion recognition. Recent studies have focused more on gait identity recognition than emotion recognition and regarded these two recognition tasks as independent and unrelated. How to train a unified model to effectively recognize the identity and emotion from gait at the same time is a novel and challenging problem. In this paper, we propose a novel Attention Enhanced Temporal Graph Convolutional Network (AT-GCN) for gait-based recognition and motion prediction. Enhanced by spatial and temporal attention, the proposed model can capture discriminative features in spatial dependency and temporal dynamics. We also present a multi-task learning architecture, which can jointly learn representations for multiple tasks. It helps the emotion recognition task with limited data considerably benefit from the identity recognition task and helps the recognition tasks benefit from the auxiliary prediction task. Furthermore, we present a new dataset (EMOGAIT) that consists of 1, 440 real gaits, annotated with identity and emotion labels. Experimental results on two datasets demonstrate the effectiveness of our approach and show that our approach achieves substantial improvements over mainstream methods for identity recognition and emotion recognition.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2021.107868", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "yao2021robust-prl-2021", "citationKey": "yao2021robust", "archiveKey": "yao2021robust", "method": "SGEI", "title": "Robust Gait Recognition Using Hybrid Descriptors Based on Skeleton Gait Energy Image", "authors": "Lingxiang Yao and Worapan Kusakunniran and Qiang Wu and Jian Zhang and Zhenmin Tang and Wankou Yang", "year": 2021, "venue": "PRL", "motivation": "Address view and clothing variations", "idea": "Hybrid gait representation with robustness", "techniques": "Skeleton–silhouette fusion; view transformation", "inputs": "Silhouette, Skeleton", "backbone": "CNN", "loss": "Similarity", "datasets": "CASIA-B", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1016/j.patrec.2019.05.012", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 113 }, { "source": "Author paper summaries", "sheet": "journals", "row": 20 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Robust-gait-recognition-using-hybrid-descriptors-base_2021_Pattern-Recogniti.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait features have been widely applied in human identification. The commonly-used representations for gait recognition can be roughly classified into two categories: model-free features and model-based features. However, due to the view variances and clothes changes, model-free features are sensitive to the appearance changes. For model-based features, there is great difficulty in extracting the underlying models from gait sequences. Based on the confidence maps and the part affinity fields produced by a twobranch multi-stage CNN network, a new model-based representation, Skeleton Gait Energy Image (SGEI), has been proposed in this paper. Another contribution is that a hybrid representation has been produced, which uses SGEI to remedy the deficiency of model-free features, Gait Energy Image (GEI) for instance. The experimental performances indicate that our proposed methods are more robust to the cloth changes, and contribute to increasing the robustness of gait recognition in the unconstrained environments with view variances and clothes changes.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patrec.2019.05.012", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "sepas2020view-tbiom-2021", "citationKey": "sepas2020view", "archiveKey": "sepas2020view", "method": "Attentive Recurrent Learning", "title": "View-Invariant Gait Recognition With Attentive Recurrent Learning of Partial Representations", "authors": "Alireza Sepas-Moghaddam and Ali Etemad", "year": 2021, "venue": "TBIOM", "motivation": "Robust gait recognition under variations", "idea": "Attention-guided recurrent partial learning", "techniques": "Convolutional maps; temporal pooling; recurrent modeling; attention weighting", "inputs": "Silhouette", "backbone": "CNN, Attention", "loss": "Triplet, Cosine Loss", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TBIOM.2020.3031470", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 121 }, { "source": "Author paper summaries", "sheet": "journals", "row": 28 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "View-Invariant_Gait_Recognition_With_Attentive_Recurrent_Learning_of_Partial_Representations.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition refers to the identification of individuals based on features acquired from their body movement during walking. Despite the recent advances in gait recognition with deep learning, variations in data acquisition and appearance, namely camera angles, subject pose, occlusions, and clothing, are challenging factors that need to be considered for achieving accurate gait recognition systems. In this article, we propose a network that first learns to extract gait convolutional energy maps (GCEM) from frame-level convolutional features. It then adopts a bidirectional recurrent neural network to learn from split bins of the GCEM, thus exploiting the relations between learned partial spatiotemporal representations. We then use an attention mechanism to selectively focus on important recurrently learned partial representations as identity information in different scenarios may lie in different GCEM bins. Our proposed model has been extensively tested on two large-scale CASIA-B and OU-MVLP gait datasets using four different test protocols and has been compared to a number of state-of-the-art and baseline solutions. Additionally, a comprehensive experiment has been performed to study the robustness of our model in the presence of six different synthesized occlusions. The experimental results show the superiority of our proposed method, outperforming the state-of-the-art, especially in scenarios where different clothing and carrying conditions are encountered. The results also revealed that our model is more robust against different occlusions as compared to the state-of-the-art methods.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TBIOM.2020.3031470", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "hou2021set-tbiom-2021", "citationKey": "hou2021set", "archiveKey": "hou2021set", "method": "Set Residual Network (SRN)", "title": "Set Residual Network for Silhouette-Based Gait Recognition", "authors": "Saihui Hou and Xu Liu and Chunshui Cao and Yongzhen Huang", "year": 2021, "venue": "TBIOM", "motivation": "Robust gait features under variations", "idea": "Parallel set–silhouette feature learning", "techniques": "Residual coordination; dual feature pyramid", "inputs": "Silhouette", "backbone": "CNN", "loss": "Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TBIOM.2021.3074963", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 120 }, { "source": "Author paper summaries", "sheet": "journals", "row": 27 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Set_Residual_Network_for_Silhouette-Based_Gait_Recognition.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Recently gait recognition receives increasing attention since it can be conducted at a long distance without the cooperation of subjects and suit for the cases of changing clothes. A key challenge is to learn gait features from the silhouettes that are invariant to the external factors such as clothing, carrying conditions and camera viewpoints. In this work, we propose a Set Residual Network for gait recognition which tries to learn more discriminative features from the silhouettes. Specifically, the silhouettes of each gait sequence are regarded as an unordered set and we propose a Set Residual Block to extract the silhouette-level and set-level features in a parallel way. Particularly, a residual connection is adopted to connect the two-level features inside each block, which enables more silhouette-set interaction and effectively coordinates the silhouette-level and set-level information for set-based feature learning from the silhouettes. Besides, we propose an efficient strategy to exploit the features from the shallow layers to learn more robust part representations for gait recognition, where the upsampling or lateral connections are unnecessary and only marginal memory cost is required. The experiments on CASIA-B and OUMVLP show that our approach can bring consistent improvement over the baselines for all walking conditions.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TBIOM.2021.3074963", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "xu2020cross-tcsvt-2021", "citationKey": "xu2020cross", "archiveKey": "xu2020cross", "method": "PSTN", "title": "Cross-View Gait Recognition Using Pairwise Spatial Transformer Networks", "authors": "Chi Xu and Yasushi Makihara and Xiang Li and Yasushi Yagi and Jianfeng Lu", "year": 2021, "venue": "TCSVT", "motivation": "Improve cross-view gait recognition", "idea": "Pairwise spatial registration for matching", "techniques": "Unified pairwise spatial transformer network", "inputs": "GEI", "backbone": "CNN", "loss": "Triplet, Contrastive", "datasets": "CASIA-B, OU-MVLP, OU-LP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TCSVT.2020.2975671", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 137 }, { "source": "Author paper summaries", "sheet": "journals", "row": 43 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Cross-View_Gait_Recognition_Using_Pairwise_Spatial_Transformer_Networks.pdf", "authorsSource": "Public scholarly metadata", "abstract": "In this paper, we propose a pairwise spatial transformer network (PSTN) for cross-view gait recognition, which reduces unwanted feature mis-alignment due to view differences before a recognition step for better performance. The proposed PSTN is a unified CNN architecture that consists of a pairwise spatial transformer (PST) and subsequent recognition network (RN). More specifically, given a matching pair of gait features from different source and target views, the PST estimates a non-rigid deformation field to register the features in the matching pair into their intermediate view, which mitigates distortion by registration compared with the case of direct deformation from the source view to target view. The registered matching pair is then fed into the RN to output a dissimilarity score. Although registration may reduce not only intra-subject variations but also inter-subject variations, we can still achieve a good trade-off between them using a loss function designed to optimize recognition accuracy. Experiments on three publicly available gait datasets demonstrate that the proposed method yields superior performance for both verification and identification scenarios by combining any gait recognition network benchmarks with the PST.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TCSVT.2020.2975671", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "chen2021multi-tip-2021", "citationKey": "chen2021multi", "archiveKey": "chen2021multi", "method": "MvGGAN", "title": "Multi-View Gait Image Generation for Cross-View Gait Recognition", "authors": "Xin Chen and Xizhao Luo and Jian Weng and Weiqi Luo and Huiting Li and Qi Tian", "year": 2021, "venue": "TIP", "motivation": "Improve cross-view gait recognition", "idea": "Augment training data via synthesis", "techniques": "Generative sample creation, multi-domain alignment, distribution matching, deep feature learning", "inputs": "Silhouette", "backbone": "CNN, GAN", "loss": "Adversarial Loss, View Loss, Cycle Loss", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TIP", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 168 }, { "source": "Author paper summaries", "sheet": "journals", "row": 70 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Multi-View_Gait_Image_Generation_for_Cross-View_Gait_Recognition.pdf", "abstract": "Gait recognition aims to recognize persons’ identities by walking styles. Gait recognition has unique advantages due to its characteristics of non-contact and long-distance compared with face and fingerprint recognition. Cross-view gait recognition is a challenge task because view variance may produce large impact on gait silhouettes. The development of deep learning has promoted cross-view gait recognition performances to a higher level. However, performances of existing deep learningbased cross-view gait recognition methods are limited by lack of gait samples under different views. In this paper, we take a Multi-view Gait Generative Adversarial Network (MvGGAN) to generate fake gait samples to extend existing gait datasets, which provides adequate gait samples for deep learning-based cross-view gait recognition methods. The proposed MvGGAN method trains a single generator for all view pairs involved in single or multiple datasets. Moreover, we perform domain alignment based on projected maximum mean discrepancy to reduce the influence of distribution divergence caused by sample generation. The experimental results on CASIA-B and OUMVLP dataset demonstrate that fake gait samples generated by the proposed MvGGAN method can improve performances of existing stateof-the-art cross-view gait recognition methods obviously on both single-dataset and cross-dataset evaluation settings.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIP", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "wang2021event-tpami-2022", "citationKey": "wang2021event", "archiveKey": "wang2021event", "method": "EV-Gait", "title": "Event-stream representation for human gaits identification using deep neural networks", "authors": "Wang, Yanxiang and Zhang, Xian and Shen, Yiran and Du, Bowen and Zhao, Guangrong and Cui, Lizhen and Wen, Hongkai", "year": 2021, "venue": "TPAMI", "motivation": "Robust gait recognition with events", "idea": "Use DVS sensors for gait", "techniques": "Event-stream representation, graph modeling, deep networks, CNN adaptation, feature distillation", "inputs": "Event-stream", "backbone": "CNN, GCN", "loss": "CE", "datasets": "CASIA-B", "codeAvailable": "No", "codeUrl": "", "url": "", "bibtex": "@article{wang2021event,\n title = {Event-stream representation for human gaits identification using deep neural networks},\n author = {Wang, Yanxiang and Zhang, Xian and Shen, Yiran and Du, Bowen and Zhao, Guangrong and Cui, Lizhen and Wen, Hongkai},\n year = {2021},\n journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},\n volume = {44},\n number = {7},\n pages = {3436--3449},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 187 }, { "source": "Author paper summaries", "sheet": "journals", "row": 87 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Ideal Gait Modalities" ], "yearNote": "Archive: 2022; bibliography: 2021", "abstract": "Dynamic vision sensors (event cameras) have recently been introduced to solve a number of different vision tasks such as object recognition, activities recognition, tracking, etc. Compared with the traditional RGB sensors, the event cameras have many unique advantages such as ultra low resources consumption, high temporal resolution and much larger dynamic range. However, these cameras only produce noisy and asynchronous events of intensity changes, i.e., event-streams rather than frames, where conventional computer vision algorithms can’t be directly applied. In our opinion the key challenge for improving the performance of event cameras in vision tasks is finding the appropriate representations of the event-streams so that cutting-edge learning approaches can be applied to fully uncover the spatio-temporal information contained in the event-streams. In this paper, we focus on the event-based human gait identification task and investigate the possible representations of the event-streams when deep neural networks are applied as the classifier. We propose new event-based gait recognition approaches basing on two different representations of the event-stream, i.e., graph and image-like representations, and use graph-based convolutional network (GCN) and convolutional neural networks (CNN) respectively to recognize gait from the event-streams. The two approaches are termed as EV-Gait-3DGraph and EV-Gait-IMG. To evaluate the performance of the proposed approaches, we collect two event-based gait datasets, one from real-world experiments and the other by converting the publicly available RGB gait recognition benchmark CASIA-B. Extensive experiments show that EV-Gait-3DGraph achieves significantly higher recognition accuracy than other competing methods when sufficient training samples are available. However, EV-Gait-IMG converges more quickly than graph-based approaches while training and shows good accuracy with only few number of training samples (less than ten). So image-like presentation is preferable when the amount of training data is limited.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-meng2020gait", "citationKey": "meng2020gait", "archiveKey": "meng2020gait", "method": "", "title": "Gait recognition for co-existing multiple people using millimeter wave sensing", "authors": "Meng, Zhen and Fu, Song and Yan, Jie and Liang, Hongyuan and Zhou, Anfu and Zhu, Shilin and Ma, Huadong and Liu, Jianhua and Yang, Ning", "year": 2020, "venue": "AAAI", "motivation": "Multiple walkers complicate nonvisual gait-based identification.", "idea": "Separate and recognize co-existing people through millimeter-wave sensing.", "techniques": "Radar sensing; multi-person gait separation; identity classification.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@inproceedings{meng2020gait,\n title = {Gait recognition for co-existing multiple people using millimeter wave sensing},\n author = {Meng, Zhen and Fu, Song and Yan, Jie and Liang, Hongyuan and Zhou, Anfu and Zhu, Shilin and Ma, Huadong and Liu, Jianhua and Yang, Ning},\n year = {2020},\n booktitle = {Proceedings of the AAAI conference on artificial intelligence},\n volume = {34},\n number = {01},\n pages = {849--856}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Task Definition" ], "abstract": "Gait recognition, i.e., recognizing persons from their walking postures, has found versatile applications in security check, health monitoring, and novel human-computer interaction. The millimeter-wave (mmWave) based gait recognition represents the most recent advance. Compared with traditional camera-based solutions, mmWave based gait recognition bears unique advantages of being still effective under non-line-of-sight scenarios, such as in black, weak light, or blockage conditions. Moreover, they are able to accomplish person identification while preserving privacy. Currently, there are only few works in mmWave gait recognition, since no public data set is available. In this paper, we build a first-of-its-kind mmWave gait data set, in which we collect gait of 95 volunteers ’seen’ from two mmWave radars in two different scenarios, which together lasts about 30 hours. Using the data set, we propose a novel deep-learning driven mmWave gait recognition method called mmGaitNet, and compare it with five state-of-the-art algorithms. We find that mmGaitNet is able to achieve 90% accuracy for single-person scenarios, 88% accuracy for five co-existing persons, while the existing methods achieve less than 66% accuracy for both scenarios.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "li2020end-accv-2020", "citationKey": "li2020end", "archiveKey": "li2020end", "method": "GaitNet", "title": "End-to-end model-based gait recognition", "authors": "Li, Xiang and Makihara, Yasushi and Xu, Chi and Yagi, Yasushi and Yu, Shiqi and Ren, Mingwu", "year": 2020, "venue": "ACCV", "motivation": "End-to-end model-based gait recognition", "idea": "Integrate 3D SMPL model", "techniques": "SMPL Recovery with Silhouette Regularization", "inputs": "RGB", "backbone": "CNN, LSTM", "loss": "Triplet, Contrastive", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "", "bibtex": "@inproceedings{li2020end,\n title = {End-to-end model-based gait recognition},\n author = {Li, Xiang and Makihara, Yasushi and Xu, Chi and Yagi, Yasushi and Yu, Shiqi and Ren, Mingwu},\n year = {2020},\n booktitle = {Proceedings of the Asian conference on computer vision}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 12 } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Most existing gait recognition approaches adopt a two-step procedure: a preprocessing step to extract silhouettes or skelet ons followed by recognition. In this paper, we propose an end-to-end mo delbased gait recognition method. Specifically, we employ a skin ned multiperson linear (SMPL) model for human modeling, and estimate it s parameters using a pre-trained human mesh recovery (HMR) network. As the pre-trained HMR is not recognition-oriented, we fine-tune it in an end-to-end gait recognition framework. To cope with differences b etween gait datasets and those used for pre-training the HMR, we introdu ce a reconstruction loss between the silhouette masks in the gait d atasets and the rendered silhouettes from the estimated SMPL model produc ed by a differentiable renderer. This enables us to adapt the HMR to th e gait dataset without supervision using the ground-truth joint l ocations. Experimental results with the OU-MVLP and CASIA-B datasets de monstrate the state-of-the-art performance of the proposed method for b oth gait identification and verification scenarios, a direct conseq uence of the explicitly disentangled pose and shape features produced by th e proposed end-to-end model-based framework.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "lin2020gait-acmmm-2020", "citationKey": "lin2020gait", "archiveKey": "lin2020gait", "method": "MT3D", "title": "Gait recognition with multiple-temporal-scale 3d convolutional neural network", "authors": "Lin, Beibei and Zhang, Shunli and Bao, Feng", "year": 2020, "venue": "ACM MM", "motivation": "Enhance temporal representation in gait", "idea": "Multi-scale temporal feature fusion", "techniques": "Dual temporal branches, low-rank convolution, frame pooling normalization", "inputs": "Silhouette", "backbone": "CNN", "loss": "Triplet", "datasets": "CASIA-B, OU-LP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1145/3394171.3413861", "bibtex": "@inproceedings{lin2020gait,\n title = {Gait recognition with multiple-temporal-scale 3d convolutional neural network},\n author = {Lin, Beibei and Zhang, Shunli and Bao, Feng},\n year = {2020},\n booktitle = {Proceedings of the 28th ACM international conference on multimedia},\n pages = {3054--3062}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 73 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait recognition which is one of the most important and effective biometric technologies has a significant advantage in long-distance recognition systems. For existing gait recognition methods, the template-based approaches may lose temporal information, while the sequence-based methods cannot fully exploit the temporal relations among the sequence. To address the above issues, we propose a novel multiple-temporal-scale gait recognition framework which integrates the temporal information in multiple temporal scales, making use of both the frame and interval fusion information. Moreover, the interval-level representation is realized by a local transformation module. Concretely, 3D convolution neural network (3D CNN) is applied in both the small and the large temporal scales to extract the spatial-temporal information. Moreover, a frame pooling method is developed to address the mismatch of the input of 3D network and video frames, and a novel 3D basic network block is designed to improve efficiency. Experiments demonstrate that the multiple-temporal-scale 3D CNN based gait recognition method can achieve better performance than most recent state-of-the-art methods in CASIA-B dataset. The proposed method obtains the rank-1 accuracy with 96.7% under normal condition, and outperforms other methods on average accuracy by at least 5.8%and 11.1%, respectively, in complex scenarios.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1145/3394171.3413861", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "fan2020gaitpart-cvpr-2020", "citationKey": "fan2020gaitpart", "archiveKey": "fan2020gaitpart", "method": "GaitPart", "title": "Gaitpart: Temporal part-based model for gait recognition", "authors": "Fan, Chao and Peng, Yunjie and Cao, Chunshui and Liu, Xu and Hou, Saihui and Chi, Jiannan and Huang, Yongzhen and Li, Qing and He, Zhiqiang", "year": 2020, "venue": "CVPR", "motivation": "Improve fine-grained description", "idea": "Learning part micro-motion features", "techniques": "Part-based spatio-temporal modeling", "inputs": "Silhouette", "backbone": "CNN", "loss": "Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "OpenGait", "codeUrl": "", "url": "https://openaccess.thecvf.com/content_CVPR_2020/html/Fan_GaitPart_Temporal_Part-Based_Model_for_Gait_Recognition_CVPR_2020_paper.html", "bibtex": "@inproceedings{fan2020gaitpart,\n title = {Gaitpart: Temporal part-based model for gait recognition},\n author = {Fan, Chao and Peng, Yunjie and Cao, Chunshui and Liu, Xu and Hou, Saihui and Chi, Jiannan and Huang, Yongzhen and Li, Qing and He, Zhiqiang},\n year = {2020},\n booktitle = {Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},\n pages = {14225--14233}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 18 }, { "source": "Manuscript Table I" }, { "source": "Verified publication metadata" } ], "summaryStatus": "Manuscript summary", "sections": [ "Background / Historical Evolution", "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Strong Gait Backbones", "Advances in Methods / Handling Real-World Noise" ], "abstract": "Gait recognition, applied to identify individual walking patterns in a long-distance, is one of the most promising video-based biometric technologies. At present, most gait recognition methods take the whole human body as a unit to establish the spatio-temporal representations. However , we have observed that different parts of human body possess evidently various visual appearances and movement patterns during walking. In the latest literature, employing partial features for human body description has been verified being beneficial to individual recognition. Taken above insights together , we assume that each part of human body needs its own spatio-temporal expression. Then, we propose a novel part-based model GaitPart and get two aspects effect of boosting the performance: On the one hand, F ocal Convolution Layer , a new applying of convolution, is presented to enhance the fine-grained learning of the partlevel spatial features. On the other hand, the Micro-motion Capture Module (MCM) is proposed and there are several parallel MCMs in the GaitPart corresponding to the predefined parts of the human body, respectively. It is worth mentioning that the MCM is a novel way of temporal modeling for gait task, which focuses on the short-range temporal features rather than the redundant long-range features for cycle gait. Experiments on two of the most popular public datasets, CASIA-B and OU-MVLP , richly exemplified that our method meets a new state-of-the-art on multiple standard benchmarks. The source code will be available on https://github.com/ChaoFan96/GaitPart.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content_CVPR_2020/html/Fan_GaitPart_Temporal_Part-Based_Model_for_Gait_Recognition_CVPR_2020_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "li2020gait-cvpr-2020", "citationKey": "li2020gait", "archiveKey": "li2020gait", "method": "ICDNet", "title": "Gait Recognition via Semi-supervised Disentangled Representation Learning to Identity and Covariate Features", "authors": "Li, Xiang and Makihara, Yasushi and Xu, Chi and Yagi, Yasushi and Ren, Mingwu", "year": 2020, "venue": "CVPR", "motivation": "Improve robustness to covariates", "idea": "Disentangled ID-covariate learning", "techniques": "GEI-based dual reconstruction", "inputs": "Silhouette", "backbone": "CNN", "loss": "Triplet, Similarity", "datasets": "OU-LP-Bag, CASIA-B", "codeAvailable": "No", "codeUrl": "", "url": "https://openaccess.thecvf.com/content_CVPR_2020/html/Li_Gait_Recognition_via_Semi-supervised_Disentangled_Representation_Learning_to_Identity_and_CVPR_2020_paper.html", "bibtex": "@inproceedings{li2020gait,\n title = {Gait Recognition via Semi-supervised Disentangled Representation Learning to Identity and Covariate Features},\n author = {Li, Xiang and Makihara, Yasushi and Xu, Chi and Yagi, Yasushi and Ren, Mingwu},\n year = {2020},\n booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},\n pages = {13309--13319},\n url = {https://openaccess.thecvf.com/content_CVPR_2020/html/Li_Gait_Recognition_via_Semi-supervised_Disentangled_Representation_Learning_to_Identity_and_CVPR_2020_paper.html}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 19 }, { "source": "Manuscript Table I" } ], "summaryStatus": "Manuscript summary", "sections": [ "Background / Historical Evolution" ], "abstract": "Existing gait recognition approaches typically focus on learning identity features that are invariant to covariates (e.g., the carrying status, clothing, walking speed, and viewing angle) and seldom involve learning features from the covariate aspect, which may lead to failure modes when variations due to the covariate overwhelm those due to the identity. We therefore propose a method of gait recognition via disentangled representation learning that considers both identity and covariate features. Specifically, we first encode an input gait template to get the disentangled identity and covariate features, and then decode the features to simultaneously reconstruct the input gait template and the canonical version of the same subject with no covariates in a semi-supervised manner to ensure successful disentanglement. We finally feed the disentangled identity features into a contrastive/triplet loss function for a verification/identification task. Moreover , we find that new gait templates can be synthesized by transferring the covariate feature from one subject to another . Experimental results on three publicly available gait data sets demonstrate the effectiveness of the proposed method compared with other state-of-the-art methods.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://openaccess.thecvf.com/content_CVPR_2020/html/Li_Gait_Recognition_via_Semi-supervised_Disentangled_Representation_Learning_to_Identity_and_CVPR_2020_paper.html", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "hou2020gait-eccv-2020", "citationKey": "hou2020gait", "archiveKey": "hou2020gait", "method": "GLN", "title": "Gait lateral network: Learning discriminative and compact representations for gait recognition", "authors": "Hou, Saihui and Cao, Chunshui and Liu, Xu and Huang, Yongzhen", "year": 2020, "venue": "ECCV", "motivation": "Reduce dimensionality without accuracy loss", "idea": "Lateral fusion with Compact Block", "techniques": "Compact, pyramid-based silhouette modeling", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "OpenGait", "codeUrl": "", "url": "", "bibtex": "@inproceedings{hou2020gait,\n title = {Gait lateral network: Learning discriminative and compact representations for gait recognition},\n author = {Hou, Saihui and Cao, Chunshui and Liu, Xu and Huang, Yongzhen},\n year = {2020},\n booktitle = {European conference on computer vision},\n pages = {382--398}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 35 } ], "summaryStatus": "Author collection", "sections": [], "abstract": "Gait recognition aims at identifying different people by the walking patterns, which can be conducted at a long distance without the cooperation of subjects. A key challenge for gait recognition is to learn representations from the silhouettes that are invariant to the factors such as clothing, carrying conditions and camera viewpoints. Besides being discriminative for identification, the gait representations should also be compact for storage to keep millions of subjects registered in the gallery. In this work, we propose a novel network named Gait Lateral Network (GLN) which can learn both discriminative and compact representations from the silhouettes for gait recognition. Specifically, GLN leverages the inherent feature pyramid in deep convolutional neural networks to enhance the gait representations. The silhouette-level and set-level features extracted by different stages are merged with the lateral connections in a top-down manner. Besides, GLN is equipped with a Compact Block which can significantly reduce the dimension of the gait representations without hindering the accuracy. Extensive experiments on CASIA-B and OUMVLP show that GLN can achieve state-of-the-art performance using the 256-dimensional representations. Under the most challenging condition of walking in different clothes on CASIA-B, our method improves the rank-1 accuracy by 6 .45%.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "xu2020gait-eccv-2020", "citationKey": "xu2020gait", "archiveKey": "xu2020gait", "method": "PA-GCRNet", "title": "Gait recognition from a single image using a phase-aware gait cycle reconstruction network", "authors": "Xu, Chi and Makihara, Yasushi and Li, Xiang and Yagi, Yasushi and Lu, Jianfeng", "year": 2020, "venue": "ECCV", "motivation": "Enable real-time gait recognition", "idea": "Phase-aware sequence regeneration", "techniques": "Single-image gait reconstruction network", "inputs": "Silhouette", "backbone": "CNN", "loss": "Triplet, Similarity", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "", "bibtex": "@inproceedings{xu2020gait,\n title = {Gait recognition from a single image using a phase-aware gait cycle reconstruction network},\n author = {Xu, Chi and Makihara, Yasushi and Li, Xiang and Yagi, Yasushi and Lu, Jianfeng},\n year = {2020},\n booktitle = {European Conference on Computer Vision},\n pages = {386--403}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 36 } ], "summaryStatus": "Author collection", "sections": [], "abstract": "We propose a method of gait recognition just from a single image for the first time, which enables latency-free gait recognition. To mitigate large intra-subject variations caused by a phase (gait pose) difference between a matching pair of input single images, we first reconstruct full gait cycles of image sequences from the single images using an auto-encoder framework, and then feed them into a state-of-the-art gait recognition network for matching. Specifically, a phase estimation network is introduced for the input single image, and the gait cycle reconstruction network exploits the estimated phase to mitigate the dependence of an encoded feature on the phase of that single image. This is called phase-aware gait cycle reconstructor (PA-GCR). In the training phase, the PA-GCR and recognition network are simultaneously optimized to achieve a good trade-off between reconstruction and recognition accuracies. Experiments on three gait datasets demonstrate the significant performance improvement of this method.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "li2020gait-pr-2020", "citationKey": "li2020gait", "archiveKey": "li2020gait", "method": "Ab-GAN", "title": "Gait Recognition Invariant to Carried Objects Using Alpha Blending Generative Adversarial Networks", "authors": "Xiang Li and Yasushi Makihara and Chi Xu and Yasushi Yagi and Mingwu Ren", "year": 2020, "venue": "PR", "motivation": "Gait recognition robust to carried objects", "idea": "Remove carried objects before recognition", "techniques": "Alpha-blending GAN, template generation, discriminative learning", "inputs": "Silhouette", "backbone": "CNN", "loss": "CE, Triplet", "datasets": "CASIA-B, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2020.107376", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 99 }, { "source": "Author paper summaries", "sheet": "journals", "row": 6 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Background / Historical Evolution" ], "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition invariant to carried objects (COs) is very difficult in a real-life scene because the COs can have various shapes and sizes, in addition to unpredictable carrying locations (e.g., front, back, and side, or multiple locations). Therefore, in this paper, we propose a robust method for gait recognition against various COs by reconstructing a gait template without COs. A straightforward approach is to directly generate a gait template without COs given a gait template with COs as the input using a conventional generative adversarial network. There is, however, a potential risk of unnecessarily altering parts that were originally unaffected by COs (e.g., leg parts for a person carrying a backpack). Because we do not want to touch such unaffected parts in the original template, we first estimate a gait template without COs, and then blend it with the original template by an estimated alpha matte that indicates the blending parameters. We then create an alpha-blended template from the original template and the generated template without COs based on the estimated alpha matte. We use two independent generators to estimate the alpha matte and the generated template without COs. Finally, we feed the alpha-blended gait template into a state-of-the-art discrimination network for gait recognition. The experimental results on three publicly available gait databases with real-life COs demonstrate the state-of-the-art performance of the proposed method.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2020.107376", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "liao2020model-pr-2020", "citationKey": "liao2020model", "archiveKey": "liao2020model", "method": "PoseGait", "title": "A model-based gait recognition method with body pose and human prior knowledge", "authors": "Liao, Rijun and Yu, Shiqi and An, Weizhi and Huang, Yongzhen", "year": 2020, "venue": "PR", "motivation": "Robust gait recognition under variations", "idea": "Use human pose as feature", "techniques": "3D pose extraction, low-dimensional features, loss fusion", "inputs": "Skeleton", "backbone": "CNN", "loss": "CE, Center Loss", "datasets": "CASIA-B, CASIA-E", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2019.107069", "bibtex": "@article{liao2020model,\n title = {A model-based gait recognition method with body pose and human prior knowledge},\n author = {Liao, Rijun and Yu, Shiqi and An, Weizhi and Huang, Yongzhen},\n year = {2020},\n journal = {Pattern Recognition},\n volume = {98},\n pages = {107069},\n publisher = {Elsevier}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 98 }, { "source": "Author paper summaries", "sheet": "journals", "row": 5 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [ "Background / Task Definition" ], "abstract": "We propose in this paper a novel model-based gait recognition method, PoseGait . Gait recognition is a challenging and attractive task in biometrics. Early approaches to gait recognition were mainly appearance-based. The appearance-based features are usually extracted from human body silhouettes, which are easy to compute and have shown to be efficient for recognition tasks. Nevertheless silhouettes shape is not invariant to changes in clothing, and can be subject to drastic variations, due to illumination changes or other external factors. An alternative to silhouette-based features are model-based features. However, they are very challenging to acquire especially for low image resolution. In contrast to previous approaches, our model PoseGait exploits human 3D pose estimated from images by Convolutional Neural Network as the input feature for gait recognition. The 3D pose, defined by the 3D coordinates of joints of the human body, is invariant to view changes and other external factors of variation. We design spatio-temporal features from the 3D pose to improve the recognition rate. Our method is evaluated on two large datasets, CASIA B and CASIA E. The experimental results show that the proposed method can achieve state-of-the-art performance and is robust to view and clothing variations.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2019.107069", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "an2020performance-tbiom-2020", "citationKey": "an2020performance", "archiveKey": "an2020performance", "method": "OU-MVLP-Pose", "title": "Performance evaluation of model-based gait on multi-view very large population database with pose sequences", "authors": "An, Weizhi and Yu, Shiqi and Makihara, Yasushi and Wu, Xinhui and Xu, Chi and Yu, Yang and Liao, Rijun and Yagi, Yasushi", "year": 2020, "venue": "TBIOM", "motivation": "Advance model-based gait recognition feasibility", "idea": "Large-scale pose sequence database", "techniques": "Deep pose estimation; multi-view dataset construction; model-based benchmarking", "inputs": "Skeleton", "backbone": "CNN, LSTM", "loss": "CE, Center Loss", "datasets": "OU-MVLP", "codeAvailable": "Yes", "codeUrl": "", "url": "", "bibtex": "@article{an2020performance,\n title = {Performance evaluation of model-based gait on multi-view very large population database with pose sequences},\n author = {An, Weizhi and Yu, Shiqi and Makihara, Yasushi and Wu, Xinhui and Xu, Chi and Yu, Yang and Liao, Rijun and Yagi, Yasushi},\n year = {2020},\n journal = {IEEE transactions on biometrics, behavior, and identity science},\n volume = {2},\n number = {4},\n pages = {421--430},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 119 }, { "source": "Author paper summaries", "sheet": "journals", "row": 26 } ], "summaryStatus": "Author collection", "sections": [ "Background / Broader Applications" ], "abstract": "Model-based gait recognition is considered to be promising due to the robustness against some variations, such as clothing and baggage carried. Although model-based gait recognition has not been fully explored due to the difficulty of human body model fitting and the lack of a large-scale gait database, recent progress in deep learning-based approaches to human body model fitting and human pose estimation is mitigating the difficulty. In this paper, we, therefore, address the remaining issue by presenting a large-scale human pose-based gait database, OUMVLP-Pose, which is based on a publicly available multiview large-scale gait database, OUMVLP. OUMVLP-Pose has many unique advantages compared with other public databases. First, OUMVLP-Pose is the first gait database that provides two datasets of human pose sequences extracted by two standard deep learning-based pose estimation algorithms, OpenPose and AlphaPose. Second, it contains multi-view large-scale data, i.e., over 10,000 subjects and 14 views for each subject. In addition, we also provide benchmarks in which different kinds of gait recognition methods, including model-based methods and appearance-based methods, have been evaluated comprehensively. The model-based gait recognition methods have shown promising performances. We believe this database, OUMVLP- Pose, will greatly promote model-based gait recognition in the next few years.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ben2019coupled-tcsvt-2020", "citationKey": "ben2019coupled", "archiveKey": "ben2019coupled", "method": "CBDP", "title": "Coupled Bilinear Discriminant Projection for Cross-View Gait Recognition", "authors": "Xianye Ben and Chen Gong and Peng Zhang and Rui Yan and Qiang Wu and Weixiao Meng", "year": 2020, "venue": "TCSVT", "motivation": "Accurate cross-view gait recognition", "idea": "Align gait images preserving structure", "techniques": "Coupled bilinear discriminant projection", "inputs": "GEI", "backbone": "", "loss": "", "datasets": "CASIA-B, OU-LP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TCSVT.2019.2893736", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 136 }, { "source": "Author paper summaries", "sheet": "journals", "row": 42 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Coupled_Bilinear_Discriminant_Projection_for_Cross-View_Gait_Recognition.pdf", "abstract": "A problem that hinders good performance of general gait recognition systems is that the appearance features of gaits are more affected-prone by views than identities, especially when the walking direction of the probe gait is different from the register gait. This problem cannot be solved by traditional projection learning methods because these methods can learn only one projection matrix, and thus for the same subject, it cannot transfer cross-view gait features into similar ones. This paper presents an innovative method to overcome this problem by aligning gait energy images (GEIs) across views with the coupled bilinear discriminant projection (CBDP). Specifically, the CBDP generates the aligned gait matrix features for two views with two sets of bilinear transformation matrices, so that the original GEIs’ spatial structure information can be preserved. By iteratively maximizing the ratio of inter-class distance metric to intra-class distance metric, the CBDP can learn the optimal matrix subspace where the GEIs across views are aligned in both horizontal and vertical coordinates. Therefore, the CBDP is also able to avoid the under-sample problem. We also theoretically prove that the upper and lower bounds of the objective function sequence of the CBDP are both monotonically increasing, so the convergence of the CBDP is demonstrated. In the terms of accuracy, the comparative experiments on the CASIA (B) and OU-ISIR gait databases show that our method is superior to the state-of-the-art cross-view gait recognition methods. More impressively, encouraging performance is obtained by our method even in matching a lateral-view gait with a frontal-view gait.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TCSVT.2019.2893736", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "zou2020deep-tifs-2020", "citationKey": "zou2020deep", "archiveKey": "zou2020deep", "method": "CNN+LSTM", "title": "Deep Learning-Based Gait Recognition Using Smartphones in the Wild", "authors": "Qin Zou and Yanling Wang and Qian Wang and Yi Zhao and Qingquan Li", "year": 2020, "venue": "TIFS", "motivation": "Unobtrusive robust person identification", "idea": "Smartphone-based inertial gait recognition", "techniques": "Time-series convolution; recurrent sequence modeling; hybrid deep learning integration", "inputs": "IMU", "backbone": "CNN, LSTM", "loss": "CE", "datasets": "whuGAIT", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TIFS.2020.2985628", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "journals", "row": 56 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "Deep_Learning-Based_Gait_Recognition_Using_Smartphones_in_the_Wild.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Compared to other biometrics, gait is difficult to conceal and has the advantage of being unobtrusive. Inertial sensors, such as accelerometers and gyroscopes, are often used to capture gait dynamics. These inertial sensors are commonly integrated into smartphones and are widely used by the average person, which makes gait data convenient and inexpensive to collect. In this paper, we study gait recognition using smartphones in the wild. In contrast to tradi tional methods, which often require a person to walk along a specified road and/or at a normal walking speed, the proposed method collects inertial gait data under unconstrained conditions without knowing when, where, and how the user walks. To obtain good person identification and authentication performance, deep-learning techniques are presented to learn and model the gait biometrics based on walking data. Specifically, a hybrid deep neural network is proposed for robust gait feature representation, where features in the space and time domains are successively abstracted by a convolutional neural network and a recurrent neural network. In the experiments, two datasets collected by smartphones for a total of 118 subjects are used for evaluations. The experiments show that the proposed method achieves higher than 93.5% and 93.7% accuracies in person identification and authentication, respectively.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIFS.2020.2985628", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "limcharoen2020view-tifs-2020", "citationKey": "limcharoen2020view", "archiveKey": "limcharoen2020view", "method": "JRC-CNN", "title": "View-Independent Gait Recognition Using Joint Replacement Coordinates JRCs and Convolutional Neural Network", "authors": "Piya Limcharoen and Nirattaya Khamsemanan and Cholwich Nattee", "year": 2020, "venue": "TIFS", "motivation": "Robust view-independent gait recognition", "idea": "Local joint movement representation", "techniques": "Frame-level joint encoding; convolutional feature learning; body-wide relation modeling", "inputs": "Skeleton", "backbone": "CNN", "loss": "CE", "datasets": "SIIT-CN-A, SIIT-CN-B", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TIFS.2020.2985535", "bibtex": "", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 153 }, { "source": "Author paper summaries", "sheet": "journals", "row": 57 }, { "source": "Verified publication metadata" } ], "summaryStatus": "Author collection", "sections": [], "metadataSource": "View-Independent_Gait_Recognition_Using_Joint_Replacement_Coordinates_JRCs_and_Convolutional_Neural_Network.pdf", "authorsSource": "Public scholarly metadata", "abstract": "Gait recognition has received increasing attention for security and authentication since it can be done unintrusively from afar and without a subject’s awareness. In this work, we propose a new model-based gait recognition technique called JRC-CNN gait recognition. We introduce three new concepts. (1) We create a new way to preprocess skeleton data by rotating skeleton data using two virtual axes. This process reduces the fluctuation in movements and resolves the multi-viewpoint issue. All postures in a walk are observed from the same angle. (2) We introduce new Joint Replacement Coordinates (JRCs), which represent the movements of the left and right joints in a group of three connected joints. These JRC gait features are designed to put more emphasis on local movements than the movements of non-connected joints. (3) We construct a new Convolution Neural Network (CNN) for the classification process, which consists of a convolutional layer on each JRC and two fully-connected layers. A convolutional layer is designed to discover relations within a group of three connected joints. Fully-connected layers also find the relations of all groups of three connected joints throughout an entire body (in a posture). Our JRC-CNN technique achieves above 98.4% accuracy and significantly outperforms other existing techniques for all freedirection walk datasets. It also performs well under the gallerysize test and the CMC curve test. This means that our proposed JRC-CNN gait recognition technique can be used in a real-world situation. Experimental results also suggest that a person can be identified by a unique posture (an entire body is observed as a whole) with the focus on the move ments of connected joints.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIFS.2020.2985535", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "wu2020condition-tip-2021", "citationKey": "wu2020condition", "archiveKey": "wu2020condition", "method": "Condition-Aware", "title": "Condition-aware comparison scheme for gait recognition", "authors": "Wu, Haoqian and Tian, Jian and Fu, Yongjian and Li, Bin and Li, Xi", "year": 2020, "venue": "TIP", "motivation": "Robust cross-condition gait recognition", "idea": "Adaptively compare discriminative body parts", "techniques": "Condition-aware weighting, geometry augmentation, temporal feature aggregation", "inputs": "Silhouette", "backbone": "CNN", "loss": "Triplet", "datasets": "CASIA-B, OU-LP", "codeAvailable": "No", "codeUrl": "", "url": "", "bibtex": "@article{wu2020condition,\n title = {Condition-aware comparison scheme for gait recognition},\n author = {Wu, Haoqian and Tian, Jian and Fu, Yongjian and Li, Bin and Li, Xi},\n year = {2020},\n journal = {IEEE Transactions on image processing},\n volume = {30},\n pages = {2734--2744},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 167 }, { "source": "Author paper summaries", "sheet": "journals", "row": 69 } ], "summaryStatus": "Author collection", "sections": [ "Advances in Methods / Gait Metric Designs" ], "yearNote": "Archive: 2021; bibliography: 2020", "abstract": "As an important and challenging problem, gait recognition has gained considerable attention. It suffers from confounding conditions, that is, it is sensitive to camera views, dressing types and so on. Interestingly, it is observed that, under different conditions, local body parts contribute differently to recognition performance. In this paper, we propose a condition-aware comparison scheme to measure gait pairs’ similarity via a novel module named Instructor. Also, we present a geometry-guided data augmentation approach (Dresser) to enrich dressing conditions. Furthermore, to enhance the gait representation, we propose to model temporal local information from coarse to fine. Our model is evaluated on two popular benchmarks, CASIA-B and OULP. Results show that our method outperforms current s tate-of-the-art methods, especially in the cross-condition scenario.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-wang2020deep", "citationKey": "wang2020deep", "archiveKey": "wang2020deep", "method": "", "title": "Deep high-resolution representation learning for visual recognition", "authors": "Wang, Jingdong and Sun, Ke and Cheng, Tianheng and Jiang, Borui and Deng, Chaorui and Zhao, Yang and Liu, Dong and Mu, Yadong and Tan, Mingkui and Wang, Xinggang and others", "year": 2020, "venue": "TPAMI", "motivation": "Repeated spatial downsampling loses detail needed for human pose estimation.", "idea": "Maintain high-resolution representations throughout the network.", "techniques": "Parallel multi-resolution branches; repeated feature exchange; HRNet.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/tpami.2020.2983686", "bibtex": "@article{wang2020deep,\n title = {Deep high-resolution representation learning for visual recognition},\n author = {Wang, Jingdong and Sun, Ke and Cheng, Tianheng and Jiang, Borui and Deng, Chaorui and Zhao, Yang and Liu, Dong and Mu, Yadong and Tan, Mingkui and Wang, Xinggang and others},\n year = {2020},\n journal = {IEEE transactions on pattern analysis and machine intelligence},\n volume = {43},\n number = {10},\n pages = {3349--3364},\n publisher = {IEEE}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Task Definition" ], "abstract": "HRNet retains a high-resolution feature stream throughout the network while exchanging information with parallel lower-resolution streams. Repeated fusion combines spatial precision with broader semantic context. Evaluations on pose estimation, segmentation and detection demonstrate the versatility of this representation, providing a general visual backbone for tasks that require accurate spatial localization.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/1908.07919", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "zhang2020learning-tpami-2022", "citationKey": "zhang2020learning", "archiveKey": "zhang2020learning", "method": "GaitNet", "title": "On learning disentangled representations for gait recognition", "authors": "Zhang, Ziyuan and Tran, Luan and Liu, Feng and Liu, Xiaoming", "year": 2020, "venue": "TPAMI", "motivation": "Robust gait recognition from RGB", "idea": "Disentangle dynamic and static features", "techniques": "Autoencoder CNN, multi-loss training, LSTM", "inputs": "RGB", "backbone": "CNN, LSTM", "loss": "CE, Triplet", "datasets": "CASIA-B, USF, FVG", "codeAvailable": "Yes", "codeUrl": "", "url": "", "bibtex": "@article{zhang2020learning,\n title = {On learning disentangled representations for gait recognition},\n author = {Zhang, Ziyuan and Tran, Luan and Liu, Feng and Liu, Xiaoming},\n year = {2020},\n journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},\n volume = {44},\n number = {1},\n pages = {345--360},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 189 }, { "source": "Author paper summaries", "sheet": "journals", "row": 89 } ], "summaryStatus": "Author collection", "sections": [ "Background / Broader Applications", "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Strong Gait Backbones", "Advances in Methods / Gait Learning Frameworks" ], "yearNote": "Archive: 2022; bibliography: 2020", "abstract": "Gait, the walking pattern of individuals, is one of the important biometrics modalities. Most of the existing gait recognition methods take silhouettes or articulated body models as gait features. These methods suffer from degraded recognition performance when handling confounding variables, such as clothing, carrying and viewing angle. To remedy this issue, we propose a novel AutoEncoder framework, GaitNet, to explicitly disentangle appearance, canonical and pose features from RGB imagery. The LSTM integrates pose features over time as a dynamic gait feature while canonical features are averaged as a static gait feature. Both of them are utilized as classification features. In addition, we collect a Frontal-View Gait (FVG) dataset to focus on gait recognition from frontal-view walking, which is a challenging problem since it contains minimal gait cues compared to other views. FVG also includes other important variations, e.g., walking speed, carrying, and clothing. With extensive experiments on CASIA-B, USF, and FVG datasets, our method demonstrates superior performance to the SOTA quantitatively, the ability of feature disentanglement qualitatively, and promising computational efficiency. We further compare our GaitNet with state-of-the-art face recognition to demonstrate the advantages of gait biometrics identification under certain scenarios, e.g., long-distance/lower resolutions, cross viewing angles. Source code is available at http://cvlab.cse.msu.edu/project-gaitnet.html", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-chao2019gaitset", "citationKey": "chao2019gaitset", "archiveKey": "chao2019gaitset", "method": "", "title": "Gaitset: Regarding gait as a set for cross-view gait recognition", "authors": "Chao, Hanqing and He, Yiwei and Zhang, Junping and Feng, Jianfeng", "year": 2019, "venue": "AAAI", "motivation": "Rigid sequence ordering and averaged templates constrain gait representation.", "idea": "Treat a walking sequence as an unordered set of silhouettes.", "techniques": "Frame-level convolution; set pooling; horizontal pyramid mapping.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1609/aaai.v33i01.33018126", "bibtex": "@inproceedings{chao2019gaitset,\n title = {Gaitset: Regarding gait as a set for cross-view gait recognition},\n author = {Chao, Hanqing and He, Yiwei and Zhang, Junping and Feng, Jianfeng},\n year = {2019},\n booktitle = {Proceedings of the AAAI conference on artificial intelligence},\n volume = {33},\n number = {01},\n pages = {8126--8133}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution", "Advances in Methods / Ideal Gait Modalities", "Advances in Methods / Strong Gait Backbones" ], "abstract": "GaitSet learns identity from an unordered set of silhouette frames. Its aggregation avoids fixed sequence-length and ordering requirements and can combine observations from different recording conditions. CASIA-B and OU-MVLP evaluations examine cross-view recognition, clothing and carrying changes, and recognition from short frame sets.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://doi.org/10.1609/aaai.v33i01.33018126", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-marsico2019survey", "citationKey": "marsico2019survey", "archiveKey": "marsico2019survey", "method": "", "title": "A survey on gait recognition via wearable sensors", "authors": "Marsico, Maria De and Mecca, Alessio", "year": 2019, "venue": "CSUR", "motivation": "Wearable gait recognition spans heterogeneous sensors and processing pipelines.", "idea": "Review wearable sensing approaches and their evaluation requirements.", "techniques": "Sensor-based taxonomy; feature extraction review; biometric protocol comparison.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@article{marsico2019survey,\n title = {A survey on gait recognition via wearable sensors},\n author = {Marsico, Maria De and Mecca, Alessio},\n year = {2019},\n journal = {ACM Computing Surveys (CSUR)},\n volume = {52},\n number = {4},\n pages = {1--39},\n publisher = {ACM New York, NY, USA}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Introduction" ], "abstract": "This survey examines gait recognition from wearable and mobile sensors. It organizes the acquisition and processing of walking signals, feature design, matching approaches and available evaluations. Alongside the practical advantages of sensor-based recognition, it discusses limitations associated with sensing conditions and deployment, positioning wearable gait biometrics relative to vision-based approaches.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://research.uniroma1.it/node/39843", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-wang2019ev", "citationKey": "wang2019ev", "archiveKey": "wang2019ev", "method": "", "title": "EV-gait: Event-based robust gait recognition using dynamic vision sensors", "authors": "Wang, Yanxiang and Du, Bowen and Shen, Yiran and Wu, Kai and Zhao, Guangrong and Sun, Jianguo and Wen, Hongkai", "year": 2019, "venue": "CVPR", "motivation": "Frame-based gait capture is sensitive to illumination and motion degradation.", "idea": "Recognize gait from asynchronous brightness-change events.", "techniques": "Dynamic vision sensors; event motion representations; robust feature learning.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@inproceedings{wang2019ev,\n title = {EV-gait: Event-based robust gait recognition using dynamic vision sensors},\n author = {Wang, Yanxiang and Du, Bowen and Shen, Yiran and Wu, Kai and Zhao, Guangrong and Sun, Jianguo and Wen, Hongkai},\n year = {2019},\n booktitle = {Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},\n pages = {6358--6367}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Task Definition", "Advances in Methods / Ideal Gait Modalities" ], "abstract": "EV-Gait explores identity recognition using asynchronous events from dynamic vision sensors. Motion consistency helps suppress noisy events before a neural model learns gait representations. Experiments use both captured events and a synthetic event version of CASIA-B, examining whether event-based sensing can retain useful walking patterns under challenging visual conditions.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://openaccess.thecvf.com/content_CVPR_2019/html/Wang_EV-Gait_Event-Based_Robust_Gait_Recognition_Using_Dynamic_Vision_Sensors_CVPR_2019_paper.html", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-zhang2019learning", "citationKey": "zhang2019learning", "archiveKey": "zhang2019learning", "method": "", "title": "Learning joint gait representation via quintuplet loss minimization", "authors": "Zhang, Kaihao and Luo, Wenhan and Ma, Lin and Liu, Wei and Li, Hongdong", "year": 2019, "venue": "CVPR", "motivation": "Gait embeddings must balance within-identity compactness and between-identity separation.", "idea": "Jointly optimize relations among several gait samples.", "techniques": "Quintuplet loss; joint representation learning; metric optimization.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@inproceedings{zhang2019learning,\n title = {Learning joint gait representation via quintuplet loss minimization},\n author = {Zhang, Kaihao and Luo, Wenhan and Ma, Lin and Liu, Wei and Li, Hongdong},\n year = {2019},\n booktitle = {Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},\n pages = {4700--4709}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "This method jointly learns representations from individual gait samples and paired observations. Its quintuplet objective coordinates these complementary representations to improve discrimination between people while accommodating variation within an identity. The study evaluates the resulting joint representation for gait recognition, emphasizing the relationship between feature learning and the structure of metric supervision.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://openaccess.thecvf.com/content_CVPR_2019/html/Zhang_Learning_Joint_Gait_Representation_via_Quintuplet_Loss_Minimization_CVPR_2019_paper.html", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-song2019gaitnet", "citationKey": "song2019gaitnet", "archiveKey": "song2019gaitnet", "method": "", "title": "Gaitnet: An end-to-end network for gait based human identification", "authors": "Song, Chunfeng and Huang, Yongzhen and Huang, Yan and Jia, Ning and Wang, Liang", "year": 2019, "venue": "PR", "motivation": "Appearance and walking dynamics entangle identity with changing conditions.", "idea": "Learn gait identity within an end-to-end video recognition network.", "techniques": "Disentangled appearance and pose features; temporal aggregation; identity supervision.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2019.106988", "bibtex": "@article{song2019gaitnet,\n title = {Gaitnet: An end-to-end network for gait based human identification},\n author = {Song, Chunfeng and Huang, Yongzhen and Huang, Yan and Jia, Ning and Wang, Liang},\n year = {2019},\n journal = {Pattern recognition},\n volume = {96},\n pages = {106988},\n publisher = {Elsevier}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Broader Applications" ], "abstract": "Gait recognition is one of the most important techniques for human identification at a distance. Most current gait recognition frameworks consist of several separate steps: silhouette segmentation, feature extraction, feature learning, and similarity measurement. These modules are mutually independent with each part fixed, resulting in a suboptimal performance in challenging conditions. In this paper, we integrate those steps into one framework, i.e., an end-to-end network for gait recognition, named GaitNet . It is composed of two convolutional neural networks: one corresponds to gait segmentation, and the other corresponds to classification. The two networks are modeled in one joint learning procedure which can be trained jointly. This strategy greatly simplifies the traditional step-by-step manner and is thus much more efficient for practical applications. Moreover, joint learning can automatically adjust each part to fit the global optimal objective, leading to obvious performance improvement over separate learning. We evaluate our method on three large scale gait datasets, including CASIA-B, SZU RGB-D Gait and a newly built database with complex dynamic outdoor backgrounds. Extensive experimental results show that the proposed method is effective and achieves the state-of-the-art results. The code and data will be released upon request.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2019.106988", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "zhang2019cross-tip-2020", "citationKey": "zhang2019cross", "archiveKey": "zhang2019cross", "method": "Angle Center Loss (ACL)", "title": "Cross-view gait recognition by discriminative feature learning", "authors": "Zhang, Yuqi and Huang, Yongzhen and Yu, Shiqi and Wang, Liang", "year": 2020, "venue": "TIP", "motivation": "Accurate cross-view gait recognition", "idea": "Learn discriminative spatial-temporal features", "techniques": "Angle-based loss, local feature partitioning, temporal attention, part-focused feature learning", "inputs": "Silhouette", "backbone": "CNN, LSTM, Attention", "loss": "Center Loss, ACL", "datasets": "CASIA-B, OU-LP, OU-MVLP", "codeAvailable": "No", "codeUrl": "", "url": "https://doi.org/10.1109/TIP.2019.2926208", "bibtex": "@article{zhang2019cross,\n title = {Cross-view gait recognition by discriminative feature learning},\n author = {Zhang, Yuqi and Huang, Yongzhen and Yu, Shiqi and Wang, Liang},\n year = {2019},\n journal = {IEEE Transactions on Image Processing},\n volume = {29},\n pages = {1001--1015},\n publisher = {IEEE}\n}", "scope": "Survey collection", "provenance": [ { "source": "Author paper summaries", "sheet": "main-table", "row": 166 }, { "source": "Author paper summaries", "sheet": "journals", "row": 68 } ], "summaryStatus": "Author collection", "sections": [ "Background / Historical Evolution" ], "yearNote": "Statistics use final publication year: IEEE TIP volume 29 (2020), pp. 1001–1015. Online-first publication was 2019.", "abstract": "Recently, deep learning-based cross-view gait recognition has become popular owing to the strong capacity of convolutional neural networks (CNNs). Current deep learning methods often rely on loss functions used widely in the task of face recognition, e.g., contrastive loss and triplet loss. These loss functions have the problem of hard negative mining. In this paper, a robust, effective, and gait-related loss function, called angle center loss (ACL), is proposed to learn discriminative gait features. The proposed loss function is robust to different local parts and temporal window sizes. Different from center loss which learns a center for each identity, the proposed loss function learns multiple sub-centers for each angle of the same identity. Only the largest distance between the anchor feature and the corresponding cross-view sub-centers is penalized, which achieves better intra-subject compactness. We also propose to extract discriminative spatial–temporal features by local feature extractors and a temporal attention model. A simplified spatial transformer network is proposed to localize the suitable horizontal parts of the human body. Local gait features for each horizontal part are extracted and then concatenated as the descriptor. We introduce long short-term memory (LSTM) units as the temporal attention model to learn the attention score for each frame, e.g., focusing more on discriminative frames and less on frames with bad quality. The temporal attention model shows better performance than the temporal average pooling or gait energy images (GEI). By combing the three aspects, we ac hieve state-of-the-art results on several cross-view gait recognition benchmarks.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": true, "venueInScope": true }, { "id": "ref-cao2019openpose", "citationKey": "cao2019openpose", "archiveKey": "cao2019openpose", "method": "", "title": "Openpose: Realtime multi-person 2d pose estimation using part affinity fields", "authors": "Cao, Zhe and Hidalgo, Gines and Simon, Tomas and Wei, Shih-En and Sheikh, Yaser", "year": 2019, "venue": "TPAMI", "motivation": "Multi-person pose estimation requires efficient joint detection and association.", "idea": "Predict body joints and their pairwise associations jointly.", "techniques": "Part affinity fields; bottom-up pose estimation; convolutional prediction.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@article{cao2019openpose,\n title = {Openpose: Realtime multi-person 2d pose estimation using part affinity fields},\n author = {Cao, Zhe and Hidalgo, Gines and Simon, Tomas and Wei, Shih-En and Sheikh, Yaser},\n year = {2019},\n journal = {IEEE transactions on pattern analysis and machine intelligence},\n volume = {43},\n number = {1},\n pages = {172--186},\n publisher = {IEEE}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Task Definition" ], "abstract": "OpenPose estimates the poses of multiple people through bottom-up detection and association of body parts. Part affinity fields encode relationships between detected joints, allowing efficient grouping without processing each person independently. The system extends to body, foot, hand and facial keypoints, and examines design choices that improve accuracy while retaining practical inference speed.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://arxiv.org/abs/1812.08008", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-an2018improving", "citationKey": "an2018improving", "archiveKey": "an2018improving", "method": "", "title": "Improving gait recognition with 3d pose estimation", "authors": "An, Weizhi and Liao, Rijun and Yu, Shiqi and Huang, Yongzhen and Yuen, Pong C", "year": 2018, "venue": "CCBR", "motivation": "2D poses vary with viewpoint and omit depth-related motion.", "idea": "Use estimated 3D poses to support gait identity recognition.", "techniques": "3D pose estimation; skeletal motion features; cross-view matching.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1007/978-3-319-97909-0_15", "bibtex": "@inproceedings{an2018improving,\n title = {Improving gait recognition with 3d pose estimation},\n author = {An, Weizhi and Liao, Rijun and Yu, Shiqi and Huang, Yongzhen and Yuen, Pong C},\n year = {2018},\n booktitle = {Chinese conference on biometric recognition},\n pages = {137--147}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "This study investigates whether estimated three-dimensional pose can improve gait recognition across camera views. It derives pose information from monocular video and combines spatial feature learning with recurrent temporal modeling. The resulting representation uses articulated motion to reduce dependence on the appearance changes that complicate recognition from two-dimensional images.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://scholars.hkbu.edu.hk/en/publications/improving-gait-recognition-with-3d-pose-estimation-2/", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-takemura2018multi", "citationKey": "takemura2018multi", "archiveKey": "takemura2018multi", "method": "", "title": "Multi-view large population gait dataset and its performance evaluation for cross-view gait recognition", "authors": "Takemura, Noriko and Makihara, Yasushi and Muramatsu, Daigo and Echigo, Tomio and Yagi, Yasushi", "year": 2018, "venue": "IPSJ CVA", "motivation": "Small populations limit evaluation of cross-view gait recognition.", "idea": "Collect a large population from multiple synchronized viewpoints.", "techniques": "OU-MVLP; multiview acquisition; cross-view evaluation.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1186/s41074-018-0039-6", "bibtex": "@article{takemura2018multi,\n title = {Multi-view large population gait dataset and its performance evaluation for cross-view gait recognition},\n author = {Takemura, Noriko and Makihara, Yasushi and Muramatsu, Daigo and Echigo, Tomio and Yagi, Yasushi},\n year = {2018},\n journal = {IPSJ transactions on Computer Vision and Applications},\n volume = {10},\n number = {1},\n pages = {4},\n publisher = {Springer}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution", "Background / Broader Applications", "Advances in Benchmarks / Popular Datasets", "Advances in Benchmarks / Research Projects" ], "abstract": "This paper describes the world’s largest gait database with wide view variation, the “OU-ISIR gait database, multi-view large population dataset (OU-MVLP)”, and its application to a statistically reliable performance evaluation of vision-based cross-view gait recognition. Specifically, we construct a gait dataset that includes 10,307 subjects (5114 males and 5193 females) from 14 view angles ranging 0°–90°, 180°–270°. In addition, we evaluate various approaches to gait recognition which are robust against view angles. By using our dataset, we can fully exploit a state-of-the-art method requiring a large number of training samples, e.g., CNN-based cross-view gait recognition method, and we validate effectiveness of such a family of the methods.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1186/s41074-018-0039-6", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-uddin2018isir", "citationKey": "uddin2018isir", "archiveKey": "uddin2018isir", "method": "", "title": "The ou-isir large population gait database with real-life carried object and its performance evaluation", "authors": "Uddin, Md Zasim and Ngo, Thanh Trung and Makihara, Yasushi and Takemura, Noriko and Li, Xiang and Muramatsu, Daigo and Yagi, Yasushi", "year": 2018, "venue": "IPSJ CVA", "motivation": "Carried objects alter silhouettes and challenge identity matching.", "idea": "Provide a large gait dataset with realistic carried-object variation.", "techniques": "OU-ISIR carried-object collection; covariate protocols; baseline evaluation.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1186/s41074-018-0041-z", "bibtex": "@article{uddin2018isir,\n title = {The ou-isir large population gait database with real-life carried object and its performance evaluation},\n author = {Uddin, Md Zasim and Ngo, Thanh Trung and Makihara, Yasushi and Takemura, Noriko and Li, Xiang and Muramatsu, Daigo and Yagi, Yasushi},\n year = {2018},\n journal = {IPSJ Transactions on Computer Vision and Applications},\n volume = {10},\n number = {1},\n pages = {5},\n publisher = {Springer}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Broader Applications" ], "abstract": "This work introduces a large OU-ISIR gait collection with naturally varied carried objects and manually annotated carrying status. The dataset supports analysis of training-set size, recognition difficulty and carrying-status prediction across a broad population. Cooperative and less cooperative evaluation settings expose how carried-object variation affects gait matching.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://doi.org/10.1186/s41074-018-0041-z", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-he2018multi", "citationKey": "he2018multi", "archiveKey": "he2018multi", "method": "", "title": "Multi-task GANs for view-specific feature learning in gait recognition", "authors": "He, Yiwei and Zhang, Junping and Shan, Hongming and Wang, Liang", "year": 2018, "venue": "TIFS", "motivation": "Viewpoint variation makes gait features difficult to compare.", "idea": "Learn view-specific representations through joint generative tasks.", "techniques": "Multi-task generative adversarial networks; view-conditioned feature learning.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@article{he2018multi,\n title = {Multi-task GANs for view-specific feature learning in gait recognition},\n author = {He, Yiwei and Zhang, Junping and Shan, Hongming and Wang, Liang},\n year = {2018},\n journal = {IEEE Transactions on Information Forensics and Security},\n volume = {14},\n number = {1},\n pages = {102--113},\n publisher = {IEEE}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "The proposed multi-task generative adversarial framework learns gait features for specific viewpoints while sharing information across related tasks. A period energy representation preserves temporal information beyond a simple sequence average. Experiments on established gait datasets investigate how generative learning and coordinated view-specific modeling improve matching when the camera viewpoint changes.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://doi.org/10.1109/TIFS.2018.2844819", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-yu2017gaitgan", "citationKey": "yu2017gaitgan", "archiveKey": "yu2017gaitgan", "method": "", "title": "Gaitgan: Invariant gait feature extraction using generative adversarial networks", "authors": "Yu, Shiqi and Chen, Haifeng and Garcia Reyes, Edel B and Poh, Norman", "year": 2017, "venue": "CVPR Workshops", "motivation": "Viewpoint and appearance covariates interfere with gait identity.", "idea": "Generate normalized gait representations that preserve identity.", "techniques": "Generative adversarial learning; identity preservation; invariant feature extraction.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@inproceedings{yu2017gaitgan,\n title = {Gaitgan: Invariant gait feature extraction using generative adversarial networks},\n author = {Yu, Shiqi and Chen, Haifeng and Garcia Reyes, Edel B and Poh, Norman},\n year = {2017},\n booktitle = {Proceedings of the IEEE conference on computer vision and pattern recognition workshops},\n pages = {30--37}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "GaitGAN transforms gait observations toward a canonical appearance to reduce differences caused by viewpoint, clothing and carried objects. Two discriminators separately encourage realistic generation and preservation of identity. The study evaluates whether this adversarial normalization yields representations that are more suitable for recognition across changing walking conditions.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://openaccess.thecvf.com/content_cvpr_2017_workshops/w6/html/Yu_GaitGAN_Invariant_Gait_CVPR_2017_paper.html", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-shiraga2016geinet", "citationKey": "shiraga2016geinet", "archiveKey": "shiraga2016geinet", "method": "", "title": "Geinet: View-invariant gait recognition using a convolutional neural network", "authors": "Shiraga, Kohei and Makihara, Yasushi and Muramatsu, Daigo and Echigo, Tomio and Yagi, Yasushi", "year": 2016, "venue": "ICB", "motivation": "Handcrafted gait descriptors struggle with cross-view variability.", "idea": "Learn discriminative gait-image features using convolutional networks.", "techniques": "Gait energy images; convolutional representation learning; view-invariant matching.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/icb.2016.7550060", "bibtex": "@inproceedings{shiraga2016geinet,\n title = {Geinet: View-invariant gait recognition using a convolutional neural network},\n author = {Shiraga, Kohei and Makihara, Yasushi and Muramatsu, Daigo and Echigo, Tomio and Yagi, Yasushi},\n year = {2016},\n booktitle = {2016 international conference on biometrics (ICB)},\n pages = {1--8}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "GEINet learns gait features directly from gait energy images using a convolutional neural network. Convolution, pooling and fully connected layers replace manually specified appearance descriptors. Experiments on the OU-ISIR large-population data consider cross-view recognition and verification settings, establishing an early deep-learning baseline for compact silhouette templates.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://doi.org/10.1109/ICB.2016.7550060", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-feng2016learning", "citationKey": "feng2016learning", "archiveKey": "feng2016learning", "method": "", "title": "Learning effective Gait features using LSTM", "authors": "Yang Feng and Yuncheng Li and Jiebo Luo", "year": 2016, "venue": "ICPR", "motivation": "Static gait templates can discard sequential motion information.", "idea": "Learn temporal gait dependencies using recurrent memory.", "techniques": "LSTM sequence modeling; temporal feature aggregation.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/ICPR.2016.7899654", "bibtex": "@inproceedings{feng2016learning,\n title = {Learning effective Gait features using LSTM},\n author = {Yang Feng and Yuncheng Li and Jiebo Luo},\n year = {2016},\n booktitle = {2016 23rd International Conference on Pattern Recognition (ICPR)},\n pages = {325-330},\n doi = {10.1109/ICPR.2016.7899654}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "This work combines pose-sensitive convolutional features with long short-term memory networks to represent walking sequences. Temporal learning retains information that would disappear in an averaged gait template, while a learned mapping helps relate observations across views. The study explores both unlabeled sequence learning and supervised cross-view training for gait recognition.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://vision.unipv.it/CV/materiale2016-17/3rd%20Choice/0213.pdf", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-wu2016comprehensive", "citationKey": "wu2016comprehensive", "archiveKey": "wu2016comprehensive", "method": "", "title": "A comprehensive study on cross-view gait based human identification with deep cnns", "authors": "Wu, Zifeng and Huang, Yongzhen and Wang, Liang and Wang, Xiaogang and Tan, Tieniu", "year": 2016, "venue": "TPAMI", "motivation": "Cross-view recognition requires representations that generalize across viewing angles.", "idea": "Systematically study deep convolutional gait matching.", "techniques": "Deep CNNs; cross-view similarity learning; benchmark evaluation.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@article{wu2016comprehensive,\n title = {A comprehensive study on cross-view gait based human identification with deep cnns},\n author = {Wu, Zifeng and Huang, Yongzhen and Wang, Liang and Wang, Xiaogang and Tan, Tieniu},\n year = {2016},\n journal = {IEEE transactions on pattern analysis and machine intelligence},\n volume = {39},\n number = {2},\n pages = {209--226},\n publisher = {IEEE}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Task Definition", "Background / Historical Evolution" ], "abstract": "This study learns gait similarity with deep convolutional networks and systematically evaluates design choices for recognition across views. It compares preprocessing and network configurations using labeled gait pairs, including changes in clothing and carried objects. The analysis investigates how learned similarity can generalize when training identities are limited and observation conditions vary.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.ee.cuhk.edu.hk/~xgwang/papers/wuHWWTtpami16.pdf", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-lecun2015deep", "citationKey": "lecun2015deep", "archiveKey": "lecun2015deep", "method": "", "title": "Deep learning", "authors": "LeCun, Yann and Bengio, Yoshua and Hinton, Geoffrey", "year": 2015, "venue": "Nature", "motivation": "Complex perception tasks require representations beyond handcrafted features.", "idea": "Explain hierarchical feature learning and its emerging applications.", "techniques": "Convolutional networks; recurrent networks; backpropagation; representation learning.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@article{lecun2015deep,\n title = {Deep learning},\n author = {LeCun, Yann and Bengio, Yoshua and Hinton, Geoffrey},\n year = {2015},\n journal = {nature},\n volume = {521},\n number = {7553},\n pages = {436--444},\n publisher = {Nature Publishing Group UK London}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "This review explains how multilayer neural networks learn increasingly abstract representations through training. It surveys convolutional architectures for perceptual data and recurrent architectures for sequences, relating these models to progress across visual, speech and language tasks. The article also discusses the broader opportunities for representation learning beyond manually designed features.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.nature.com/articles/nature14539", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-hofmann2014tum", "citationKey": "hofmann2014tum", "archiveKey": "hofmann2014tum", "method": "", "title": "The TUM Gait from Audio, Image and Depth (GAID) database: Multimodal recognition of subjects and traits", "authors": "Hofmann, Martin and Geiger, Jurgen and Bachmann, Sebastian and Schuller, Bjorn and Rigoll, Gerhard", "year": 2014, "venue": "JVCIR", "motivation": "Gait recognition needs evaluation across complementary sensor channels and personal traits.", "idea": "Provide synchronized audio, image and depth observations.", "techniques": "TUM GAID; multimodal sensing; subject and trait recognition protocols.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1016/j.jvcir.2013.02.006", "bibtex": "@article{hofmann2014tum,\n title = {The TUM Gait from Audio, Image and Depth (GAID) database: Multimodal recognition of subjects and traits},\n author = {Hofmann, Martin and Geiger, Jurgen and Bachmann, Sebastian and Schuller, Bjorn and Rigoll, Gerhard},\n year = {2014},\n journal = {Journal of Visual Communication and Image Representation},\n volume = {25},\n number = {1},\n pages = {195--206},\n publisher = {Elsevier}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Broader Applications" ], "abstract": "TUM GAID combines synchronized color, depth and audio observations of walking people. The dataset includes changes in clothing and carried objects, together with a smaller repeated-session subset. Recognition protocols and baseline experiments assess identity and related attributes, providing a resource for studying the complementary value of visual, geometric and acoustic gait information.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://www.sciencedirect.com/science/article/pii/S1047320313000266", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-borras2012depth", "citationKey": "borras2012depth", "archiveKey": "borras2012depth", "method": "", "title": "Depth information in human gait analysis: an experimental study on gender recognition", "authors": "Borras, Ricard and Lapedriza, Agata and Igual, Laura", "year": 2012, "venue": "ICIAR", "motivation": "Depth sensing may reveal gait characteristics obscured in conventional images.", "idea": "Evaluate depth information for gait-based gender recognition.", "techniques": "Depth acquisition; gait descriptors; gender classification.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@inproceedings{borras2012depth,\n title = {Depth information in human gait analysis: an experimental study on gender recognition},\n author = {Borras, Ricard and Lapedriza, Agata and Igual, Laura},\n year = {2012},\n booktitle = {Image Analysis and Recognition: 9th International Conference, ICIAR 2012},\n pages = {98--105}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Task Definition" ], "abstract": "This study introduces DGait to investigate depth information in gait analysis. Videos capture people walking in multiple directions and include identity, gender and age annotations, with complete walking cycles marked. Gender-classification experiments illustrate how depth can provide additional information for analyzing walking patterns beyond conventional intensity images.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://doi.org/10.1007/978-3-642-31298-4_12", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-iwama2012isir", "citationKey": "iwama2012isir", "archiveKey": "iwama2012isir", "method": "", "title": "The OU-ISIR gait database comprising the large population dataset and performance evaluation of gait recognition", "authors": "Iwama, Haruyuki and Okumura, Mayu and Makihara, Yasushi and Yagi, Yasushi", "year": 2012, "venue": "TIFS", "motivation": "Recognition scalability cannot be assessed reliably on small subject sets.", "idea": "Establish a large-population gait database and systematic evaluation.", "techniques": "OU-ISIR large-population dataset; gait matching; population-scale analysis.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/tifs.2012.2204253", "bibtex": "@article{iwama2012isir,\n title = {The OU-ISIR gait database comprising the large population dataset and performance evaluation of gait recognition},\n author = {Iwama, Haruyuki and Okumura, Mayu and Makihara, Yasushi and Yagi, Yasushi},\n year = {2012},\n journal = {IEEE Transactions on Information Forensics and Security},\n volume = {7},\n number = {5},\n pages = {1511--1521},\n publisher = {IEEE}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Broader Applications" ], "abstract": "The OU-ISIR large-population dataset expands gait evaluation to thousands of people across a broad age range. Its size enables statistically meaningful comparisons between gait representations that smaller collections cannot reliably separate. The study also examines how recognition performance varies with age and gender, including changes associated with human growth.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://doi.org/10.1109/TIFS.2012.2204253", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-wang2010chrono", "citationKey": "wang2010chrono", "archiveKey": "wang2010chrono", "method": "", "title": "Chrono-gait image: A novel temporal template for gait recognition", "authors": "Wang, Chen and Zhang, Junping and Pu, Jian and Yuan, Xiaoru and Wang, Liang", "year": 2010, "venue": "ECCV", "motivation": "Averaged gait images suppress the temporal order of motion.", "idea": "Encode chronological information in a compact gait template.", "techniques": "Chrono-gait image; temporal encoding; template matching.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@inproceedings{wang2010chrono,\n title = {Chrono-gait image: A novel temporal template for gait recognition},\n author = {Wang, Chen and Zhang, Junping and Pu, Jian and Yuan, Xiaoru and Wang, Liang},\n year = {2010},\n booktitle = {European Conference on Computer Vision},\n pages = {257--270}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "Chrono-Gait Image encodes the ordering of gait contours into a compact temporal template. Color mapping distinguishes successive walking phases before their information is aggregated. The method also generates synthetic template variations to improve training coverage, investigating how a single-image representation can preserve more temporal structure than uniform silhouette averaging.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://researchportal.bath.ac.uk/en/publications/chrono-gait-image-a-novel-temporal-template-for-gait-recognition/", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-hossain2010clothing", "citationKey": "hossain2010clothing", "archiveKey": "hossain2010clothing", "method": "", "title": "Clothing-invariant gait identification using part-based clothing categorization and adaptive weight control", "authors": "Hossain, Md Altab and Makihara, Yasushi and Wang, Junqiu and Yagi, Yasushi", "year": 2010, "venue": "PR", "motivation": "Clothing changes affect different body regions unequally.", "idea": "Adapt part weights according to clothing-related reliability.", "techniques": "Body-part decomposition; clothing categorization; adaptive weighting.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2009.12.020", "bibtex": "@article{hossain2010clothing,\n title = {Clothing-invariant gait identification using part-based clothing categorization and adaptive weight control},\n author = {Hossain, Md Altab and Makihara, Yasushi and Wang, Junqiu and Yagi, Yasushi},\n year = {2010},\n journal = {Pattern Recognition},\n volume = {43},\n number = {6},\n pages = {2281--2291},\n publisher = {Elsevier}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Broader Applications", "Advances in Benchmarks / Research Projects" ], "abstract": "Variations in clothing alter an individual’s appearance, making the problem of gait identification much more difficult. If the type of clothing differs between the gallery and a probe, certain parts of the silhouettes are likely to change and the ability to discriminate subjects decreases with respect to these parts. A part-based approach, therefore, has the potential of selecting the appropriate parts. This paper proposes a method for part-based gait identification in the light of substantial clothing variations. We divide the human body into eight sections, including four overlapping ones, since the larger parts have a higher discrimination capability, while the smaller parts are more likely to be unaffected by clothing variations. Furthermore, as there are certain clothes that are common to different parts, we present a categorization for items of clothing that groups similar clothes. Next, we exploit the discrimination capability as a matching weight for each part and control the weights adaptively based on the distribution of distances between the probe and all the galleries. The results of the experiments using our large-scale gait dataset with clothing variations show that the proposed method achieves far better performance than other approaches. & 2010 Elsevier Ltd. All rights reserved.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2009.12.020", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-cho2009vision", "citationKey": "cho2009vision", "archiveKey": "cho2009vision", "method": "", "title": "A vision-based analysis system for gait recognition in patients with Parkinson's disease", "authors": "Cho, Chien-Wen and Chao, Wen-Hung and Lin, Sheng-Huang and Chen, You-Yin", "year": 2009, "venue": "ESWA", "motivation": "Parkinsonian gait assessment benefits from objective movement measurements.", "idea": "Analyze patient walking through a vision-based system.", "techniques": "Video gait analysis; movement descriptors; patient gait evaluation.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1016/j.eswa.2008.08.076", "bibtex": "@article{cho2009vision,\n title = {A vision-based analysis system for gait recognition in patients with Parkinson's disease},\n author = {Cho, Chien-Wen and Chao, Wen-Hung and Lin, Sheng-Huang and Chen, You-Yin},\n year = {2009},\n journal = {Expert Systems with applications},\n volume = {36},\n number = {3},\n pages = {7033--7039},\n publisher = {Elsevier}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Introduction", "Background / Broader Applications" ], "abstract": "This study develops a video-based approach to distinguish Parkinsonian walking patterns from those of control participants. Principal component analysis reduces image dimensionality, while linear discriminant analysis emphasizes differences between gait categories. Experiments compare the resulting features for gait classification, assessing the potential of visual analysis as an aid to clinical assessment.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://scholar.nycu.edu.tw/en/publications/a-vision-based-analysis-system-for-gait-recognition-in-patients-w/", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-seely2008soton", "citationKey": "seely2008soton", "archiveKey": "seely2008soton", "method": "", "title": "The University of Southampton Multi-Biometric Tunnel and Introducing a Novel 3D Gait Dataset", "authors": "Seely, Richard D. and Samangooei, Sina and Middleton, Lee and Carter, John N. and Nixon, Mark S.", "year": 2008, "venue": "ICB", "motivation": "Multimodal and 3D biometrics require coordinated acquisition resources.", "idea": "Develop a capture tunnel and a 3D gait dataset.", "techniques": "Multi-biometric acquisition; synchronized observations; 3D gait reconstruction.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/btas.2008.4699353", "bibtex": "@inproceedings{seely2008soton,\n title = {The University of Southampton Multi-Biometric Tunnel and Introducing a Novel 3D Gait Dataset},\n author = {Seely, Richard D. and Samangooei, Sina and Middleton, Lee and Carter, John N. and Nixon, Mark S.},\n year = {2008},\n booktitle = {Proceedings of the IEEE International Conference on Biometrics: Theory, Applications and Systems (BTAS)},\n pages = {1--6}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Advances in Benchmarks / Research Projects" ], "abstract": "The Southampton multi-biometric tunnel acquires gait, face and ear observations as people walk through a constrained passage. Synchronized cameras support three-dimensional gait capture without requiring subjects to stop. The study describes the collection system and early recognition experiments, exploring how automated acquisition can support larger multimodal datasets and high-throughput settings.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://eprints.soton.ac.uk/266970/", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-yu2006framework", "citationKey": "yu2006framework", "archiveKey": "yu2006framework", "method": "", "title": "A framework for evaluating the effect of view angle, clothing and carrying condition on gait recognition", "authors": "Yu, Shiqi and Tan, Daoliang and Tan, Tieniu", "year": 2006, "venue": "ICPR", "motivation": "View, clothing and carrying effects require controlled comparisons.", "idea": "Introduce an evaluation framework that separates key gait covariates.", "techniques": "CASIA-B; multiview acquisition; clothing and carrying protocols.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@inproceedings{yu2006framework,\n title = {A framework for evaluating the effect of view angle, clothing and carrying condition on gait recognition},\n author = {Yu, Shiqi and Tan, Daoliang and Tan, Tieniu},\n year = {2006},\n booktitle = {18th international conference on pattern recognition (ICPR'06)},\n volume = {4},\n pages = {441--444}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution", "Background / Broader Applications", "Advances in Benchmarks / Popular Datasets", "Advances in Benchmarks / Research Projects" ], "abstract": "Gait recognition has gained increasing interest from researchers, but there is still no standard evaluation method to compare the performance of different gait recognition algorithms. In this paper , a framework is proposed in an attempt to tackle this problem. The framework consists of a large gait database, a large set of well designed experiments and some evaluation metrics. There are 124 subjects in the database, and the gait data was captured from 11 views. Three variations, namely view angle, clothing and carrying condition changes, are separately considered in the database. The database is one of the largest database among the existing databases. Three sets of experiments, including a total of 363 experiments, are designed in the framework. Some metrics are proposed to evaluate gait recognition algorithms.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-tan2006efficient", "citationKey": "tan2006efficient", "archiveKey": "tan2006efficient", "method": "", "title": "Efficient night gait recognition based on template matching", "authors": "Tan, Daoliang and Huang, Kaiqi and Yu, Shiqi and Tan, Tieniu", "year": 2006, "venue": "ICPR", "motivation": "Low illumination limits conventional gait acquisition and recognition.", "idea": "Use compact templates for nighttime gait matching.", "techniques": "Night gait imaging; template construction; efficient similarity matching.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@inproceedings{tan2006efficient,\n title = {Efficient night gait recognition based on template matching},\n author = {Tan, Daoliang and Huang, Kaiqi and Yu, Shiqi and Tan, Tieniu},\n year = {2006},\n booktitle = {18th international conference on pattern recognition (ICPR'06)},\n volume = {3},\n pages = {1000--1003}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Broader Applications" ], "abstract": "Gait is a useful biometric which can be used to recognize people at a distance when other biometrics are incapable. However, most work on gait recognition has been visible spectrum-oriented over the past decade, ignoring recognition at night which is in reality demandimperative. This paper deals with the problem of night gait recognition via thermal infrared imagery. First of all, human detection is accomplished, based on the Gaussian mixture modeling of the background. Then, human silhouettes are extracted on the basis of preceding detection results. Moreover, a new gait representation called HTI is proposed to characterize gait signatures for recognition. An infrared night gait database was built to provide a foundation for night gait recognition. Experimental results on two gait datasets show the effectiveness of this method.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-nixon2006automatic", "citationKey": "nixon2006automatic", "archiveKey": "nixon2006automatic", "method": "", "title": "Automatic recognition by gait", "authors": "Nixon, Mark S and Carter, John N", "year": 2006, "venue": "Proceedings of the IEEE", "motivation": "Early gait biometrics involved diverse models and evaluation practices.", "idea": "Organize the foundations of automatic identity recognition from walking.", "techniques": "Model-based and appearance-based methods; gait acquisition; biometric evaluation.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@article{nixon2006automatic,\n title = {Automatic recognition by gait},\n author = {Nixon, Mark S and Carter, John N},\n year = {2006},\n journal = {Proceedings of the IEEE},\n volume = {94},\n number = {11},\n pages = {2013--2024},\n publisher = {IEEE}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Task Definition" ], "abstract": "This review surveys gait recognition methods, datasets and evidence supporting walking as a biometric. It examines the appeal of recognition at a distance alongside sensitivity to clothing, footwear, surface and viewing conditions. Connections with medical and psychological research provide context for understanding both the capabilities and limitations of early automatic gait recognition.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://eprints.soton.ac.uk/263369/", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-han2005individual", "citationKey": "han2005individual", "archiveKey": "han2005individual", "method": "", "title": "Individual recognition using gait energy image", "authors": "Han, Jinguang and Bhanu, Bir", "year": 2006, "venue": "TPAMI", "motivation": "Sequence-level gait comparison can be costly and sensitive to silhouette noise.", "idea": "Summarize a gait cycle as an average silhouette image.", "techniques": "Gait energy image; silhouette normalization; compact template matching.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/tpami.2006.38", "bibtex": "@article{han2005individual,\n title = {Individual recognition using gait energy image},\n author = {Han, Jinguang and Bhanu, Bir},\n year = {2006},\n journal = {IEEE transactions on pattern analysis and machine intelligence},\n volume = {28},\n number = {2},\n pages = {316--322},\n publisher = {IEEE}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "The gait energy image summarizes a walking cycle into a compact silhouette-based template. To address limited training examples, the method combines features from observed sequences with synthetic templates that simulate silhouette distortions. Evaluation on the USF HumanID benchmark studies the efficiency of this representation and the benefit of combining real and generated training evidence.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://doi.org/10.1109/TPAMI.2006.38", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-sarkar2005humanid", "citationKey": "sarkar2005humanid", "archiveKey": "sarkar2005humanid", "method": "", "title": "The humanid gait challenge problem: Data sets, performance, and analysis", "authors": "Sarkar, Sudeep and Phillips, P Jonathon and Liu, Zongyi and Vega, Isidro Robledo and Grother, Patrick and Bowyer, Kevin W", "year": 2005, "venue": "TPAMI", "motivation": "Algorithm comparisons need common data and controlled covariate protocols.", "idea": "Define the HumanID gait challenge and a reproducible baseline.", "techniques": "USF HumanID dataset; silhouette matching; covariate-based evaluation.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@article{sarkar2005humanid,\n title = {The humanid gait challenge problem: Data sets, performance, and analysis},\n author = {Sarkar, Sudeep and Phillips, P Jonathon and Liu, Zongyi and Vega, Isidro Robledo and Grother, Patrick and Bowyer, Kevin W},\n year = {2005},\n journal = {IEEE transactions on pattern analysis and machine intelligence},\n volume = {27},\n number = {2},\n pages = {162--177},\n publisher = {IEEE}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Broader Applications", "Advances in Benchmarks / Competitions", "Advances in Benchmarks / Research Projects" ], "abstract": "Identification of people by analysis of gait patterns extracted from video has recently become a popular research problem. However, the conditions under which the problem is “solvable” are not understood or characterized. To provide a means for measuring progress and characterizing the properties of gait recognition, we introduce the HumanID Gait Challenge Problem. The challenge problem consists of a baseline algorithm, a set of 12 experiments, and a large data set. The baseline algorithm estimates silhouettes by background subtraction and performs recognition by temporal correlation of silhouettes. The 12 experiments are of increasing difficulty, as measured by the baseline algorithm, and examine the effects of five covariates on performance. The covariates are: change in viewing angle, change in shoe type, change in walking surface, carrying or not carrying a briefcase, and elapsed time between sequences being compared. Identification rates for the 12 experiments range from 78 percent on the easiest experiment to 3 percent on the hardest. All five covariates had statistically significant effects on performance, with walking surface and time difference having the greatest impact. The data set consists of 1,870 sequences from 122 subjects spanning five covariates (1.2 Gigabytes of data). The gait data, the source code of the baseline algorithm, and scripts to run, score, and analyze the challenge experiments are available at http://www.GaitChallenge.org. This infrastructure supports further development of gait recognition algorithms and additional experiments to understand the strengths and weaknesses of new algorithms. The more detailed the experimental results presented, the more detailed is the possible meta-analysis and greater is the understanding. It is this potential from the adoption of this challenge problem that represents a radical departure from traditional computer vision research methodology.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-shutler2004large", "citationKey": "shutler2004large", "archiveKey": "shutler2004large", "method": "", "title": "On a large sequence-based human gait database", "authors": "Shutler, Jamie D and Grant, Michael G and Nixon, Mark S and Carter, John N", "year": 2004, "venue": "Applications and science in soft computing", "motivation": "Gait research needs sufficiently large sequence collections for evaluation.", "idea": "Build a sequence-based human gait database.", "techniques": "Video acquisition; gait sequences; benchmark organization.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1007/978-3-540-45240-9_46", "bibtex": "@incollection{shutler2004large,\n title = {On a large sequence-based human gait database},\n author = {Shutler, Jamie D and Grant, Michael G and Nixon, Mark S and Carter, John N},\n year = {2004},\n booktitle = {Applications and science in soft computing},\n pages = {339--346},\n publisher = {Springer}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "This work presents the design of a large sequence-based gait database for studying person recognition. Multiple observations per subject support analysis of identity variation, while different capture conditions connect controlled evaluation with less ideal imagery. The paper explains the collection structure and motivates shared datasets as a basis for comparing gait recognition approaches.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://eprints.soton.ac.uk/257901/1/Shutler_2002.pdf", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-veres2004what", "citationKey": "veres2004what", "archiveKey": "veres2004what", "method": "", "title": "What Image Information Is Important in Silhouette-Based Gait Recognition?", "authors": "Veres, G.V. and Gordon, L. and Carter, J.N. and Nixon, M.S.", "year": 2004, "venue": "CVPR", "motivation": "It is unclear which silhouette information contributes most to identity.", "idea": "Analyze the discriminative value of different image regions and measurements.", "techniques": "Silhouette feature analysis; feature selection; recognition evaluation.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/CVPR.2004.1315243", "bibtex": "@inproceedings{veres2004what,\n title = {What Image Information Is Important in Silhouette-Based Gait Recognition?},\n author = {Veres, G.V. and Gordon, L. and Carter, J.N. and Nixon, M.S.},\n year = {2004},\n booktitle = {Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2004. CVPR 2004.},\n volume = {2},\n pages = {776--782},\n publisher = {IEEE},\n doi = {10.1109/CVPR.2004.1315243}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "Gait recognition has recently gained significant attention, especially in vision-based automated human identification at a distance in visual surveillance and monitoring applications. Silhouette-based gait recognition is the one of the most popular methods for recognising moving shapes. This paper aims to investigate the important features in silhouette-based gait recognition from point of view of statistical analysis. It is shown that the average silhouette includes a static component of gait (head and body) as the most important image part, while dynamic component of gait (swings of legs and arms) is ignored as the least important information. At the same time ignoring dynamic part of gait can result in loss in recognition rate in some cases, and the importance of better motion estimation is underlined.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/CVPR.2004.1315243", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-yam2004automated", "citationKey": "yam2004automated", "archiveKey": "yam2004automated", "method": "", "title": "Automated Person Recognition by Walking and Running via Model-Based Approaches", "authors": "Yam, ChewYean and Nixon, Mark S. and Carter, John N.", "year": 2004, "venue": "PR", "motivation": "Person recognition should account for both walking and running dynamics.", "idea": "Describe individual locomotion using an explicit movement model.", "techniques": "Model-based gait extraction; walking and running analysis; identity matching.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1016/j.patcog.2003.09.012", "bibtex": "@article{yam2004automated,\n title = {Automated Person Recognition by Walking and Running via Model-Based Approaches},\n author = {Yam, ChewYean and Nixon, Mark S. and Carter, John N.},\n year = {2004},\n journal = {Pattern Recognition},\n volume = {37},\n number = {5},\n pages = {1057--1072},\n doi = {10.1016/j.patcog.2003.09.012}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "Gait enjoys advantages over other biometrics in that it can be perceived from a distance and is di cult to disguise. Current approaches are mostly statistical and concentrate on walking only. By analysing leg motion we show how we can recognise people not only by the walking gait, but also by the running gait. This is achieved by either of two new modelling approaches which employ coupled oscillators and the biomechanics of human locomotion as the underlying concepts. These models give a plausible method for data reduction by providing estimates of the inclination of the thigh and of the leg, from the image data. Both approaches derive a phase-weighted Fourier description gait signature by automated non-invasive means. One approach is completely automated whereas the other requires speci cation of a single parameter to distinguish between walking and running. Results show that both gaits are potential biometrics, with running being more potent. By its basis in evidence gathering, this new technique can tolerate noise and low resolution.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/j.patcog.2003.09.012", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-kale2004identification", "citationKey": "kale2004identification", "archiveKey": "kale2004identification", "method": "", "title": "Identification of Humans Using Gait", "authors": "Kale, A. and Sundaresan, A. and Rajagopalan, A.N. and Cuntoor, N.P. and Roy-Chowdhury, A.K. and Kruger, V. and Chellappa, R.", "year": 2004, "venue": "TIP", "motivation": "Identity must be inferred from variable-length walking observations.", "idea": "Model the evolution of silhouette-based gait features over time.", "techniques": "Silhouette descriptors; gait dynamics; sequence matching.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/TIP.2004.832865", "bibtex": "@article{kale2004identification,\n title = {Identification of Humans Using Gait},\n author = {Kale, A. and Sundaresan, A. and Rajagopalan, A.N. and Cuntoor, N.P. and Roy-Chowdhury, A.K. and Kruger, V. and Chellappa, R.},\n year = {2004},\n journal = {IEEE Transactions on Image Processing},\n volume = {13},\n number = {9},\n pages = {1163--1173},\n doi = {10.1109/TIP.2004.832865}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "We propose a view-based approach to recognize humans from their gait. Two different image features have been considered: the width of the outer contour of the binarized silhouette of the walking person and the entire binary silhouette itself. To obtain the observation vector from the image features, we employ two different methods. In the first method, referred to as the indirect approach, the high-dimensional image feature is transformed to a lower dimensional space by generating what we call the frame to exemplar (FED) distance. The FED vector captures both structural and dynamic traits of each individual. For compact and effective gait representation and recognition, the gait information in the FED vector sequences is captured in a hidden Markov model (HMM). In the second method, referred to as the direct approach, we work with the feature vector directly (as opposed to computing the FED) and train an HMM. We estimate the HMM parameters (specifically the observation probability ) based on the distance between the exemplars and the image features. In this way, we avoid learning high-dimensional probability density functions. The statistical nature of the HMM lends overall robustness to representation and recognition. The performance of the methods is illustrated using several databases.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/TIP.2004.832865", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-kale2003towards", "citationKey": "kale2003towards", "archiveKey": "kale2003towards", "method": "", "title": "Towards a View Invariant Gait Recognition Algorithm", "authors": "Kale, A. and Chowdhury, A.K.R. and Chellappa, R.", "year": 2003, "venue": "AVSS", "motivation": "Different camera views alter the apparent walking pattern.", "idea": "Reduce viewpoint dependence when comparing gait observations.", "techniques": "View normalization; gait representation; cross-view matching.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/AVSS.2003.1217914", "bibtex": "@inproceedings{kale2003towards,\n title = {Towards a View Invariant Gait Recognition Algorithm},\n author = {Kale, A. and Chowdhury, A.K.R. and Chellappa, R.},\n year = {2003},\n booktitle = {Proceedings of the IEEE Conference on Advanced Video and Signal Based Surveillance, 2003.},\n pages = {143--150},\n publisher = {IEEE Comput. Soc},\n doi = {10.1109/AVSS.2003.1217914}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "Human gait is a spatio-temporal phenomenon and typifies the motion characteristics of an individual. The gait of a person is easily recognizable when extracted from a sideview of the person. Accordingly, gait-recognition algorithms work best when presented with images where the person walks parallel to the camera (i.e. the image plane). However, it is not realistic to expect that this assumption will be valid in most real-life scenarios. Hence it is important to develop methods whereby the side-view can be generated from any other arbitrary view in a simple, yet accurate, manner. That is the main theme of this paper. We show that if the person is far enough from the camera, it is possible to synthesize a side view (referred to as canonical view) from any other arbitrary view using a single camera. Two methods are proposed for doing this: i) by using the perspective projection model, and ii) by using the optical flow based structure from motion equations. A simple camera calibration scheme for this method is also proposed. Examples of synthesized views are presented. Preliminary testing with gait recognition algorithms gives encouraging results. A by-product of this method is a simple algorithm for synthesizing novel views of a planar scene.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/AVSS.2003.1217914", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-cunado2003automatic", "citationKey": "cunado2003automatic", "archiveKey": "cunado2003automatic", "method": "", "title": "Automatic Extraction and Description of Human Gait Models for Recognition Purposes", "authors": "Cunado, David and Nixon, Mark S. and Carter, John N.", "year": 2003, "venue": "CVIU", "motivation": "Automatic gait recognition requires repeatable descriptions of body motion.", "idea": "Extract and parameterize human gait models for identity matching.", "techniques": "Model-based motion extraction; gait parameterization; biometric comparison.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1016/S1077-3142(03)00008-0", "bibtex": "@article{cunado2003automatic,\n title = {Automatic Extraction and Description of Human Gait Models for Recognition Purposes},\n author = {Cunado, David and Nixon, Mark S. and Carter, John N.},\n year = {2003},\n journal = {Computer Vision and Image Understanding},\n volume = {90},\n number = {1},\n pages = {1--41},\n doi = {10.1016/S1077-3142(03)00008-0}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution", "Advances in Benchmarks / Research Projects" ], "abstract": "Using gait as a biometric is of emerging interest. We describe a new model-based moving feature extraction analysis is presented that automatically extracts and describes human gait for recognition. The gait signature is extracted directly from the evidence gathering process. This is possible by using a Fourier series to describe the motion of the upper leg and apply temporal evidence gathering techniques to extract the moving model from a sequence of images. Simulation results highlight potential performance benefits in the presence of noise. Classification uses the k-nearest neighbour rule applied to the Fourier components of the motion of the upper leg. Experimental analysis demonstrates that an improved classification rate is given by the phase-weighted Fourier magnitude information over the use of the magnitude information alone. The improved classification capability of the phase-weighted magnitude information is verified using statistical analysis of the separation of clusters in the feature space. Furthermore, the technique is shown to be able to handle high levels of occlusion, which is of especial importance in gait as the human body is self-occluding. As such, a new technique has been developed to automatically extract and describe a moving articulated shape, the human leg, and shown its potential in gait as a biometric.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/S1077-3142(03)00008-0", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-cuntoor2003combining", "citationKey": "cuntoor2003combining", "archiveKey": "cuntoor2003combining", "method": "", "title": "Combining multiple evidences for gait recognition", "authors": "Cuntoor, Naresh and Kale, Amit and Chellappa, Rama", "year": 2003, "venue": "ICASSP", "motivation": "Individual gait cues provide incomplete identity evidence.", "idea": "Combine complementary recognition evidence.", "techniques": "Multiple gait descriptors; evidence fusion; identity comparison.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@inproceedings{cuntoor2003combining,\n title = {Combining multiple evidences for gait recognition},\n author = {Cuntoor, Naresh and Kale, Amit and Chellappa, Rama},\n year = {2003},\n booktitle = {2003 IEEE International Conference on Acoustics, Speech, and Signal Processing, 2003. Proceedings.(ICASSP'03).},\n volume = {3},\n pages = {III--33}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "This study combines complementary gait cues for identifying people from frontal and side views. Static measurements such as height are considered alongside arm, leg and upper-body motion. Probabilistic and non-probabilistic matching methods feed simple score-combination rules, with experiments across four datasets assessing whether fused evidence improves recognition.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://doi.org/10.1109/ICASSP.2003.1199100", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-sundaresan2003hidden", "citationKey": "sundaresan2003hidden", "archiveKey": "sundaresan2003hidden", "method": "", "title": "A Hidden Markov Model Based Framework for Recognition of Humans from Gait Sequences", "authors": "Sundaresan, Aravind and RoyChowdhury, Amit and Chellappa, Rama", "year": 2003, "venue": "ICIP", "motivation": "Walking sequences vary in timing and observed phase.", "idea": "Model gait as a sequence of latent temporal states.", "techniques": "Hidden Markov models; silhouette features; sequence likelihood matching.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/ICIP.2003.1246624", "bibtex": "@inproceedings{sundaresan2003hidden,\n title = {A Hidden Markov Model Based Framework for Recognition of Humans from Gait Sequences},\n author = {Sundaresan, Aravind and RoyChowdhury, Amit and Chellappa, Rama},\n year = {2003},\n booktitle = {Proceedings 2003 International Conference on Image Processing (Cat. No.03CH37429)},\n volume = {2},\n pages = {II-93},\n doi = {10.1109/ICIP.2003.1246624}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "In this paper we propose a generic framework based on hidden Markov models (HMMs) for recognition of individuals from their gait. The HMM framework is suitable, because the gait of an individual can be visualized as his adopting postures from a set, in a sequence which has an underlying structured probabilistic nature. The postures that the individual adopts can be regarded as the states of the HMM and are typical to that individual and provide a means of discrimination. The framework assumes that, during gait, the individual transitions between N discrete postures or states but it is not dependent on the particular feature vector used to represent the gait information contained in the postures. The framework, thus, provides flexibility in the selection of the feature vector. The statistical nature of the HMM lends robustness to the model. In this paper we use the binarized background-subtracted image as the feature vector and use different distance metrics, such as those based on the L/sub 1/ and L/sub 2/ norms of the vector difference, and the normalized inner product of the vectors, to measure the similarity between feature vectors. The results we obtain are better than the baseline recognition rates reported before.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1109/ICIP.2003.1246624", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-wang2003silhouette", "citationKey": "wang2003silhouette", "archiveKey": "wang2003silhouette", "method": "", "title": "Silhouette analysis-based gait recognition for human identification", "authors": "Wang, Liang and Tan, Tieniu and Ning, Huazhong and Hu, Weiming", "year": 2003, "venue": "TPAMI", "motivation": "Remote human identification needs cues that do not require facial detail.", "idea": "Recognize individuals from the shape and evolution of silhouettes.", "techniques": "Silhouette analysis; gait feature extraction; identity matching.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@article{wang2003silhouette,\n title = {Silhouette analysis-based gait recognition for human identification},\n author = {Wang, Liang and Tan, Tieniu and Ning, Huazhong and Hu, Weiming},\n year = {2003},\n journal = {IEEE transactions on pattern analysis and machine intelligence},\n volume = {25},\n number = {12},\n pages = {1505--1518},\n publisher = {IEEE}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Introduction", "Background / Broader Applications" ], "abstract": "Human identification at a distance has recently gained growing interest from computer vision researchers. Gait recognition aims essentially to address this problem by identifying people based on the way they walk. In this paper, a simple but efficient gait recognition algorithm using spatial-temporal silhouette analysis is proposed. For each image sequence, a background subtraction algorithm and a simple correspondence procedure are first used to segment and track the moving silhouettes of a walking figure. Then, eigenspace transformation based on Principal Component Analysis (PCA) is applied to time-varying distance signals derived from a sequence of silhouette images to reduce the dimensionality of the input feature space. Supervised pattern classification techniques are finally performed in the lower-dimensional eigenspace for recognition. This method implicitly captures the structural and transitional characteristics of gait. Extensive experimental results on outdoor image sequences demonstrate that the proposed algorithm has an encouraging recognition performance with relatively low computational cost.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-collins2002silhouette", "citationKey": "collins2002silhouette", "archiveKey": "collins2002silhouette", "method": "", "title": "Silhouette-based human identification from body shape and gait", "authors": "Collins, Robert T and Gross, Ralph and Shi, Jianbo", "year": 2002, "venue": "FG", "motivation": "Body shape and walking motion offer complementary remote identity cues.", "idea": "Use silhouette observations to combine shape and gait evidence.", "techniques": "Silhouette templates; body-shape descriptors; gait comparison.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@inproceedings{collins2002silhouette,\n title = {Silhouette-based human identification from body shape and gait},\n author = {Collins, Robert T and Gross, Ralph and Shi, Jianbo},\n year = {2002},\n booktitle = {Proceedings of fifth IEEE international conference on automatic face gesture recognition},\n pages = {366--371}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "This baseline identifies people by comparing silhouette key frames extracted from walking cycles. Normalized correlation and nearest-neighbor matching use body proportions and motion-dependent shape without an elaborate articulated model. Evaluation across four datasets examines varying viewpoints, backgrounds, walking styles and image resolutions, providing a reference point for more complex methods.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://publications.ri.cmu.edu/silhouette-based-human-identification-from-body-shape-and-gait", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-huang1999recognising", "citationKey": "huang1999recognising", "archiveKey": "huang1999recognising", "method": "", "title": "Recognising Humans by Gait via Parametric Canonical Space", "authors": "Huang, P.S. and Harris, C.J. and Nixon, M.S.", "year": 1999, "venue": "Artificial Intelligence in Engineering", "motivation": "Gait observations require a compact discriminative representation for comparison.", "idea": "Map gait patterns into a parametric canonical space.", "techniques": "Canonical-space projection; gait parameterization; identity classification.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1016/S0954-1810(99)00008-4", "bibtex": "@article{huang1999recognising,\n title = {Recognising Humans by Gait via Parametric Canonical Space},\n author = {Huang, P.S. and Harris, C.J. and Nixon, M.S.},\n year = {1999},\n journal = {Artificial Intelligence in Engineering},\n volume = {13},\n number = {4},\n pages = {359--366},\n doi = {10.1016/S0954-1810(99)00008-4}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "Based on principal component analysis (PCA), eigenspace transformation (EST) was demonstrated to be a potent metric in automatic face recognition and gait analysis by template matching, but without using data analysis to increase classification capability. Gait is a new biometric aimed to recognise subjects by the way they walk. In this article, we propose a new approach which combines canonical space transformation (CST) based on Canonical Analysis (CA), with EST for feature extraction. This method can be used to reduce data dimensionality and to optimise the class separability of different gait classes simultaneously. Each image template is projected from the high-dimensional image space to a low-dimensional canonical space. Using template matching, recognition of human gait becomes much more accurate and robust in this new space. Experimental results on a small database show how subjects can be recognised with 100% accuracy by their gait, using this method. q 1999 Elsevier Science Ltd. All rights reserved.", "abstractKind": "Original abstract", "abstractSource": "Original paper abstract", "abstractUrl": "https://doi.org/10.1016/S0954-1810(99)00008-4", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-cunado1997using", "citationKey": "cunado1997using", "archiveKey": "cunado1997using", "method": "", "title": "Using gait as a biometric, via phase-weighted magnitude spectra", "authors": "Cunado, David and Nixon, Mark S and Carter, John N", "year": 1997, "venue": "AVBPA", "motivation": "Individual leg-motion patterns may support biometric identification.", "idea": "Represent periodic motion with phase-sensitive spectral descriptors.", "techniques": "Model-based motion analysis; phase-weighted magnitude spectra.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "", "bibtex": "@inproceedings{cunado1997using,\n title = {Using gait as a biometric, via phase-weighted magnitude spectra},\n author = {Cunado, David and Nixon, Mark S and Carter, John N},\n year = {1997},\n booktitle = {International Conference on Audio-and Video-based Biometric Person Authentication},\n pages = {93--102}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Introduction" ], "abstract": "This early model-based approach describes leg movement using an articulated pendulum interpretation. The Hough transform estimates leg orientations, least-squares fitting smooths incomplete measurements, and Fourier analysis characterizes their periodic variation. Phase-sensitive spectral features are classified with nearest-neighbor matching to investigate the identifying information carried by walking dynamics.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://eprints.soton.ac.uk/250040/1/cunado_avbpa.pdf", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-niyogi1994analyzing", "citationKey": "niyogi1994analyzing", "archiveKey": "niyogi1994analyzing", "method": "", "title": "Analyzing and recognizing walking figures in XYT", "authors": "Niyogi and Adelson", "year": 1994, "venue": "CVPR", "motivation": "Walking contains distinctive motion structure across space and time.", "idea": "Analyze walking figures through their trajectories in an XYT volume.", "techniques": "Spatiotemporal patterns; walking-motion analysis; identity comparison.", "inputs": "", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "", "codeUrl": "", "url": "https://doi.org/10.1109/cvpr.1994.323868", "bibtex": "@inproceedings{niyogi1994analyzing,\n title = {Analyzing and recognizing walking figures in XYT},\n author = {Niyogi and Adelson},\n year = {1994},\n booktitle = {1994 Proceedings of IEEE Conference on Computer Vision and Pattern Recognition},\n pages = {469--474}\n}", "scope": "Cited background", "provenance": [ { "source": "Manuscript bibliography" }, { "source": "Verified publication metadata" }, { "source": "Companion editorial summary", "basis": "Survey text and the recorded literature evidence" } ], "summaryStatus": "Editorial summary", "sections": [ "Background / Historical Evolution" ], "abstract": "This work analyzes walking through the patterns it creates in space and time. Rather than recovering body structure independently in each frame, it uses characteristic motion signatures to constrain a simple articulated representation. These temporal patterns provide a basis for detecting walkers and investigating differences between individuals' gait.", "abstractKind": "Abstract summary", "abstractSource": "Editorial summary of the linked research paper", "abstractUrl": "https://persci.mit.edu/pub_abstracts/niyogi_XYT_abs.html", "complete": true, "statisticsEligible": false, "venueInScope": true }, { "id": "ref-mitra2025gaitcrafter", "citationKey": "mitra2025gaitcrafter", "archiveKey": "mitra2025gaitcrafter", "method": "GaitCrafter", "title": "GaitCrafter: Diffusion Model for Biometric Preserving Gait Synthesis", "authors": "Mitra, Sirshapan and Rawat, Yogesh Singh", "year": 2025, "venue": "ICCV Workshops", "motivation": "Collecting diverse walking sequences for each identity is costly.", "idea": "Generate controllable silhouette sequences and study their value for gait recognition.", "techniques": "Conditional video diffusion; identity embedding interpolation; silhouette augmentation.", "inputs": "Silhouette", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://openaccess.thecvf.com/content/ICCV2025W/FoundGen-Bio/html/Mitra_GaitCrafter_Diffusion_Model_for_Biometric_Preserving_Gait_Synthesis_ICCVW_2025_paper.html", "bibtex": "@inproceedings{mitra2025gaitcrafter,\n author = {Mitra, Sirshapan and Rawat, Yogesh Singh},\n booktitle = {Proc. IEEE/CVF Int. Conf. Comput. Vis. Workshops},\n pages = {3726--3735},\n title = {{GaitCrafter}: Diffusion Model for Biometric\nPreserving Gait Synthesis},\n url = {https://openaccess.thecvf.com/content/ICCV2025W/FoundGen-Bio/html/Mitra_GaitCrafter_Diffusion_Model_for_Biometric_Preserving_Gait_Synthesis_ICCVW_2025_paper.html},\n year = {2025}\n}", "scope": "Survey collection", "provenance": [ { "source": "Manuscript alignment, 2026-10-14", "evidence": "Author-supplied manuscript and linked primary publication page; metadata cross-checked against the bibliography audit." } ], "summaryStatus": "Editorial summary", "notes": "Outside the frozen 2020-2025 statistical snapshot.", "sections": [ "Advances in Methods / Synthetic Gait Data and Generative Learning" ], "abstract": "GaitCrafter synthesizes silhouettes conditioned on identity, clothing, carried objects and view. Interpolating identity embeddings creates candidate novel identities. The survey discusses its controlled-benchmark gains while distinguishing added sample count from identity diversity and unconstrained real-world transfer.", "abstractKind": "Abstract summary", "abstractSource": "Concise editorial summary of the linked publication", "abstractUrl": "https://openaccess.thecvf.com/content/ICCV2025W/FoundGen-Bio/html/Mitra_GaitCrafter_Diffusion_Model_for_Biometric_Preserving_Gait_Synthesis_ICCVW_2025_paper.html", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-mcdaniel2026frailty", "citationKey": "mcdaniel2026frailty", "archiveKey": "mcdaniel2026frailty", "method": "Gait Signature of Frailty", "title": "The Gait Signature of Frailty: Transfer Learning based Deep Gait Models for Scalable Frailty Assessment", "authors": "McDaniel, Laura and Pal, Basudha and Szczesny, Crystal and Guo, Yuxiang and Wang, Zhaoyang and Roemmich, Ryan and Abadir, Peter and Chellappa, Rama", "year": 2026, "venue": "arXiv", "motivation": "Frailty assessment needs scalable measures despite limited clinical training data.", "idea": "Adapt pretrained gait recognition representations to frailty classification.", "techniques": "Silhouette features; selective freezing; transfer learning; frailty assessment.", "inputs": "Silhouette", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://arxiv.org/abs/2603.24434", "bibtex": "@article{mcdaniel2026frailty,\n author = {McDaniel, Laura and Pal, Basudha and Szczesny, Crystal and Guo, Yuxiang and Wang, Zhaoyang and Roemmich, Ryan and Abadir, Peter and Chellappa, Rama},\n journal = {arXiv:2603.24434},\n title = {{The Gait Signature of Frailty: Transfer Learning based Deep Gait Models for Scalable Frailty Assessment}},\n url = {https://arxiv.org/abs/2603.24434},\n year = {2026}\n}", "scope": "Related research", "provenance": [ { "source": "Manuscript alignment, 2026-10-14", "evidence": "Author-supplied manuscript and linked primary publication page; metadata cross-checked against the bibliography audit." } ], "summaryStatus": "Editorial summary", "notes": "Outside the frozen 2020-2025 statistical snapshot.", "sections": [ "Challenges and Outlooks / Gait Representation Learning for Healthcare" ], "abstract": "This preprint studies frailty classification using gait models and a silhouette dataset that includes older adults using walking aids. It examines how selectively adapting pretrained representations affects performance under limited data. Clinical value requires evaluation beyond identity-recognition accuracy.", "abstractKind": "Abstract summary", "abstractSource": "Concise editorial summary of the linked publication", "abstractUrl": "https://arxiv.org/abs/2603.24434", "complete": true, "statisticsEligible": false, "venueInScope": false }, { "id": "ref-huang2020movienet", "citationKey": "huang2020movienet", "archiveKey": "huang2020movienet", "method": "MovieNet", "title": "MovieNet: A Holistic Dataset for Movie Understanding", "authors": "Huang, Qingqiu and Xiong, Yu and Rao, Anyi and Wang, Jiaze and Lin, Dahua", "year": 2020, "venue": "ECCV", "motivation": "Long-video understanding requires linking characters, scenes and narrative information.", "idea": "Provide a movie dataset with complementary annotations and evaluation tasks.", "techniques": "Character identity annotations; scene boundaries; multimodal movie understanding.", "inputs": "RGB video", "backbone": "", "loss": "", "datasets": "", "codeAvailable": "Unverified", "codeUrl": "", "url": "https://arxiv.org/abs/2007.10937", "bibtex": "@inproceedings{huang2020movienet,\n author = {Huang, Qingqiu and Xiong, Yu and Rao, Anyi\nand Wang, Jiaze and Lin, Dahua},\n booktitle = {Proc. Eur. Conf. Comput. Vis.},\n title = {{MovieNet}: A Holistic Dataset for Movie Understanding},\n url = {https://arxiv.org/abs/2007.10937},\n year = {2020}\n}", "scope": "Related research", "provenance": [ { "source": "Manuscript alignment, 2026-10-14", "evidence": "Author-supplied manuscript and linked primary publication page; metadata cross-checked against the bibliography audit." } ], "summaryStatus": "Editorial summary", "notes": "Outside the frozen 2020-2025 statistical snapshot.", "sections": [ "Challenges and Outlooks / Gait Recognition for Agents" ], "abstract": "MovieNet provides a dataset and benchmarks for movie understanding, including character identity annotations. It motivates the identity-continuity problem discussed for virtual agents; it does not demonstrate gait recognition or a gait-enabled agent.", "abstractKind": "Abstract summary", "abstractSource": "Concise editorial summary of the linked publication", "abstractUrl": "https://arxiv.org/abs/2007.10937", "complete": true, "statisticsEligible": false, "venueInScope": true } ] }