Research map / Label-efficient and robust learning / Recognition and long-tail
Free Lunch for Gait Recognition: A Novel Relation Descriptor
Jilong Wang, Saihui Hou, Yan Huang, Chunshui Cao, Xu Liu, Yongzhen Huang, Tianzhu Zhang, Liang Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace embedding with relation descriptor, FAS and SVD | GaitBase · GREW | Rank-1 accuracy (%): 60.1 -> 65.5 | Improved |
| replace embedding with relation descriptor and FAS | GaitBase · Gait3D | Rank-1 accuracy (%): 64.6 -> 70.1 | Improved |
| use relation descriptor on OU-MVLP | GaitBase · OU-MVLP | mean rank-1 accuracy (%): 90.8 -> 91.3 | Improved |
| use plain relation descriptor with only 74 anchors (ablation) | GaitGL · CASIA-B | mean accuracy (%): 92.8 -> 91.4 | Hurt |
| add orthogonal regularization loss to relation descriptor (ablation) | GaitGL · CASIA-B | mean accuracy (%): 91.4 -> 93.4 | Improved |
| use plain relation descriptor with all classifier weights as anchors (ablation) | GaitBase · Gait3D | Rank-1 accuracy (%): 64.6 -> 68.8 | Improved |
| add Farthest gait-Anchor Selection (ablation) | relation descriptor on GaitBase · Gait3D | Rank-1 accuracy (%): 68.8 -> 70.1 | Improved |
| apply relation descriptor to other backbones | GaitSet, GaitPart, GaitGL · Gait3D | Rank-1 accuracy (%): 36.7 -> 42.5, 28.2 -> 38.0, 29.7 -> 37.1 | Improved |
| vary number of selected anchors N, and use all weights or too few (ablation) | FAS · Gait3D | Rank-1 accuracy (%) | Mixed |
| select anchors with FAS instead of randomly (ablation) | relation descriptor · Gait3D | Rank-1 accuracy (%) | Improved |
| use relation descriptor instead of embedding in cross-domain testing | GaitBase · cross-domain (Gait3D to CCPG) | Rank-1 accuracy (%): +7.0% | Improved |
| use relation descriptor on CASIA-B with GaitBase | GaitBase · CASIA-B | mean rank-1 accuracy (%): 94.0 -> 94.1 (BG), 77.4 -> 77.9 (CL) | No clear effect |
Closest papers by meaning
- Learning Gait Representation From Massive Unlabelled Walking Videos: A Benchmark (TPAMI 2023 · similarity 0.90)
- Gait Recognition via Semi-supervised Disentangled Representation Learning to Identity and Covariate Features (CVPR 2020 · similarity 0.88)
- RA-GAR: A Richly Annotated Benchmark for Gait Attribute Recognition (AAAI 2025 · similarity 0.88)
- Vocabulary-Guided Gait Recognition (NeurIPS 2025 · similarity 0.86)
- Cross-Covariate Gait Recognition: A Benchmark (AAAI 2024 · similarity 0.86)
- BiggerGait: Unlocking Gait Recognition with Layer-wise Representations from Large Vision Models (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
