Research map / Label-efficient and robust learning / Recognition and long-tail
GaitSet: Cross-view Gait Recognition through Utilizing Gait as a Deep Set
Hanqing Chao, Kun Wang, Yiwei He, Junping Zhang, Jianfeng Feng
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 |
|---|---|---|---|
| treat gait as an unordered set instead of averaging into a GEI (ablation) | GaitSet · CASIA-B LT NM | rank-1 accuracy: 89.0 -> 95.4 | Improved |
| treat gait as a set instead of GEI (ablation) | GaitSet · CASIA-B LT CL | rank-1 accuracy: 50.7 -> 69.9 | Improved |
| add MGP (multilayer global pipeline) (ablation) | GaitSet · CASIA-B LT | rank-1 accuracy: 95.4 -> 96.1 NM, 88.7 -> 90.8 BG, 69.9 -> 70.3 CL | Improved |
| use max set pooling instead of mean (ablation) | GaitSet · CASIA-B LT NM | rank-1 accuracy: 95.0 -> 95.4 | Improved |
| use pixel-wise attention set pooling instead of max (ablation) | GaitSet · CASIA-B LT NM | rank-1 accuracy: 95.4 -> 95.6 | No clear effect |
| use independent HPM weights instead of shared (ablation) | GaitSet · CASIA-B LT | rank-1 accuracy: 91.1 -> 96.1 NM, 82.0 -> 90.8 BG, 60.9 -> 70.3 CL | Improved |
| increase HPM scales from 1 to 5 (ablation) | GaitSet · CASIA-B LT | rank-1 accuracy: 91.8 -> 96.1 NM, 59.4 -> 70.3 CL | Improved |
| add dropout to cross-entropy training (ablation) | GaitSet · CASIA-B LT CL | rank-1 accuracy: 46.7 -> 64.8 | Improved |
| add batch normalization to training (ablation) | GaitSet · CASIA-B LT | rank-1 accuracy: 94.4 -> 95.6 NM (CE loss) | Improved |
| combine CE and triplet loss instead of either alone (ablation) | GaitSet · CASIA-B LT NM | rank-1 accuracy: 95.6 -> 96.1 | Improved |
| compress feature from 15,872 dims to 1024 with new FC (ablation) | GaitSet · CASIA-B LT NM | rank-1 accuracy: 96.1 -> 95.0 | No clear effect |
| reduce feature dimension too low (128) (ablation) | GaitSet · CASIA-B LT | rank-1 accuracy: 95.0 -> 91.7 NM | Hurt |
Closest papers by meaning
- GaitSet: Regarding Gait as a Set for Cross-View Gait Recognition (AAAI 2019 · similarity 0.97)
- A Comprehensive Study on Cross-View Gait Based Human Identification with Deep CNNs (TPAMI 2016 · similarity 0.93)
- Learning Visual Prompt for Gait Recognition (CVPR 2024 · similarity 0.90)
- Vocabulary-Guided Gait Recognition (NeurIPS 2025 · similarity 0.89)
- BiggerGait: Unlocking Gait Recognition with Layer-wise Representations from Large Vision Models (NeurIPS 2025 · similarity 0.89)
- BigGait: Learning Gait Representation You Want by Large Vision Models (CVPR 2024 · similarity 0.88)
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.
