Research map / Label-efficient and robust learning / Recognition and long-tail
On Learning Disentangled Representations for Gait Recognition
Ziyuan Zhang, Luan Tran, Feng Liu, Xiaoming Liu
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 |
|---|---|---|---|
| add cross reconstruction loss (ablation) | GaitNet dynamic feature · CASIA-B ablation (NM vs CL) | Rank-1 accuracy: 56.0 -> 60.2 | Improved |
| add pose similarity loss (ablation) | GaitNet · CASIA-B ablation (NM vs CL) | Rank-1 accuracy: 60.2 -> 85.6 | Improved |
| add canonical consistency loss and static feature (ablation) | GaitNet · CASIA-B ablation (NM vs CL) | Rank-1 accuracy: 85.6 -> 92.1 | Improved |
| use incremental identity loss instead of final-step loss (ablation) | GaitNet · CASIA-B ablation (NM vs CL) | Rank-1 accuracy: 72.5 -> 92.1 | Improved |
| use incremental identity loss instead of averaged loss (ablation) | GaitNet · CASIA-B ablation (NM vs CL) | Rank-1 accuracy: 82.6 -> 92.1 | Improved |
| use incremental identity loss instead of autoencoder loss (ablation) | GaitNet · CASIA-B ablation (NM vs CL) | Rank-1 accuracy: 76.5 -> 92.1 | Improved |
| fuse static with dynamic feature (ablation) | GaitNet · CASIA-B ablation (NM vs CL) | Rank-1 accuracy: 85.9 -> 92.1 | Improved |
| use canonical static feature instead of appearance feature (ablation) | GaitNet · CASIA-B ablation (NM vs CL) | Rank-1 accuracy: 33.4 -> 76.3 | Improved |
| add canonical feature (GaitNet vs GaitNet-pre) (ablation) | GaitNet · CASIA-B protocol 2 (NM) | average Rank-1: 81.8 -> 87.3 | Improved |
| add canonical feature (GaitNet vs GaitNet-pre) (ablation) | GaitNet · CASIA-B protocol 3 CL | mean accuracy: 89.8 -> 96.7 | Improved |
| add canonical feature (GaitNet vs GaitNet-pre) (ablation) | GaitNet · FVG ALL | TAR@1% FAR: 81.2 -> 91.9 | Improved |
| add canonical feature (GaitNet vs GaitNet-pre) (ablation) | GaitNet · USF Probe A | accuracy: 99.5 -> 99.7 | No clear effect |
Closest papers by meaning
- Gait Recognition via Disentangled Representation Learning (CVPR 2019 · similarity 0.98)
- Gait Recognition via Semi-supervised Disentangled Representation Learning to Identity and Covariate Features (CVPR 2020 · similarity 0.91)
- Exploring More from Multiple Gait Modalities for Human Identification (AAAI 2025 · similarity 0.88)
- Learning Visual Prompt for Gait Recognition (CVPR 2024 · similarity 0.88)
- On Denoising Walking Videos for Gait Recognition (CVPR 2025 · similarity 0.88)
- BiggerGait: Unlocking Gait Recognition with Layer-wise Representations from Large Vision Models (NeurIPS 2025 · 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.
