Research map / Architectures and efficiency / RNNs and normalization
GaitEdge: Beyond Plain End-to-End Gait Recognition for Better Practicality
Junhao Liang, Chao Fan, Saihui Hou, Chuanfu Shen, Yongzhen Huang, Shiqi Yu
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 the trainable float mask with Gait Synthesis (trainable edge only, binary interior) in end-to-end GaitGL | GaitGL-E2E · CASIA-B* to TTG-200 cross-domain | rank-1 accuracy (%): 37.18 -> 49.12 | Improved |
| replace the trainable float mask with Gait Synthesis | GaitGL-E2E · TTG-200 to CASIA-B* cross-domain (mean) | rank-1 accuracy (%): 41.45 -> 47.59 | Improved |
| replace the trainable float mask with Gait Synthesis | GaitGL-E2E · CASIA-B* single-domain | rank-1 accuracy (%): 95.58 -> 93.45 | Hurt |
| replace the trainable float mask with Gait Synthesis | GaitGL-E2E · TTG-200 single-domain | rank-1 accuracy (%): 90.37 -> 88.66 | Hurt |
| replace binary silhouette input with float-coded segmentation output (end-to-end) | GaitGL · TTG-200 to CASIA-B* cross-domain (CL) | rank-1 accuracy (%): 40.34 -> 27.18 | Hurt |
| replace binary silhouette input with float-coded segmentation output (end-to-end) | GaitGL · CASIA-B* single-domain (mean) | rank-1 accuracy (%): 88.52 -> 95.58 | Improved |
| replace binary silhouette input with float-coded segmentation output (end-to-end) | GaitGL · CASIA-B* to TTG-200 cross-domain | rank-1 accuracy (%): 48.74 -> 37.18 | Hurt |
| enlarge structuring element (wider trainable edge) from 3x3 to 9x9 (ablation) | GaitEdge · CASIA-B* single-domain (mean) | rank-1 accuracy (%): 93.45 -> 95.42 | Improved |
| enlarge structuring element (wider trainable edge) from 3x3 to 9x9 (ablation) | GaitEdge · CASIA-B* to TTG-200 cross-domain | rank-1 accuracy (%): 49.12 -> 44.47 | Hurt |
| enlarge structuring element (wider trainable edge) from 3x3 to 9x9 (ablation) | GaitEdge · TTG-200 to CASIA-B* cross-domain (mean) | rank-1 accuracy (%): 47.59 -> 39.41 | Hurt |
| enlarge structuring element from 3x3 to 9x9 (ablation) | GaitEdge · TTG-200 single-domain | rank-1 accuracy (%): 88.66 -> 90.39 | Improved |
| add GaitAlign module under disturbed (offset) detection boxes (ablation) | GaitEdge · CASIA-B* with random pixel offset (NM) | rank-1 accuracy (%): -0.38% | Hurt |
Closest papers by meaning
- QAGait: Revisit Gait Recognition from a Quality Perspective (AAAI 2024 · similarity 0.89)
- Gait Transformer: End-to-End Transformer Backbone for Gait Recognition (AAAI 2026 · similarity 0.88)
- Exploring More from Multiple Gait Modalities for Human Identification (AAAI 2025 · similarity 0.87)
- BigGait: Learning Gait Representation You Want by Large Vision Models (CVPR 2024 · similarity 0.87)
- RA-GAR: A Richly Annotated Benchmark for Gait Attribute Recognition (AAAI 2025 · similarity 0.85)
- On Denoising Walking Videos for Gait Recognition (CVPR 2025 · similarity 0.85)
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.
