Research map / Label-efficient and robust learning / Recognition and long-tail
Pose-Guided Feature Alignment for Occluded Person Re-Identification
Jiaxu Miao, Yu Wu, Ping Liu, Yuhang Ding, Yi Yang
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 partial feature branch to the pose-guided global feature (ablation) | PGFA · Occluded-DukeMTMC | Rank-1 (%): 42.5 -> 51.4 | Improved |
| add pose-guided global feature branch to the partial features (ablation) | PGFA · Occluded-DukeMTMC | Rank-1 (%): 46.0 -> 51.4 | Improved |
| add pose-guided global feature branch to the partial features (ablation) | PGFA · Occluded-DukeMTMC | mAP (%): 34.4 -> 37.3 | Improved |
| use shared-region matching on partial features instead of all parts (lambda=1 vs PCB) (ablation) | PCB · Occluded-DukeMTMC | Rank-1 (%): 42.6 -> 46.0 | Improved |
| partition features into parts (p>1) instead of a single global feature (p=1) (ablation) | PGFA · Occluded-DukeMTMC | Rank-1 (%): 51.4% Rank-1 accuracy and 37.2% mAP | Improved |
| increase the number of parts p beyond 3 (ablation) | PGFA · Occluded-DukeMTMC | Rank-1 (%) | Hurt |
| set landmark confidence threshold gamma too small (e.g. 0) or too large (ablation) | PGFA · Occluded-DukeMTMC | Rank-1 (%) | Hurt |
| vary loss coefficient lambda between 0.2 and 0.8 (ablation) | PGFA · Occluded-DukeMTMC | Rank-1 (%) | No clear effect |
| replace AlphaPose with OpenPose as pose estimator (ablation) | PGFA · Occluded-DukeMTMC | Rank-1 (%): 51.4 -> 49.1 | No clear effect |
| use PGFA instead of PCB | PCB · Occluded-DukeMTMC | Rank-1 (%): 42.6 -> 51.4 | Improved |
| use PGFA instead of SFR | SFR · Partial-REID | Rank-1 (%): 56.9 -> 68.0 | Improved |
| use PGFA instead of SFR | SFR · Partial-iLIDS | Rank-1 (%): 63.9 -> 69.1 | Improved |
Closest papers by meaning
- Dynamic Feature Pruning and Consolidation for Occluded Person Re-identification (AAAI 2024 · similarity 0.92)
- Learning To Know Where To See: A Visibility-Aware Approach for Occluded Person Re-Identification (ICCV 2021 · similarity 0.92)
- Pose-Guided Feature Disentangling for Occluded Person Re-identification Based on Transformer (AAAI 2022 · similarity 0.92)
- Matching on Sets: Conquer Occluded Person Re-identification Without Alignment (AAAI 2021 · similarity 0.91)
- Foreground-Aware Pyramid Reconstruction for Alignment-Free Occluded Person Re-Identification (ICCV 2019 · similarity 0.91)
- From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (CVPR 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.
