Research map / 3D vision / Human pose estimation
Learning To Know Where To See: A Visibility-Aware Approach for Occluded Person Re-Identification
Jinrui Yang, Jiawei Zhang, Fufu Yu, Xinyang Jiang, Mengdan Zhang, Xing Sun, Ying-Cong Chen, Wei-Shi Zheng
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 region visibility discriminator (VRD) (ablation) | PCB baseline with VGTri · Occluded-Duke | Rank-1: 47.0 -> 62.2 | Improved |
| add region visibility discriminator (VRD) (ablation) | PCB baseline · Occluded-REID | Rank-1: 71.8 -> 80.6 | Improved |
| add visibility-guided hard triplet loss (VGTri) (ablation) | PCB baseline · Occluded-Duke | Rank-1: 41.7 -> 47.0 | Improved |
| add visibility-guided hard triplet loss (VGTri) (ablation) | PCB with VRD · Occluded-Duke | Rank-1: 61.3 -> 62.2 | Improved |
| VGTri instead of traditional hard triplet loss (ablation) | Ours · Occluded-Duke | Rank-1: 61.2 -> 62.2 | Improved |
| learned soft visibility scores instead of hard pose-detector part labels (ablation) | Ours · Occluded-Duke | Rank-1: 48.1 -> 62.2 | Improved |
| use manual ground-truth part labels instead of learned visibility (ablation) | Ours · Partial-REID | Rank-1: 85.7 -> 88.0 | No clear effect |
| reduce keypoints from 18 to 6 (ablation) | Ours · Occluded-Duke | Rank-1: 62.2 -> 60.7 | No clear effect |
| reduce keypoints from 18 to 6 (ablation) | Ours · Occluded-REID | Rank-1: 81.0 -> 77.7 | Hurt |
| reduce keypoints from 14 to 6 (ablation) | HOReID · Occluded-Duke | Rank-1: 55.1 -> 49.2 | Hurt |
| use manual ground-truth pose cues | PGFA · Occluded-Duke | Rank-1: 51.9 -> 52.8 | Improved |
| use Mask R-CNN instead of AlphaPose keypoints (ablation) | Ours · Occluded-Duke | Rank-1: 62.2 -> 62.9 | No clear effect |
Closest papers by meaning
- Occlude Them All: Occlusion-Aware Attention Network for Occluded Person Re-ID (ICCV 2021 · similarity 0.92)
- Adversarially Occluded Samples for Person Re-Identification (CVPR 2018 · similarity 0.92)
- Pose-Guided Feature Alignment for Occluded Person Re-Identification (ICCV 2019 · similarity 0.92)
- Dynamic Feature Pruning and Consolidation for Occluded Person Re-identification (AAAI 2024 · similarity 0.91)
- From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (CVPR 2025 · similarity 0.89)
- Exploring the Camera Bias of Person Re-identification (ICLR 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.
