Research map / Label-efficient and robust learning / Recognition and long-tail
Feature Completion for Occluded Person Re-Identification
Ruibing Hou, Bingpeng Ma, Hong Chang, Xinqian Gu, Shiguang Shan, Xilin Chen
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 SRFC (spatial region feature completion) (ablation) | baseline ResNet50 · Occluded-DukeMTMC (image) | mAP (%): 45.8 -> 51.8 | Improved |
| add SRFC instead of only foreground masks (ablation) | Foreground · Occluded-DukeMTMC-VideoReID | mAP (%): 74.1 -> 82.9 | Improved |
| use adaptive clustering in SRFC instead of keypoint-select (ablation) | Keypoint-Select · Occluded-DukeMTMC (image) | mAP (%): 48.1 -> 51.8 | Improved |
| combine appearance and position assignment in SRFC (ablation) | SRFC-A · Occluded-DukeMTMC-VideoReID | mAP (%): 80.6 -> 82.9 | Improved |
| add TRFC (temporal completion) (ablation) | baseline ResNet50 · Occluded-DukeMTMC-VideoReID | mAP (%): 69.3 -> 82.2 | Improved |
| add TRFC after SRFC (ablation) | SRFC · Occluded-DukeMTMC-VideoReID | mAP (%): 82.9 -> 89.4 | Improved |
| arrange SRFC and TRFC sequentially instead of in parallel (ablation) | S+T-RFC · Occluded-DukeMTMC-VideoReID | mAP (%): 86.5 -> 89.4 | Improved |
| spatial-first instead of temporal-first order (ablation) | T-S-RFC · Occluded-DukeMTMC-VideoReID | mAP (%): 88.9 -> 89.4 | Improved |
| insert RFC after stage 4 instead of stage 3 (ablation) | RFC · Occluded-DukeMTMC (image) | mAP (%): 52.8 -> 49.5 | Hurt |
| insert RFC after stage 3 instead of stage 2 (ablation) | RFC · Occluded-DukeMTMC (image) | mAP (%): 51.8 -> 52.8 | Improved |
| add keypoint loss L_k (ablation) | RFCnet · Occluded-DukeMTMC (image) | mAP (%): 52.9 -> 54.5 | Improved |
| add L_ap loss (ablation) | RFCnet · Occluded-DukeMTMC-VideoReID | mAP (%): 90.6 -> 92.0 | Improved |
Closest papers by meaning
- Dynamic Feature Pruning and Consolidation for Occluded Person Re-identification (AAAI 2024 · similarity 0.91)
- Occluded Person Re-Identification With Single-Scale Global Representations (ICCV 2021 · similarity 0.91)
- Occlude Them All: Occlusion-Aware Attention Network for Occluded Person Re-ID (ICCV 2021 · similarity 0.91)
- Text-Based Occluded Person Re-identification via Multi-Granularity Contrastive Consistency Learning (AAAI 2024 · similarity 0.86)
- Occluded Person Re-identification via Saliency-Guided Patch Transfer (AAAI 2024 · similarity 0.86)
- From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (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.
