Research map / Label-efficient and robust learning / Recognition and long-tail
Efficient Person Search: An Anchor-Free Approach
Yichao Yan, Jinpeng Li, Jie Qin, Peng Zheng, Shengcai Liao, Xiaokang 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 lateral deformable conv, output deformable conv and feature concatenation (ablation) | FCOS-based AlignPS · CUHK-SYSU | mAP: 83.4 -> 93.1 | Improved |
| add lateral deformable conv, output deformable conv and feature concatenation (ablation) | FCOS-based AlignPS · PRW | mAP: 27.4 -> 45.9 | Improved |
| use only the P3 feature level instead of P3,P4,P5 (ablation) | AlignPS · PRW | mAP: 39.5 -> 45.9 | Improved |
| use only the P3 feature level instead of P3,P4,P5 (ablation) | AlignPS · CUHK-SYSU | mAP: 90.0 -> 93.1 | Improved |
| separate re-id head from detection (T3) vs shared with regression head (T1) (ablation) | AlignPS · CUHK-SYSU | mAP: 80.3 -> 80.7 | No clear effect |
| use the AlignPS re-id-first design instead of independent heads T3 (ablation) | AlignPS · PRW | mAP: 35.8 -> 45.9 | Improved |
| replace OIM with TOIM using LUT (ablation) | AlignPS · CUHK-SYSU | mAP: 92.4 -> 93.1 | Improved |
| replace OIM with TOIM using LUT (ablation) | AlignPS · PRW | mAP: 45.7 -> 45.9 | No clear effect |
| add deformable conv in backbone res3-res5 (ablation) | AlignPS · CUHK-SYSU | mAP: 93.1 -> 94.0 | Improved |
| add deformable conv in backbone res3-res5 (ablation) | AlignPS · PRW | mAP: 45.9 -> 46.1 | No clear effect |
| jointly train ROI-Align branch (w/o RPN) before concatenating re-id features (ablation) | ROI-AlignPS · CUHK-SYSU | mAP: 93.0 -> 94.3 | Improved |
| jointly train ROI-Align branch with RPN (ablation) | ROI-AlignPS · CUHK-SYSU | mAP: 94.3 -> 95.0 | Improved |
Closest papers by meaning
- Anchor-Free Person Search (CVPR 2021 · similarity 0.95)
- Person Search by Multi-Scale Matching (ECCV 2018 · similarity 0.89)
- OIMNet++: Prototypical Normalization and Localization-Aware Learning for Person Search (ECCV 2022 · similarity 0.88)
- Diverse Knowledge Distillation for End-to-End Person Search (AAAI 2021 · similarity 0.88)
- Cheb-GR: Rethinking K-nearest Neighbor Search in Re-ranking for Person Re-identification (CVPR 2025 · similarity 0.87)
- Divide and Conquer: Hybrid Pre-training for Person Search (AAAI 2024 · similarity 0.87)
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.
