Research map / Label-efficient and robust learning / Recognition and long-tail
Weakly Supervised Person Search With Region Siamese Networks
Chuchu Han, Kai Su, Dongdong Yu, Zehuan Yuan, Changxin Gao, Nong Sang, Yi Yang, Changhu Wang
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 |
|---|---|---|---|
| train on cropped instance path instead of scene path (ablation) | R-SiamNet baseline · CUHK-SYSU | mAP (%): 51.85 -> 63.79 | Improved |
| add self-instance consistency loss L_ins between two paths (ablation) | R-SiamNet · CUHK-SYSU | mAP (%): 51.85 -> 76.06 | Improved |
| add inter-instance similarity consistency loss L_int (ablation) | R-SiamNet · CUHK-SYSU | rank-1 (%): 78.21 -> 82.17 | Improved |
| replace instance recognition with cluster-level contrastive loss L_clu (ablation) | R-SiamNet · CUHK-SYSU | mAP (%): 79.62 -> 85.72 | Improved |
| train with more data (20% to 100%) (ablation) | R-SiamNet · CUHK-SYSU | mAP (%): 76.71 -> 85.72 | Improved |
| add PRW training data (ablation) | R-SiamNet · PRW | mAP (%): 21.16 -> 23.50 | Improved |
| add INRIA pedestrian detection data (ablation) | R-SiamNet · PRW | mAP (%): 23.50 -> 25.53 | Improved |
| use both paths at inference instead of search path only (ablation) | R-SiamNet · CUHK-SYSU | mAP (%): 85.72 -> 85.65 | No clear effect |
| swap RepPoints detector for Faster R-CNN (ablation) | R-SiamNet · CUHK-SYSU | mAP (%): 85.72 -> 84.84 | No clear effect |
| swap RepPoints detector for RetinaNet (ablation) | R-SiamNet · CUHK-SYSU | mAP (%): 85.72 -> 85.39 | No clear effect |
| add filter forbidding clustering of persons from the same scene image (ablation) | R-SiamNet · CUHK-SYSU | rank-1 (%): 85.64 -> 86.86 | Improved |
| add filter forbidding clustering of persons from the same scene image (ablation) | R-SiamNet · PRW | rank-1 (%): 72.23 -> 73.36 | Improved |
Closest papers by meaning
- Exploring Visual Context for Weakly Supervised Person Search (AAAI 2022 · similarity 0.88)
- Self-similarity Driven Scale-invariant Learning for Weakly Supervised Person Search (ICCV 2023 · similarity 0.88)
- Re-Identification With Consistent Attentive Siamese Networks (CVPR 2019 · similarity 0.88)
- Optimizing Label Assignment for Weakly Supervised Person Search (AAAI 2025 · similarity 0.86)
- Divide and Conquer: Hybrid Pre-training for Person Search (AAAI 2024 · similarity 0.85)
- Domain Generalizable Person Search Using Unreal Dataset (AAAI 2024 · similarity 0.84)
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.
