Research map / Detection and segmentation / CNN features
Towards Robustness of Person Search against Corruptions
Woojung Son, Yoonki Cho, Guoyuan An, Chanmi Lee, Sung-Eui Yoon
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 robust proposal regularizer (ablation) | SeqNet · CUHK-SYSU-C | mAP: 41.6 -> 53.7 | Improved |
| add foreground-aware augmentation on top of regularizer (ablation) | SeqNet · CUHK-SYSU-C | mAP: 53.7 -> 63.1 | Improved |
| add IoU score weighting to regularizer (ablation) | SeqNet · CUHK-SYSU-C | mAP: 63.1 -> 65.0 | Improved |
| add robust proposal regularizer (ablation) | SeqNet · CUHK-SYSU (clean) | mAP: 93.8 -> 93.4 | Hurt |
| apply foreground-aware augmentation and regularizer to six person search models | OIMNet, NAE, OIMNet++, SeqNet, COAT, SAT · CUHK-SYSU-C | mAP: 41.6 -> 65.0 (SeqNet) | Improved |
| apply method to six person search models | OIMNet, NAE, OIMNet++, SeqNet, COAT, SAT · PRW-C | mAP: 10.5 -> 15.9 (SeqNet) | Improved |
| apply method to COAT on clean data | COAT · CUHK-SYSU (clean) | R@1: 94.7 -> 92.8 | Hurt |
| apply method to SAT on real rain scenario benchmark | SeqNet · BDD100K rain | R@1: 28.3 -> 43.5 | Improved |
| apply method to SeqNet on real fog scenario | SeqNet · BDD100K fog | mAP: 42.8 -> 46.2 | Improved |
| train on CUHK-SYSU-C instead of clean CUHK-SYSU | SeqNet · BDD100K rain | R@1: 28.3 -> 38.4 | Improved |
| combine OADG robust detector and CIL robust reID | OADG+CIL · CUHK-SYSU-C | mAP: 46.0 vs 44.4 (SeqNet) | No clear effect |
| corrupt person representation versus detection at severity level 5 | SeqNet · CUHK-SYSU-C | R@1: 19.3 vs 69.8 | Hurt |
Closest papers by meaning
- Benchmarks for Corruption Invariant Person Re-identification (NeurIPS 2021 · similarity 0.89)
- Divide and Conquer: Hybrid Pre-training for Person Search (AAAI 2024 · similarity 0.83)
- Person De-reidentification: A Variation-guided Identity Shift Modeling (CVPR 2025 · similarity 0.83)
- Diverse Person: Customize Your Own Dataset for Text-Based Person Search (AAAI 2024 · similarity 0.83)
- Cheb-GR: Rethinking K-nearest Neighbor Search in Re-ranking for Person Re-identification (CVPR 2025 · similarity 0.83)
- Exploring the Camera Bias of Person Re-identification (ICLR 2025 · similarity 0.82)
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.
