Research map / Label-efficient and robust learning / Recognition and long-tail
Adversarially Occluded Samples for Person Re-Identification
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 adversarially selected occluded samples using Sampling | ResNet-50 ReID model · Market1501 | Rank-1: 84.42 -> 86.49 | Helped |
| Add adversarially selected occluded samples using Sampling | ResNet-50 ReID model · Market1501 | mAP: 66.95 -> 70.43 | Helped |
| Add adversarially selected occluded samples using Sampling | ResNet-50 ReID model · CUHK03 detected | Rank-1: 40.23 -> 47.14 | Helped |
| Add adversarially selected occluded samples using Sampling | ResNet-50 ReID model · CUHK03 detected | mAP: 37.00 -> 43.33 | Helped |
| Add adversarially selected occluded samples using Sampling | ResNet-50 ReID model · DukeMTMC-reID | Rank-1: 75.60 -> 79.17 | Helped |
| Add adversarially selected occluded samples using Sampling | ResNet-50 ReID model · DukeMTMC-reID | mAP: 57.34 -> 62.10 | Helped |
| Replace adversarial selection with selecting the easiest occluded sample (ablation) | ReID model retraining · Market1501 | Rank-1: 84.42 -> 84.01 | Hurt |
| Replace adversarial selection with selecting the easiest occluded sample (ablation) | ReID model retraining · CUHK03 detected | Rank-1: 40.23 -> 39.41 | Hurt |
| Replace adversarial selection with selecting the easiest occluded sample (ablation) | ReID model retraining · DukeMTMC-reID | Rank-1: 75.60 -> 76.39 | Helped |
| Choose occlusion locations randomly rather than by model sensitivity (ablation) | ReID model retraining · Market1501 | Rank-1: 84.42 -> 86.16 | Helped |
| Choose occlusion locations randomly rather than by model sensitivity (ablation) | ReID model retraining · DukeMTMC-reID | Rank-1: 75.60 -> 78.19 | Helped |
| Always select the hardest occlusion instead of sampling among influential locations (ablation) | ReID model retraining · DukeMTMC-reID | Rank-1: 75.60 -> 77.33 | Helped |
Closest papers by meaning
- Exploring the Camera Bias of Person Re-identification (ICLR 2025 · similarity 0.91)
- Learning Disentangled Representation for Robust Person Re-identification (NeurIPS 2019 · similarity 0.91)
- From Poses to Identity: Training-Free Person Re-Identification via Feature Centralization (CVPR 2025 · similarity 0.91)
- Pose Transferrable Person Re-Identification (CVPR 2018 · similarity 0.90)
- Dynamic Feature Pruning and Consolidation for Occluded Person Re-identification (AAAI 2024 · similarity 0.90)
- Person De-reidentification: A Variation-guided Identity Shift Modeling (CVPR 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.