Research map / Label-efficient and robust learning / Recognition and long-tail
CSOR: Coreset Selection for Object Re-identification via Class Pruning
Minyoung Oh, Jae-Young Sim
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 diversity-driven class pruning (DCP) to random sampling (ablation) | Random coreset · Market1501 (budget 1,000 images) | mAP: 54.5 -> 58.0 | Improved |
| add coverage-prioritized sampling (CPS) with easy-to-hard class order (ablation) | Random coreset · Market1501 (budget 500 images) | mAP: 37.0 -> 54.1 | Improved |
| add DCP on top of CPS (ablation) | CPS · Market1501 (budget 1,000 images) | mAP: 64.3 -> 67.1 | Improved |
| add DCP on top of CPS at lowest budget (ablation) | CPS · Market1501 (budget 500 images) | mAP: 54.1 -> 54.8 | No clear effect |
| prioritize easy classes first instead of hard-to-easy (ablation) | CSOR · Market1501 (budget 500 images) | mAP: 44.2 -> 54.8 | Improved |
| use class-prioritized easy-to-hard allocation instead of sample-wise allocation (ablation) | CSOR · Market1501 (budget 500 images) | mAP: 14.8 -> 54.8 | Improved |
| use easy-to-hard class order instead of random class selection (ablation) | CSOR · Market1501 (budget 1,000 images) | mAP: 64.7 -> 67.1 | Improved |
| prune class with low diversity scores at ratio alpha about 0.2 (ablation) | CSOR · Market1501 | mAP | Improved |
| raise termination threshold tau from too small to 0.01 (ablation) | CSOR · Market1501 | mAP | Improved |
| use CSOR coreset instead of AUM coreset (ablation) | SBS · Market1501 (budget 1,000 images) | mAP: 31.8 -> 39.7 | Improved |
| use CSOR coreset instead of EL2N coreset | SOLIDER · Market1501 (budget 1,500 images) | mAP: 72.4 -> 76.0 | Improved |
| use CSOR coreset instead of AUM coreset | Cross-dataset MS->MA · Market1501 (trained on MSMT17, 1,000 images) | mAP: 10.1 -> 35.3 | Improved |
Closest papers by meaning
- Catalyst for Clustering-Based Unsupervised Object Re-identification: Feature Calibration (AAAI 2024 · similarity 0.86)
- Alice Benchmarks: Connecting Real World Re-Identification with the Synthetic (ICLR 2024 · similarity 0.86)
- TOP-ReID: Multi-Spectral Object Re-identification with Token Permutation (AAAI 2024 · similarity 0.85)
- Exploring the Camera Bias of Person Re-identification (ICLR 2025 · similarity 0.85)
- $\mathbb{D}^2$ Pruning: Message Passing for Balancing Diversity & Difficulty in Data Pruning (ICLR 2024 · similarity 0.85)
- 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.
