Research map / Label-efficient and robust learning / Recognition and long-tail
Intra-Camera Supervised Person Re-Identification
Xiangping Zhu, Xiatian Zhu, Minxian Li, Pietro Morerio, Vittorio Murino, Shaogang Gong
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 cross-camera multi-label learning (CCML) (ablation) | PCMT · Market-1501 (ICS) | Rank-1 / mAP (%): 78.4 -> 85.3 Rank-1, 52.1 -> 65.2 mAP | Improved |
| add curriculum thresholding (CT) (ablation) | PCMT+CCML · Market-1501 (ICS) | Rank-1 / mAP (%): 85.3 -> 88.7 Rank-1, 65.2 -> 71.1 mAP | Improved |
| add cross-camera association on top of per-camera multi-task learning (full MATE) | PCMT · Market-1501 (ICS) | Rank-1 / mAP (%): 78.4 -> 88.7 Rank-1, 52.1 -> 71.1 mAP | Improved |
| use per-camera multi-task branches instead of merged label space | MCST · Market-1501 (ICS) | Rank-1 (%): 34.9 -> 78.4 | Improved |
| train separate per-camera models instead of merged label space (EPCS vs MCST) | MCST · Market-1501 (ICS) | Rank-1 (%): 34.9 -> 42.6 | Improved |
| enlarge cyclic association from 2 to 3 camera views (ablation) | MATE w/o CT · Market-1501 (ICS) | Rank-1 / mAP (%): 85.3 -> 83.5 Rank-1, 65.2 -> 64.2 mAP | Hurt |
| enlarge cyclic association from 2 to 4 camera views (ablation) | MATE w/o CT · Market-1501 (ICS) | Rank-1 / mAP (%): 85.3 -> 80.7 Rank-1, 65.2 -> 58.9 mAP | Hurt |
| explicitly exploit transitive association (ablation) | MATE · Market-1501 (ICS) | Rank-1 / mAP (%): 88.7 -> 88.9 Rank-1, 71.1 -> 71.2 mAP | No clear effect |
| intra-camera labelling instead of cross-camera labelling | annotation process · MSMT17 subset (714 identities) | person-days: 8 -> 2.5 person-days | Improved |
| vary loss weight lambda and curriculum bounds (ablation) | MATE · Market-1501 (ICS) | Rank-1 / mAP (%) | No clear effect |
Closest papers by meaning
- Intra-Inter Camera Similarity for Unsupervised Person Re-Identification (CVPR 2021 · similarity 0.93)
- Intra-Inter Domain Similarity for Unsupervised Person Re-Identification (TPAMI 2022 · similarity 0.93)
- Deep Learning for Person Re-Identification: A Survey and Outlook (TPAMI 2021 · similarity 0.92)
- Weakly Supervised Person Re-Identification (CVPR 2019 · similarity 0.91)
- Camera-Aware Proxies for Unsupervised Person Re-Identification (AAAI 2021 · similarity 0.91)
- Unsupervised Person Re-Identification by Soft Multilabel Learning (CVPR 2019 · similarity 0.91)
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.
