Research map / Label-efficient and robust learning / Domain adaptation and generalization
Tracklet Self-Supervised Learning for Unsupervised 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 tracklet frame coherence loss Lf to cluster loss (ablation) | TSSL (Lc) · Market-1501 | mAP: 35.1 -> 42.5 | Helped |
| add tracklet frame coherence loss Lf (ablation) | TSSL (Lc) · Market-1501 | rank-1: 65.8 -> 70.7 | Helped |
| add neighbourhood compactness loss Ln (ablation) | TSSL (Lf+Lc) · Market-1501 | mAP: 42.5 -> 43.3 | Helped |
| add neighbourhood compactness loss Ln (ablation) | TSSL (Lf+Lc) · Market-1501 | rank-1: 70.7 -> 71.2 | Helped |
| use distribution-aware cluster pairwise distance (ablation) | TSSL (Lc only) · Market-1501 | mAP: +3.2% mAP | Helped |
| use TSSL positive surrogate design vs single-frame, mean, two-stream variants (ablation) | TSSL · Market-1501 | mAP/rank-1 | Helped |
| increase lambda from 0.1 to 0.3 (ablation) | TSSL · Market-1501 | rank-1 | Hurt |
| use more than 4 neighbour clusters (Nk > 4) (ablation) | TSSL · Market-1501 | rank-1 | Hurt |
| TSSL vs BUC (no labels) | unsupervised re-id · Market-1501 | mAP: 38.3 -> 43.3 | Helped |
| TSSL vs BUC | unsupervised re-id · MARS | mAP: 38.0 -> 30.5 | Hurt |
| TSSL vs TAUDL (uses camera labels) | unsupervised re-id · DukeMTMC-ReID | mAP: 43.5 -> 38.5 | Hurt |
Closest papers by meaning
- Exploit the Unknown Gradually: One-Shot Video-Based Person Re-Identification by Stepwise Learning (CVPR 2018 · similarity 0.92)
- Camera-Aware Proxies for Unsupervised Person Re-Identification (AAAI 2021 · similarity 0.90)
- Unsupervised Person Re-Identification by Soft Multilabel Learning (CVPR 2019 · similarity 0.90)
- Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification (ICLR 2020 · similarity 0.89)
- Unsupervised Person Re-Identification via Softened Similarity Learning (CVPR 2020 · similarity 0.89)
- Unsupervised Pre-Training for Person Re-Identification (CVPR 2021 · similarity 0.89)
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.