Research map / Label-efficient and robust learning / Domain adaptation and generalization
Unsupervised Tracklet Person Re-Identification
Minxian Li, Xiatian Zhu, Shaogang Gong
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 per-camera tracklet discrimination (multi-task) instead of single-task classification (ablation) | UTAL · MARS | Rank-1: 15.9 -> 43.8 | Improved |
| add per-camera tracklet discrimination instead of single-task classification (ablation) | UTAL · DukeTracklet | Rank-1: 3.9 -> 31.7 | Improved |
| use soft labels instead of hard one-hot labels (ablation) | UTAL · MARS | Rank-1: 35.5 -> 49.9 | Improved |
| use soft labels instead of hard one-hot labels (ablation) | UTAL · DukeTracklet | Rank-1: 24.5 -> 43.8 | Improved |
| soft labels vs hard labels with one tracklet per ID per camera (cross-ID transfer) (ablation) | UTAL · MARS | Rank-1: 45.1 -> 46.5 | Improved |
| add cross-camera tracklet association learning (ablation) | UTAL · MARS | Rank-1: 43.8 -> 49.9 | Improved |
| add cross-camera tracklet association learning (ablation) | UTAL · DukeTracklet | Rank-1: 31.7 -> 43.8 | Improved |
| reciprocal 2-way 1-NN instead of 1-way 1-NN (ablation) | UTAL · MARS | Rank-1 / mAP: 50.5 -> 49.9 Rank-1; 33.2 -> 35.2 mAP | Mixed |
| use 10 neighbours instead of 1 in 2-way matching (ablation) | UTAL · DukeTracklet | Rank-1: 43.8 -> 33.9 | Hurt |
| use all tracklets instead of SSTT sampling (ablation) | UTAL · MARS | Rank-1: 45.1 -> 49.9 | Improved |
| use all tracklets instead of SSTT sampling (ablation) | UTAL · DukeTracklet | Rank-1: 31.7 -> 43.8 | Improved |
| replace ResNet-50 backbone with DenseNet-121 (ablation) | UTAL · MARS | Rank-1: 49.9 -> 51.6 | Improved |
Closest papers by meaning
- Unsupervised Person Re-identification by Deep Learning Tracklet Association (ECCV 2018 · similarity 0.97)
- Tracklet Self-Supervised Learning for Unsupervised Person Re-Identification (AAAI 2020 · similarity 0.95)
- Stepwise Metric Promotion for Unsupervised Video Person Re-Identification (ICCV 2017 · similarity 0.93)
- Exploit the Unknown Gradually: One-Shot Video-Based Person Re-Identification by Stepwise Learning (CVPR 2018 · similarity 0.92)
- Weakly Supervised Tracklet Association Learning With Video Labels for Person Re-Identification (TPAMI 2023 · similarity 0.92)
- Unsupervised Graph Association for Person Re-Identification (ICCV 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.
