Research map / Label-efficient and robust learning / Domain adaptation and generalization
Unsupervised Graph Association for Person Re-Identification
Jinlin Wu, Yang Yang, Hao Liu, Shengcai Liao, Zhen Lei, Stan Z. Li
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 embedding block (BN-Dropout-FC-BN) after backbone (ablation) | intra-camera stage · Market-1501 | mAP: 27.9 -> 54.8 | Improved |
| add BN layer after backbone (ablation) | intra-camera stage · Market-1501 | mAP: 27.9 -> 55.6 | Improved |
| use embedding block instead of plain BN (ablation) | intra-camera stage · Market-1501 and MARS | mAP: Market-1501 55.6 -> 54.8; MARS 33.2 -> 35.1 | Mixed |
| add self-camera constraint to cross-camera loss (ablation) | UGA · Market-1501 | mAP: 65.8 -> 71.0 | Improved |
| add self-camera constraint to cross-camera loss (ablation) | UGA · MARS | mAP: 36.7 -> 40.5 | Improved |
| add inter-camera learning stage after intra-camera stage (ablation) | UGA · Market-1501 | Rank-1: 77.5 -> 87.2 | Improved |
| add inter-camera learning stage after intra-camera stage (ablation) | UGA · DukeMTMC-ReID | mAP: 52.5 -> 53.3 | No clear effect |
| add inter-camera learning stage after intra-camera stage (ablation) | UGA · MARS | Rank-1: 55.1 -> 58.1 | Improved |
| add symmetric constraint on cross-view graph edges (ablation) | CVG · graph edge precision | precision | Improved |
| vary threshold lambda between 0.55 and 0.75 (ablation) | UGA · seven RE-ID datasets | Rank-1: lambda 0.65 best, 0.75 worst | Mixed |
| add ID duplication noise (20% of persons) (ablation) | intra-camera stage · MARS | Rank-1: 55.1 -> 54.1 | Hurt |
| add ID duplication noise (50% of persons) (ablation) | intra-camera stage · MARS | Rank-1: 55.1 -> 51.4 | Hurt |
Closest papers by meaning
- Unsupervised Tracklet Person Re-Identification (TPAMI 2019 · similarity 0.91)
- Unsupervised Person Re-identification by Deep Learning Tracklet Association (ECCV 2018 · similarity 0.90)
- Tracklet Self-Supervised Learning for Unsupervised Person Re-Identification (AAAI 2020 · similarity 0.90)
- Stepwise Metric Promotion for Unsupervised Video Person Re-Identification (ICCV 2017 · similarity 0.90)
- Cross-View Asymmetric Metric Learning for Unsupervised Person Re-Identification (ICCV 2017 · similarity 0.89)
- Weakly Supervised Tracklet Association Learning With Video Labels for Person Re-Identification (TPAMI 2023 · 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.
