Research map / Label-efficient and robust learning / Recognition and long-tail
Part-Based Pseudo Label Refinement for Unsupervised Person Re-Identification
Check 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 agreement-aware label smoothing and part-guided label refinement (ablation) | PPLR without camera labels · Market-1501 | mAP: 73.5 -> 81.5 | Helped |
| Add agreement-aware label smoothing and part-guided label refinement (ablation) | PPLR with camera labels · MSMT17 | mAP: 36.8 -> 42.2 | Helped |
| Add agreement-aware label smoothing (ablation) | PPLR without AALS · Market-1501 | mAP: 77.8 -> 81.5 | Helped |
| Replace agreement-aware smoothing with part-to-part label refinement (ablation) | PPLR without AALS · MSMT17 | mAP: 29.3 -> 27.9 | Hurt |
| Replace agreement-aware smoothing with constant-weight vanilla label smoothing (ablation) | PPLR without AALS · Market-1501 | mAP: 77.8 -> 79.1; PPLR: 81.5 | Hurt |
| Add part-guided label refinement (ablation) | PPLR without PGLR · MSMT17 | mAP: 28.8 -> 31.4 | Helped |
| Replace part-guided refinement with mean-teaching (ablation) | PPLR without PGLR · Market-1501 | mAP: 78.9 -> 79.0; PPLR: 81.5 | Hurt |
| Remove cross-agreement weights and average the part predictions (ablation) | PGLR · MSMT17 | mAP: 31.4 -> 30.1 | Hurt |
| Use camera-aware inter-camera contrastive loss (ablation) | PPLR · Market-1501 | mAP: 81.5 -> 84.4 | Helped |
| Set PGLR mixing parameter β to 0, using only ensembled part predictions (ablation) | PPLR · Market-1501 | mAP | Hurt |
| Increase the number of nearest neighbors k substantially (ablation) | PPLR cross-agreement scoring · Market-1501 | mAP | Hurt |
| Increase the number of uniformly partitioned parts (ablation) | Part-based unsupervised re-ID baseline · Market-1501 | mAP | Hurt |
Closest papers by meaning
- Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification (ICLR 2020 · similarity 0.91)
- Unsupervised Person Re-Identification by Soft Multilabel Learning (CVPR 2019 · similarity 0.90)
- Group-aware Label Transfer for Domain Adaptive Person Re-identification (CVPR 2021 · similarity 0.90)
- SECRET: Self-Consistent Pseudo Label Refinement for Unsupervised Domain Adaptive Person Re-identification (AAAI 2022 · similarity 0.90)
- Unsupervised Person Re-Identification via Multi-Label Classification (CVPR 2020 · similarity 0.90)
- Intra-Inter Camera Similarity for Unsupervised 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.