Research map / Label-efficient and robust learning / Recognition and long-tail
Patch-Based Discriminative Feature Learning for Unsupervised Person Re-Identification
Qize Yang, Hong-Xing Yu, Ancong Wu, Wei-Shi Zheng
Check 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 |
|---|---|---|---|
| Learn patch features with PEDAL rather than using the PatchNet baseline alone (ablation) | PAUL · Market-1501 | Rank-1: 59.3 -> 66.2 | Improved |
| Learn patch features with PEDAL rather than using the PatchNet baseline alone (ablation) | PAUL · DukeMTMC-reID | mAP: 45.6 -> 52.1 | Improved |
| Add image-level IPFL to the PatchNet baseline (ablation) | PAUL · Market-1501 | Rank-1: 59.3 -> 65.4 | Improved |
| Add image-level IPFL to the PatchNet baseline (ablation) | PAUL · DukeMTMC-reID | Rank-1: 65.7 -> 67.1 | Improved |
| Use both PEDAL and IPFL rather than either loss alone (ablation) | PAUL · Market-1501 | mAP: 38.0 -> 40.1 | Improved |
| Use both PEDAL and IPFL rather than either loss alone (ablation) | PAUL · DukeMTMC-reID | mAP: 52.1 -> 53.2 | Improved |
| Learn patch-level features rather than use image-level ResNet-50 features (ablation) | PAUL · Market-1501 | Rank-1: 46.6 -> 59.3 | Improved |
| Apply PEDAL to patches rather than pull similar whole-image features together (ablation) | PAUL · Market-1501 | Rank-1: 25.6 -> 66.2 | Improved |
| Use learnable PGN patch selection rather than random patch selection (ablation) | PAUL · Market-1501 | Rank-1: 24.1 -> 68.5 | Improved |
| Use learnable PGN patch selection rather than equal horizontal stripes (ablation) | PAUL · DukeMTMC-reID | Rank-1: 56.2 -> 72.0 | Improved |
| Use 6 patches instead of 4 (ablation) | PAUL · Market-1501 | Rank-1: 67.3 -> 68.5 | Improved |
| Use 8 patches instead of 6 (ablation) | PAUL · Market-1501 | mAP: 40.1 -> 37.2 | Hurt |
Closest papers by meaning
- Towards Discriminative Representation Learning for Unsupervised Person Re-Identification (ICCV 2021 · similarity 0.89)
- Deeply Unsupervised Patch Re-Identification for Pre-Training Object Detectors (TPAMI 2022 · similarity 0.89)
- Unsupervised Person Re-Identification via Softened Similarity Learning (CVPR 2020 · similarity 0.88)
- Unsupervised Person Re-Identification by Soft Multilabel Learning (CVPR 2019 · similarity 0.88)
- Weakly Supervised Discriminative Feature Learning With State Information for Person Identification (CVPR 2020 · similarity 0.88)
- Part-Based Pseudo Label Refinement for Unsupervised Person Re-Identification (CVPR 2022 · similarity 0.87)
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.
