Research map / Label-efficient and robust learning / Recognition and long-tail
Implicit Sample Extension for Unsupervised Person Re-Identification
Xinyu Zhang, Dongdong Li, Zhigang Wang, Jian Wang, Errui Ding, Javen Qinfeng Shi, Zhaoxiang Zhang, Jingdong Wang
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 |
|---|---|---|---|
| Add progressive linear interpolation support samples (ablation) | Cluster-Contrast baseline · Market-1501 | mAP: 82.5 -> 83.9 | Improved |
| Add progressive linear interpolation support samples (ablation) | Cluster-Contrast baseline · MSMT17 | mAP: 30.1 -> 33.5 | Improved |
| Add label-preserving loss on support samples (ablation) | ISE with PLI · Market-1501 | mAP: 83.9% to 84.7% | Improved |
| Add label-preserving loss on support samples (ablation) | ISE with PLI · MSMT17 | mAP: 33.5% to 35.0% | Improved |
| Add logarithmically progressive rather than constant support-sample degree (ablation) | ISE with PLI and LP · Market-1501 | mAP: 83.3 -> 84.7 | Improved |
| Add logarithmically progressive rather than constant support-sample degree (ablation) | ISE with PLI and LP · MSMT17 | mAP: 33.8 -> 35.0 | Improved |
| Use logarithmic rather than square degree schedule (ablation) | ISE · MSMT17 | mAP: 32.4 -> 35.0 | Improved |
| Use nearest-cluster rather than random direction (ablation) | ISE · MSMT17 | mAP: 28.3 -> 35.0 | Improved |
| Use nearest-cluster rather than farthest direction (ablation) | ISE · MSMT17 | mAP: 27.8 -> 35.0 | Improved |
| Increase nearest clusters from K=1 to K=5 (ablation) | ISE · Market-1501 | mAP: 84.7 -> 84.8 | Improved |
| Increase nearest clusters from K=1 to K=5 (ablation) | ISE · MSMT17 | mAP: 35.0 -> 31.9 | Hurt |
| Add GeM pooling | ISE · MSMT17 | mAP: 35.0 -> 37.0 | Improved |
Closest papers by meaning
- Asymmetric Co-Teaching for Unsupervised Cross-Domain Person Re-Identification (AAAI 2020 · similarity 0.90)
- Self-Similarity Grouping: A Simple Unsupervised Cross Domain Adaptation Approach for Person Re-Identification (ICCV 2019 · similarity 0.90)
- A Bottom-Up Clustering Approach to Unsupervised Person Re-Identification (AAAI 2019 · similarity 0.89)
- Camera-Aware Proxies for Unsupervised Person Re-Identification (AAAI 2021 · similarity 0.89)
- Unsupervised Pre-Training for Person Re-Identification (CVPR 2021 · similarity 0.89)
- Part-Based Pseudo Label Refinement for Unsupervised Person Re-Identification (CVPR 2022 · 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.
