Research map / Label-efficient and robust learning / Recognition and long-tail
Cross-View Asymmetric Metric Learning for Unsupervised Person Re-Identification
Hong-Xing Yu, Ancong Wu, Wei-Shi Zheng
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 asymmetric view-specific projections instead of a symmetric metric (ablation) | CAMEL · VIPeR (single-shot) | rank-1 accuracy: 27.5 -> 30.9 | Improved |
| add asymmetric view-specific projections instead of a symmetric metric (ablation) | CAMEL · CUHK01 (multi-shot) | rank-1 accuracy: 54.9 -> 61.9 | Improved |
| add asymmetric view-specific projections instead of a symmetric metric (ablation) | CAMEL · Market-1501 (multi-shot) | rank-1 accuracy: 47.6 -> 54.5 | Improved |
| add asymmetric view-specific projections instead of a symmetric metric (ablation) | CAMEL · ExMarket (multi-shot) | rank-1 accuracy: 48.7 -> 55.9 | Improved |
| add asymmetric view-specific projections instead of a symmetric metric (ablation) | CAMEL · CUHK03 (single-shot) | rank-1 accuracy: 29.8 -> 31.9 | Improved |
| increase number of clusters K from 500 to 1250 (ablation) | CAMEL · CUHK01 (single-shot) | rank-1 accuracy: 57.35 -> 52.75 | Hurt |
| vary number of clusters K from 250 to 1250 (ablation) | CAMEL · Market-1501 (multi-shot) | rank-1 accuracy: 54.48 -> 54.48 | No clear effect |
| run alternating optimization to convergence (K=500) (ablation) | CAMEL · CUHK01 clustering | rate of clusters with multiple persons: 57.0% -> 34.3% | Improved |
| replace cluster pseudo-labels with true labels (supervised CAMEL_s) | CAMEL · CUHK03 (single-shot) | rank-1 accuracy: 31.9 -> 45.1 | Improved |
| use CAMEL on LOMO features instead of L2 distance | CAMEL · VIPeR (single-shot, LOMO) | rank-1 accuracy: 11.6 -> 26.4 | Improved |
| use CAMEL instead of Dic on JSTL features | CAMEL · CUHK01 (single-shot) | rank-1 accuracy: 49.3 -> 57.3 | Improved |
Closest papers by meaning
- Unsupervised Person Re-Identification by Deep Asymmetric Metric Embedding (TPAMI 2018 · similarity 0.96)
- Stepwise Metric Promotion for Unsupervised Video Person Re-Identification (ICCV 2017 · similarity 0.92)
- Unsupervised Person Re-Identification by Camera-Aware Similarity Consistency Learning (ICCV 2019 · similarity 0.91)
- Person Re-Identification by Camera Correlation Aware Feature Augmentation (TPAMI 2017 · similarity 0.91)
- A Systematic Evaluation and Benchmark for Person Re-Identification: Features, Metrics, and Datasets (TPAMI 2018 · similarity 0.90)
- Asymmetric Co-Teaching for Unsupervised Cross-Domain Person Re-Identification (AAAI 2020 · similarity 0.90)
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.
