Research map / Label-efficient and robust learning / Recognition and long-tail
Deep Metric Learning by Online Soft Mining and Class-Aware Attention
Read the paperCheck 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 OSM soft pair weighting, scoring local positives and difficult negatives (ablation) | Weighted contrastive learning · CUB-200-2011 | Recall@1 (%): 52.0 -> 54.2 | Helped |
| Add OSM soft pair weighting, scoring local positives and difficult negatives (ablation) | Weighted contrastive learning · LPW | CMC-1 (%): 61.9 -> 69.3 | Helped |
| Add OSM soft pair weighting, scoring local positives and difficult negatives (ablation) | Weighted contrastive learning · MARS | CMC-1 (%): 79.1 -> 82.3 | Helped |
| Add CAA attention to downweight outlier-like pairs (ablation) | OSM · CUB-200-2011 | Recall@1 (%): 54.2 -> 55.3 | Helped |
| Add CAA attention to downweight outlier-like pairs (ablation) | OSM · LPW | CMC-1 (%): 69.3 -> 71.7 | Helped |
| Add CAA attention to downweight outlier-like pairs (ablation) | OSM · MARS | CMC-1 (%): 82.3 -> 84.7 | Helped |
| Use OSM+CAA instead of HDC* | Deep metric learning on CARS196 raw images · CARS196 raw images | Recall@1 (%): 73.7 -> 74.0 | Helped |
| Use OSM+CAA instead of HDC* | Deep metric learning on CARS196 cropped images · CARS196 cropped images | Recall@1 (%): 83.8 -> 85.5 | Helped |
| Use OSM+CAA instead of HDC* | Deep metric learning on CUB-200-2011 raw images · CUB-200-2011 raw images | Recall@1 (%): 53.6 -> 55.3 | Helped |
| Use OSM+CAA instead of HDC* | Deep metric learning on CUB-200-2011 cropped images · CUB-200-2011 cropped images | Recall@1 (%): 60.7 -> 62.3 | Helped |
| Use OSM+CAA instead of CAE | Video person re-identification · MARS | CMC-1 (%): 82.4 -> 84.7 | Helped |
Closest papers by meaning
- Stochastic Class-Based Hard Example Mining for Deep Metric Learning (CVPR 2019 · similarity 0.90)
- Deep Metric Learning via Lifted Structured Feature Embedding (CVPR 2016 · similarity 0.88)
- Divide and Conquer the Embedding Space for Metric Learning (CVPR 2019 · similarity 0.87)
- Learning Low-Dimensional Metrics (NeurIPS 2017 · similarity 0.87)
- Deep Metric Learning with Self-Supervised Ranking (AAAI 2021 · similarity 0.87)
- Deep Disentangled Metric Learning (AAAI 2025 · similarity 0.86)
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.