Research map / Label-efficient and robust learning / Recognition and long-tail
It Takes Two to Tango: Mixup for Deep Metric Learning
Shashanka Venkataramanan, Bill Psomas, Ewa Kijak, laurent amsaleg, Konstantinos Karantzalos, Yannis Avrithis
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 Metrix feature mixup with interpolated labels | Multi-Similarity · CUB200 | Recall@1: 67.8 -> 71.4 | Improved |
| Add Metrix feature mixup with interpolated labels | Multi-Similarity · Cars196 | Recall@1: 87.8 -> 89.6 | Improved |
| Add Metrix feature mixup with interpolated labels | Multi-Similarity · SOP | Recall@1: 76.9 -> 81.0 | Improved |
| Add Metrix feature mixup with interpolated labels | Multi-Similarity · In-Shop | Recall@1: 90.1 -> 92.2 | Improved |
| Use feature-level mixup instead of input or embedding mixup (ablation) | Contrastive loss · Cars196 | Recall@1: 84.7 vs 85.1 | Improved |
| Use all pairs rather than 20 or 40 hard negatives for feature mixup (ablation) | Metrix feature mixup · Cars196 | Recall@1: 83.5, 84.0 -> 85.1 | Improved |
| Increase input mixup hard negatives from k=1 to k=3 (ablation) | Metrix input mixup · Cars196 | Recall@1: 82.0 -> 82.9 | Improved |
| Mix positive-negative pairs instead of positive-positive pairs (ablation) | Metrix feature mixup · Cars196 | Recall@1: 81.1 -> 84.0 | Improved |
| Mix anchor-negative pairs instead of positive-positive pairs (ablation) | Metrix feature mixup · Cars196 | Recall@1: 81.1 -> 83.7 | Improved |
| Mix positive-positive pairs only (ablation) | Contrastive loss · Cars196 | Recall@1: 81.6 -> 81.1 | Hurt |
| Combine input, feature, and embedding mixup (ablation) | Metrix · Cars196 | Recall@1: 85.1 -> 85.3 | Improved |
| Combine feature and embedding mixup (ablation) | Metrix · Cars196 | Recall@1: 85.1 -> 84.7 | Hurt |
Closest papers by meaning
- Sharing Matters for Generalization in Deep Metric Learning (TPAMI 2020 · similarity 0.88)
- Improving Deep Metric Learning by Divide and Conquer (TPAMI 2021 · similarity 0.88)
- Nonlinear Mixup: Out-Of-Manifold Data Augmentation for Text Classification (AAAI 2020 · similarity 0.87)
- Divide and Conquer the Embedding Space for Metric Learning (CVPR 2019 · similarity 0.87)
- Embedding Transfer With Label Relaxation for Improved Metric Learning (CVPR 2021 · similarity 0.87)
- Potential Field Based Deep Metric Learning (CVPR 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.
