Research map / Label-efficient and robust learning / Recognition and long-tail
Deep Metric Learning Beyond Binary Supervision
Check 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 |
|---|---|---|---|
| Use a caption-aware visual feature instead of an ImageNet feature | Att2all2 captioning model · Karpathy test split, RL | CIDEr-D: 1.131 -> 1.159 | Helped |
| Use a caption-aware visual feature instead of an ImageNet feature | Topdown captioning model · Karpathy test split, RL | CIDEr-D: 1.201 -> 1.213 | Helped |
| Replace the log-ratio loss with triplet loss while retaining dense mining (ablation) | Continuous-label image retrieval · MPII human pose retrieval | Retrieval performance | Hurt |
| Replace dense mining with binary nearest-neighbor mining while retaining triplet loss (ablation) | Continuous-label image retrieval · MPII human pose retrieval | Retrieval performance | Hurt |
| Replace the log-ratio loss with triplet loss while retaining dense mining (ablation) | Continuous-label image retrieval · LSUN room layout retrieval | Retrieval performance | Hurt |
| Replace dense mining with binary nearest-neighbor mining while retaining triplet loss (ablation) | Continuous-label image retrieval · LSUN room layout retrieval | Retrieval performance | No clear effect |
| Replace the log-ratio loss with triplet loss while retaining dense mining (ablation) | Continuous-label image retrieval · MS-COCO caption-aware retrieval | Retrieval performance | Hurt |
| Replace dense mining with binary nearest-neighbor mining while retaining triplet loss (ablation) | Continuous-label image retrieval · MS-COCO caption-aware retrieval | Retrieval performance | Hurt |
| Reduce the embedding dimension from 128-D to 16-D (ablation) | Log-ratio model · MPII human pose retrieval | Retrieval performance | Hurt |
| Reduce the embedding dimension from 128-D to 16-D (ablation) | Triplet loss with dense mining · MPII human pose retrieval | Retrieval performance | Hurt |
Closest papers by meaning
- Divide and Conquer the Embedding Space for Metric Learning (CVPR 2019 · similarity 0.89)
- Supervised Metric Learning to Rank for Retrieval via Contextual Similarity Optimization (ICML 2023 · similarity 0.89)
- Deep Compositional Metric Learning (CVPR 2021 · similarity 0.87)
- Deep Metric Learning via Facility Location (CVPR 2017 · similarity 0.87)
- Ranked List Loss for Deep Metric Learning (CVPR 2019 · similarity 0.87)
- Implicit Relative Labeling-Importance Aware Multi-Label 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.