Research map / Label-efficient and robust learning / Knowledge distillation
DarkRank: Accelerating Deep Metric Learning via Cross Sample Similarities Transfer
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 |
|---|---|---|---|
| Transfer teacher cross-sample rankings with hard DarkRank | NIN-BN student · Market1501 single query | mAP: 58.1 -> 63.5 | Helped |
| Transfer teacher cross-sample rankings with soft DarkRank | NIN-BN student · Market1501 multiple query | Rank-1: 86.7 -> 88.8 | Helped |
| Combine KD with hard DarkRank | NIN-BN student · Market1501 single query | mAP: 66.7 -> 68.5 | Helped |
| Combine KD with soft DarkRank | NIN-BN student · Market1501 multiple query | Rank-1: 90.4 -> 91.4 | Helped |
| Directly match teacher and student distances rather than ranking them | NIN-BN student · Market1501 single query | Rank-1: 80.3 -> 80.3 | No clear effect |
| Add DarkRank to FitNet embedding matching | FitNet · Market1501 single query | mAP: 64.0 -> 67.3 | Helped |
| Transfer cross-sample rankings | NIN-BN student · CUB-200-2011 test species | F_1: 0.153 -> 0.168 | Helped |
| Use hard transfer instead of soft transfer (ablation) | DarkRank · Market1501 | mAP: 63.1 -> 63.5; 71.4 -> 71.2 | Mixed |
| Set contrast parameter β to 3.0 (ablation) | Soft DarkRank · CUHK03 validation set | Rank-1 | Helped |
| Set scaling factor α to 3.0 (ablation) | Soft DarkRank · CUHK03 validation set | Rank-1 | Helped |
| Set transfer-loss weight λ to 2.0 (ablation) | Soft DarkRank · CUHK03 validation set | Rank-1 | Helped |
| Vary transfer-loss weight λ across a large range (ablation) | Soft DarkRank · CUHK03 validation set | Rank-1 | No clear effect |
Closest papers by meaning
- Learning Intra-Batch Connections for Deep Metric Learning (ICML 2021 · similarity 0.88)
- Revisiting Training Strategies and Generalization Performance in Deep Metric Learning (ICML 2020 · similarity 0.88)
- Ranked List Loss for Deep Metric Learning (CVPR 2019 · similarity 0.88)
- Deep Metric Learning to Rank (CVPR 2019 · similarity 0.87)
- Divide and Conquer the Embedding Space for Metric Learning (CVPR 2019 · similarity 0.87)
- Denoising Multi-Similarity Formulation: A Self-Paced Curriculum-Driven Approach for Robust Metric Learning (AAAI 2023 · 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.