Research map / Label-efficient and robust learning / Recognition and long-tail
Symmetrical Synthesis for Deep Metric Learning
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 symmetrical synthesis to N-pair loss | N-pair · SOP | Recall@1: 66.4 -> 73.2 | Helped |
| add symmetrical synthesis to triplet loss | Triplet · CUB200-2011 | Recall@1: 35.9 -> 51.4 | Helped |
| add symmetrical synthesis to lifted structure loss | Lifted-Struct · CARS196 | Recall@1: 59.9 -> 66.6 | Helped |
| add symmetrical synthesis to angular loss | Angular · CUB200-2011 | Recall@1: 53.6 -> 54.9 | Helped |
| add symmetrical synthesis to N-pair loss | N-pair · CARS196 | Recall@1: 68.9 -> 76.5 | Helped |
| use symmetrical synthesis instead of HDML generation (N-pair) | N-pair · CARS196 | Recall@1: 79.1 -> 76.5 | Hurt |
| use symmetrical synthesis instead of HDML generation (N-pair) | N-pair · SOP | Recall@1: 68.7 -> 73.2 | Helped |
| change synthetic point cosine similarity/norm away from symmetry (alpha 1.5 or 2.5, beta 1.5) (ablation) | Symm + N-pair · CARS196 | Recall@1 | Hurt |
| reduce synthetic point norm (beta 0.5) (ablation) | Symm + N-pair · CARS196 | Recall@1 | Hurt |
| select harder top-k negative pairs (ablation) | Symm + N-pair · CARS196 | Recall@1 | Helped |
| add symmetrical synthesis to semi-hard triplet | Triplet-semihard · CARS196 | Recall@1: 53.2 -> 68.5 | Helped |
| add symmetrical synthesis to N-pair | N-pair · CARS196 | training time per batch: 8.852e-1 s -> 8.866e-1 s | No clear effect |
Closest papers by meaning
- Proxy Synthesis: Learning with Synthetic Classes for Deep Metric Learning (AAAI 2021 · similarity 0.88)
- Deep Adversarial Metric Learning (CVPR 2018 · similarity 0.87)
- Stochastic Class-Based Hard Example Mining for Deep Metric Learning (CVPR 2019 · similarity 0.87)
- Embedding Expansion: Augmentation in Embedding Space for Deep Metric Learning (CVPR 2020 · similarity 0.86)
- Signal-To-Noise Ratio: A Robust Distance Metric for Deep Metric Learning (CVPR 2019 · similarity 0.86)
- Mathematical Justification of Hard Negative Mining via Isometric Approximation Theorem (ICLR 2024 · similarity 0.84)
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.