Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Proxy Synthesis: Learning with Synthetic Classes 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 Proxy Synthesis to Proxy-anchor | Proxy-anchor · CUB200, conventional evaluation | Recall@1: 68.4 -> 69.2 | Helped |
| Add Proxy Synthesis to Proxy-anchor | Proxy-anchor · CARS196, conventional evaluation | Recall@1: 86.1 -> 86.9 | Helped |
| Add Proxy Synthesis to Proxy-anchor | Proxy-anchor · SOP, conventional evaluation | Recall@1: 79.1 -> 79.8 | Helped |
| Add Proxy Synthesis to Proxy-anchor | Proxy-anchor · In-Shop, conventional evaluation | Recall@1: 91.5 -> 91.9 | Helped |
| Use only synthetic embeddings and proxies, excluding originals (ablation) | Norm-softmax · CARS196 | Recall@1: 83.3 -> 83.1 | Hurt |
| Add synthetic proxies while retaining original embeddings and proxies (ablation) | Norm-softmax · CARS196 | Recall@1: 83.3 -> 83.7 | Helped |
| Use synthetic embeddings with original and synthetic proxies (ablation) | Norm-softmax · CARS196 | Recall@1: 83.3 -> 83.7 | Helped |
| Use original and synthetic embeddings and proxies together (ablation) | Norm-softmax · CARS196 | Recall@1: 83.3 -> 84.7 | Helped |
| Set static interpolation coefficient λ to 0.2 rather than 0.1 (ablation) | Proxy Synthesis + Norm-softmax · CARS196 | Recall@1: 83.1 -> 83.8 | Helped |
| Use stochastic interpolation with α=0.4 rather than static λ=0.2 (ablation) | Proxy Synthesis + Norm-softmax · CARS196 | Recall@1: 83.8 -> 84.7 | Helped |
| Set stochastic interpolation parameter α to 0.4 rather than 0.2 (ablation) | Proxy Synthesis + Norm-softmax · CARS196 | Recall@1: 84.0 -> 84.7 | Helped |
| Add Proxy Synthesis to Norm-softmax | Norm-softmax · CARS196, input deformation | Recall@1: 75.3 -> 77.0; 59.7 -> 62.2; 64.3 -> 65.6; 78.3 -> 80.0; 70.8 -> 72.1; 70.3 -> 72.0; 65.1 -> 67.2; 74.4 -> 76.3 | Helped |
Closest papers by meaning
- Non-Isotropy Regularization for Proxy-Based Deep Metric Learning (CVPR 2022 · similarity 0.91)
- Proxy Anchor Loss for Deep Metric Learning (CVPR 2020 · similarity 0.89)
- Towards Improved Proxy-Based Deep Metric Learning via Data-Augmented Domain Adaptation (AAAI 2024 · similarity 0.88)
- Deep Disentangled Metric Learning (AAAI 2025 · similarity 0.87)
- Mean Field Theory in Deep Metric Learning (ICLR 2024 · similarity 0.84)
- Potential Field Based Deep Metric Learning (CVPR 2025 · 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.