Research map / Label-efficient and robust learning / Recognition and long-tail
Deep Metric Learning With Angular Loss
Jian Wang, Feng Zhou, Shilei Wen, Xiao Liu, Yuanqing Lin
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 |
|---|---|---|---|
| replace triplet loss (T-II) with angular loss, same N-pair sampling | triplet embedding · CUB-200-2011 | Recall@1: 42.8 -> 53.6 | Improved |
| replace triplet loss (T-II) with angular loss | triplet embedding · Stanford Car | Recall@1: 46.3 -> 71.3 | Improved |
| replace N-pair loss with angular loss | N-pair loss · Stanford Car | Recall@1: 68.9 -> 71.3 | Improved |
| replace N-pair loss with angular loss | N-pair loss · CUB-200-2011 | Recall@1: 51.9 -> 53.6 | Improved |
| add angular loss to N-pair loss (NL&AL) | N-pair loss · CUB-200-2011 | Recall@1: 51.9 -> 54.7 | Improved |
| add angular loss to N-pair loss (NL&AL) | N-pair loss · Online Products | Recall@1: 66.9 -> 70.9 | Improved |
| add angular loss to N-pair loss (NL&AL) | N-pair loss · Online Products | F1: 26.3 -> 29.9 | Improved |
| add angular loss to N-pair loss (NL&AL) | N-pair loss · CUB-200-2011 | F1: 30.2 -> 29.4 | Hurt |
| add angular loss to N-pair loss (NL&AL) | N-pair loss · Stanford Car | Recall@1: 68.9 -> 71.4 | Improved |
| raise angle margin alpha from 36 to 45 degrees (ablation) | NL&AL · Stanford Car | Recall@1: 69.9 -> 71.4 | Improved |
| raise angle margin alpha from 45 to 55 degrees (ablation) | NL&AL · Stanford Car | Recall@1: 71.4 -> 69.0 | Hurt |
| lower angle margin alpha from 45 to 36 degrees (ablation) | NL&AL · Online Products | Recall@1: 69.2 -> 70.9 | Improved |
Closest papers by meaning
- Deep Metric Learning Beyond Binary Supervision (CVPR 2019 · similarity 0.87)
- Circle Loss: A Unified Perspective of Pair Similarity Optimization (CVPR 2020 · similarity 0.87)
- Improved Deep Metric Learning with Multi-class N-pair Loss Objective (NeurIPS 2016 · similarity 0.87)
- Deep Metric Learning with Spherical Embedding (NeurIPS 2020 · similarity 0.87)
- Hard negative examples are hard, but useful (ECCV 2020 · similarity 0.86)
- DIML: Deep Interpretable Metric Learning via Structural Matching (TPAMI 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.
