Research map / Label-efficient and robust learning / Recognition and long-tail
Towards Interpretable Deep Metric Learning With Structural Matching
Wenliang Zhao, Yongming Rao, Ziyi Wang, Jiwen Lu, Jie Zhou
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 |
|---|---|---|---|
| add DIML structural matching at test time (no training) | ProxyAnchor · Cars196 | Precision@1: 82.36 -> 86.13 | Improved |
| add DIML at test time | Margin · CUB200-2011 | Precision@1: 62.47 -> 65.16 | Improved |
| add DIML at test time | Contrastive · Cars196 | Precision@1: 67.45 -> 73.35 | Improved |
| add DIML at test time | ProxyAnchor · SOP | Precision@1: 79.10 -> 79.22 | No clear effect |
| add multi-scale matching to structural similarity (ablation) | Margin + DIML · SOP | Precision@1: 77.30 -> 78.53 | Improved |
| add structural similarity alone replacing cosine similarity (ablation) | Margin · SOP | Precision@1: 78.39 -> 77.30 | Hurt |
| add structural similarity alone replacing cosine similarity (ablation) | MS · SOP | Precision@1: 77.90 -> 75.81 | Hurt |
| add structural similarity alone replacing cosine similarity (ablation) | Margin · Cars196 | Precision@1: 72.18 -> 74.86 | Improved |
| add cross-correlation marginals (ablation) | Margin + DIML (SS+MM) · Cars196 | Precision@1: 76.02 -> 76.62 | Improved |
| add cross-correlation marginals (ablation) | MS + DIML (SS+MM) · SOP | Precision@1: 77.87 -> 78.53 | Improved |
| increase truncation number K from 10 to 100 (ablation) | Margin + DIML · CUB200-2011 | Precision@1: 65.16 -> 65.16 | No clear effect |
| increase feature map size G from 1 to 7 (ablation) | Margin + DIML · Cars196 | Precision@1: 72.18 -> 76.96 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Arcface — ArcFace: Additive Angular Margin Loss for Deep Face Recognition (CVPR 2019) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DIML: Deep Interpretable Metric Learning via Structural Matching (TPAMI 2023 · similarity 0.97)
- Improving Deep Metric Learning by Divide and Conquer (TPAMI 2021 · similarity 0.90)
- Divide and Conquer the Embedding Space for Metric Learning (CVPR 2019 · similarity 0.89)
- Deep Factorized Metric Learning (CVPR 2023 · similarity 0.89)
- Deep Compositional Metric Learning (CVPR 2021 · similarity 0.88)
- Deep Shape Matching (ECCV 2018 · similarity 0.88)
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.
