Research map / Label-efficient and robust learning / Recognition and long-tail
LoOp: Looking for Optimal Hard Negative Embeddings for Deep Metric Learning
Bhavya Vasudeva, Puneesh Deora, Saumik Bhattacharya, Umapada Pal, Sukalpa Chanda
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 LoOp optimal hard negatives | Triplet loss · CUB-200-2011 | Recall@1: 35.9 -> 50.3 | Improved |
| add LoOp optimal hard negatives | Triplet loss · SOP | Recall@1: 53.9 -> 69.6 | Improved |
| add LoOp optimal hard negatives | HPHN-triplet loss · Cars196 | Recall@1: 60.3 -> 74.5 | Improved |
| add LoOp optimal hard negatives | HPHN-triplet loss · SOP | NMI: 91.4 -> 90.8 | Hurt |
| add LoOp optimal hard negatives | HPHN-triplet loss · SOP | F1: 43.3 -> 36.9 | Hurt |
| add LoOp optimal hard negatives | Lifted structure loss · CUB-200-2011 | Recall@1: 46.9 -> 52.8 | Improved |
| add LoOp optimal hard negatives | Lifted structure loss · Cars196 | NMI: 57.8 -> 57.4 | Hurt |
| add LoOp optimal hard negatives | Multi-similarity loss · Cars196 | Recall@1: 71.0 -> 72.6 | Improved |
| add LoOp optimal hard negatives | Multi-similarity loss · Cars196 | F1: 31.7 -> 30.6 | Hurt |
| compute optimal points on l2-normalized curves instead of line segments (ablation) | LoOp (Triplet) · Cars196 | Recall@1 | Improved |
| use LoOp on line segments without l2-normalization versus plain triplet (ablation) | Triplet loss · Cars196 | Recall@1 | Improved |
| increase batch size (ablation) | LoOp (HPHNtri) · Cars196 | Recall@1 | Improved |
Closest papers by meaning
- Divide and Conquer the Embedding Space for Metric Learning (CVPR 2019 · similarity 0.89)
- Deep Metric Learning via Lifted Structured Feature Embedding (CVPR 2016 · similarity 0.89)
- Improving Deep Metric Learning by Divide and Conquer (TPAMI 2021 · similarity 0.89)
- Symmetrical Synthesis for Deep Metric Learning (AAAI 2020 · similarity 0.89)
- Stochastic Class-Based Hard Example Mining for Deep Metric Learning (CVPR 2019 · similarity 0.89)
- Mathematical Justification of Hard Negative Mining via Isometric Approximation Theorem (ICLR 2024 · similarity 0.87)
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.
