Research map / Label-efficient and robust learning / Recognition and long-tail
Mining on Manifolds: Metric Learning Without Labels
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 |
|---|---|---|---|
| Use manifold-guided positive and negative pools instead of Euclidean neighbors and non-neighbors (ablation) | Unsupervised metric learning · CUB200-2011 | Recall@1: 38.5 -> 43.5 | Helped |
| Use manifold-guided positive and negative pools instead of Euclidean pools (ablation) | Unsupervised metric learning · Oxford5k | mAP: 41.9 -> 73.7 | Helped |
| Add manifold-guided hard negatives (ablation) | Manifold-guided metric learning · Oxford5k | mAP: 38.1 -> 73.7 | Helped |
| Add manifold-guided hard positives (ablation) | Manifold-guided metric learning · Oxford5k | mAP: 71.3 -> 73.7 | Helped |
| Weight tuples by positive manifold similarity (ablation) | Manifold-guided metric learning · CUB200-2011 | Recall@1: 43.5 -> 45.3 | Helped |
| Weight tuples by positive manifold similarity (ablation) | Manifold-guided metric learning · Oxford5k | mAP: 73.7 -> 76.7 | Helped |
| Select anchors by graph-node importance rather than randomly (ablation) | Manifold-guided retrieval training · Oxford5k | mAP: 67.0 -> 76.7 | Helped |
| Fine-tune with manifold-mined examples | GoogLeNet · CUB200-2011 | Recall@1: 35.0 -> 45.3 | Helped |
| Fine-tune with manifold-mined examples | VGG with MAC · Oxford5k | mAP: 58.5 -> 78.7 | Helped |
| Use manifold-mined training instead of SfM-supervised training | VGG with MAC · Oxford5k | mAP: 79.7 -> 78.7 | Hurt |
| Use manifold-mined training instead of SfM-supervised training | VGG with MAC · Instre | mAP: 48.5 -> 55.5 | Helped |
Closest papers by meaning
- Ensemble Deep Manifold Similarity Learning Using Hard Proxies (CVPR 2019 · similarity 0.86)
- Stochastic Class-Based Hard Example Mining for Deep Metric Learning (CVPR 2019 · similarity 0.86)
- Deep Networks and the Multiple Manifold Problem (ICLR 2021 · similarity 0.86)
- Efficient Representation of Low-Dimensional Manifolds using Deep Networks (ICLR 2017 · similarity 0.85)
- Co-manifold learning with missing data (ICML 2019 · similarity 0.85)
- Fast Spectral Ranking for Similarity Search (CVPR 2018 · similarity 0.85)
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.