Research map / Label-efficient and robust learning / Recognition and long-tail
Deep Representation Learning on Long-Tailed Data: A Learnable Embedding Augmentation Perspective
Check 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 feature-cloud augmentation to CosFace | CosFace · Market-1501, ⟨H20,S5⟩ | Rank-1: 78.6 -> 83.2 | Helped |
| Add feature-cloud augmentation to CosFace | CosFace · DukeMTMC, ⟨H20,S5⟩ | Rank-1: 66.0 -> 72.7; +6.7% | Helped |
| Add feature-cloud augmentation to CosFace | CosFace · Market-1501, ⟨H20,S3⟩ | Rank-1: 55.5 -> 67.2; +11.7% | Helped |
| Add feature-cloud augmentation to CosFace | CosFace · Market-1501, ⟨H20,S4⟩ | Rank-1: 67.7 -> 76.8 | Helped |
| Add feature-cloud augmentation to CosFace | CosFace · LFW, ⟨H5K,T10K⟩ | Rank-1: 98.73 -> 98.88 | Helped |
| Add feature-cloud augmentation to CosFace | CosFace · MegaFace, ⟨H3K,T20K⟩ | Rank-1: 74.06 -> 75.18 | Helped |
| Add feature-cloud augmentation to CosFace | CosFace · IJB-C TPR@FPR 1e-5, ⟨H3K,T10K⟩ | TPR@FPR: 56.52 -> 58.89 | Helped |
| Use the full version's distribution-weighted adaptive selection rather than explicit head/tail thresholding (ablation) | LEAP with CosFace · Market-1501, ⟨H20,S5⟩ | Rank-1: 83.2 -> 83.4 | No clear effect |
| Reduce the number of tail-class samples from 5 to 3 (ablation) | LEAP-CF · Market-1501, ⟨H20,S3⟩ | Rank-1: 76.8 -> 67.2 | Hurt |
| Construct the feature cloud during epochs 10–30 rather than outside that interval (ablation) | LEAP with CosFace · Market-1501 and DukeMTMC, ⟨H20,S4⟩ | Rank-1 | No clear effect |
| Construct the feature cloud during epochs 20–30 rather than outside that interval (ablation) | LEAP with ArcFace · Market-1501 and DukeMTMC, ⟨H20,S4⟩ | Rank-1 | No clear effect |
Closest papers by meaning
- Geometry of Long-Tailed Representation Learning: Rebalancing Features for Skewed Distributions (ICLR 2025 · similarity 0.90)
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.90)
- Learning from Neighbors: Category Extrapolation for Long-Tail Learning (CVPR 2025 · similarity 0.87)
- Long-Tailed Recognition via Information-Preservable Two-Stage Learning (NeurIPS 2025 · similarity 0.87)
- DiffuLT: Diffusion for Long-tail Recognition Without External Knowledge (NeurIPS 2024 · similarity 0.86)
- Decoupled Contrastive Learning for Long-Tailed Recognition (AAAI 2024 · 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.