Research map / Label-efficient and robust learning / Noisy labels
Unequal-Training for Deep Face Recognition With Long-Tailed Noisy Data
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 CD tail training to NRA on the MF2 head subset (ablation) | NRA · MF2 FaceScrub, 1M distractors | Rank1 accuracy: 79.52 -> 80.02 | Helped |
| Add CD tail training to NRA on the MF2 head subset (ablation) | NRA · MF2 FGNet, 1M distractors | Rank1 accuracy: 59.89 -> 60.39 | Helped |
| Add CD tail training to NRA on the MF2 head subset (ablation) | NRA · MF2 FaceScrub | TAR at 10⁻⁶ FAR: 88.55 -> 89.93 | Helped |
| Add CD tail training to NRA on the MF2 head subset (ablation) | NRA · MF2 FGNet | TAR at 10⁻⁶ FAR: 60.09 -> 60.99 | Helped |
| Use NRA instead of ArcFace on the MF2 head subset | MF2-h9 training · MF2 FaceScrub, 1M distractors | Rank1 accuracy: 78.97 -> 79.52 | Helped |
| Use the ≥9 images/id head subset rather than the ≥10 images/id subset (ablation) | MF2 training subset selection · MF2 FaceScrub, 1M distractors | Rank1 accuracy: 76.24 -> 78.97 | Helped |
| Use the ≥9 images/id head subset rather than the ≥8 images/id subset (ablation) | MF2 training subset selection · MF2 FaceScrub, 1M distractors | Rank1 accuracy: 78.32 -> 78.97 | Helped |
| Train on all long-tailed MF2 data with rolling softmax rather than only the head subset (ablation) | Softmax · MF2 FaceScrub, 1M distractors | Rank1 accuracy: 54.72 -> 42.87 | Hurt |
| Replace NRA+CD with ArcFace on the WebFace+ one-shot set | ResNet50 face recognition · CPLFW | accuracy: 88.03 -> 86.32 | Hurt |
| Use the proposed NR+CD training instead of the corresponding margin-loss baseline | ResNet50 · CPLFW | accuracy: 87.67 -> 87.92; 87.20 -> 87.53 | Mixed |
Closest papers by meaning
- Noise-Tolerant Paradigm for Training Face Recognition CNNs (CVPR 2019 · similarity 0.89)
- Killing Two Birds With One Stone: Efficient and Robust Training of Face Recognition CNNs by Partial FC (CVPR 2022 · similarity 0.87)
- AdaptiveFace: Adaptive Margin and Sampling for Face Recognition (CVPR 2019 · similarity 0.87)
- Towards Universal Representation Learning for Deep Face Recognition (CVPR 2020 · similarity 0.86)
- Difficulty-aware Balancing Margin Loss for Long-tailed Recognition (AAAI 2025 · similarity 0.85)
- Long-tailed Recognition with Model Rebalancing (NeurIPS 2025 · similarity 0.84)
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.