Research map / Label-efficient and robust learning / Recognition and long-tail
CurricularFace: Adaptive Curriculum Learning Loss for Deep Face Recognition
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 |
|---|---|---|---|
| Use CurricularFace instead of ArcFace | ResNet100 face recognition · CFP-FP | verification accuracy (%): 98.27 -> 98.37 | Helped |
| Use CurricularFace instead of ArcFace | ResNet100 face recognition · IJB-B 1:1 verification TAR (@FAR=1e-4) | TAR (%): 94.2 -> 94.8 | Helped |
| Use CurricularFace instead of ArcFace | ResNet100 face recognition · IJB-C 1:1 verification TAR (@FAR=1e-4) | TAR (%): 95.6 -> 96.1 | Helped |
| Adapt t during training rather than fixing it (ablation) | CurricularFace · LFW | verification accuracy (%): 99.45 -> 99.47 | Helped |
| Adapt t during training rather than fixing it (ablation) | CurricularFace · CFP-FP | verification accuracy (%): 93.94 -> 96.96 | Helped |
| Estimate t with mean positive cosine similarity instead of its mode (ablation) | CurricularFace · CFP-FP | verification accuracy (%): 96.49 -> 96.96 | Helped |
| Estimate t with mean positive cosine similarity instead of mean predicted probability (ablation) | CurricularFace · CFP-FP | verification accuracy (%): 95.39 -> 96.96 | Helped |
| Train with CurricularFace instead of ArcFace at margin m=0.5 (ablation) | MobileFaceNet trained on CASIA-WebFace · LFW | verification accuracy (%): ArcFace does not converge; CurricularFace 99.25% | Helped |
| Train with CurricularFace instead of ArcFace at margin m=0.45 (ablation) | MobileFaceNet trained on CASIA-WebFace · LFW | verification accuracy (%): 99.10% -> 99.20% | Helped |
Closest papers by meaning
- AdaptiveFace: Adaptive Margin and Sampling for Face Recognition (CVPR 2019 · similarity 0.89)
- Mis-Classified Vector Guided Softmax Loss for Face Recognition (AAAI 2020 · similarity 0.88)
- AdaFace: Quality Adaptive Margin for Face Recognition (CVPR 2022 · similarity 0.88)
- CosFace: Large Margin Cosine Loss for Deep Face Recognition (CVPR 2018 · similarity 0.87)
- ArcFace: Additive Angular Margin Loss for Deep Face Recognition (CVPR 2019 · similarity 0.87)
- Killing Two Birds With One Stone: Efficient and Robust Training of Face Recognition CNNs by Partial FC (CVPR 2022 · 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.