Research map / Label-efficient and robust learning / Recognition and long-tail
MagFace: A Universal Representation for Face Recognition and Quality Assessment
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 |
|---|---|---|---|
| Replace ArcFace's fixed margin with a magnitude-adaptive margin and magnitude regularizer | ResNet100 face recognition model · LFW | verification accuracy (%): 0.02% | Helped |
| Replace ArcFace's fixed margin with a magnitude-adaptive margin and magnitude regularizer | ResNet100 face recognition model · CFP-FP | verification accuracy (%): 0.06% | Helped |
| Replace ArcFace's fixed margin with a magnitude-adaptive margin and magnitude regularizer | ResNet100 face recognition model · CALFW | verification accuracy (%): 0.19% | Hurt |
| Replace ArcFace's fixed margin with a magnitude-adaptive margin and magnitude regularizer | ResNet100 face recognition model · IJB-B, TAR@FAR=1e-6 | TAR: 2.23% | Helped |
| Aggregate identity features weighted by their magnitudes | MagFace · IJB-C, TAR@FAR=1e-6 | TAR: 1.41% | Helped |
| Use magnitude-aware face embeddings instead of ArcFace embeddings | Face clustering with L-GCN · IJB-B-1845 | F-score (%) | Helped |
| Set margin bounds to 0.45 and 0.65 rather than the balanced 0.40 and 0.80 (ablation) | MagFace · CFP-FP | verification accuracy (%): 97.47 -> 97.23 | Hurt |
| Set margin bounds to 0.25 and 1.60 rather than the balanced 0.40 and 0.80 (ablation) | MagFace · CFP-FP | verification accuracy (%): 97.47 -> 97.30 | Hurt |
| Use balanced margin bounds 0.40 and 0.80 (ablation) | MagFace · CFP-FP | verification accuracy (%): 97.32 -> 97.47 | Helped |
| Use balanced margin bounds 0.40 and 0.80 (ablation) | MagFace · IJB-C, TAR@FAR=1e-6 | TAR: 83.88 -> 85.82 | Helped |
Closest papers by meaning
- AdaFace: Quality Adaptive Margin for Face Recognition (CVPR 2022 · similarity 0.88)
- SER-FIQ: Unsupervised Estimation of Face Image Quality Based on Stochastic Embedding Robustness (CVPR 2020 · similarity 0.88)
- Attributes Shape the Embedding Space of Face Recognition Models (ICML 2025 · similarity 0.85)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.85)
- HyperFace: Generating Synthetic Face Recognition Datasets by Exploring Face Embedding Hypersphere (ICLR 2025 · similarity 0.84)
- RepFace: Refining Closed-Set Noise with Progressive Label Correction for Face Recognition (AAAI 2025 · similarity 0.83)
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.