Research map / Label-efficient and robust learning / Recognition and long-tail
Variational Prototype Learning 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 |
|---|---|---|---|
| Add memorized feature injection to ArcFace prototypes | ArcFace (ResNet50) · CFP-FP | verification accuracy (%): 98.54 -> 98.96 | Helped |
| Add memorized feature injection to ArcFace prototypes | ArcFace (ResNet50) · IJB-C | TAR@FAR=1e-4: 96.21 -> 96.61 | Helped |
| Add memorized feature injection to CosFace prototypes | CosFace (ResNet50) · CFP-FP | verification accuracy (%): 98.51 -> 98.81 | Helped |
| Add memorized feature injection to CurricularFace prototypes | CurricularFace (ResNet50) · IJB-C | TAR@FAR=1e-4: 96.30 -> 96.65 | Helped |
| Set feature injection weight λ to 0.15 rather than 0 (ablation) | VPL-ArcFace (ResNet50) · IJB-C | TAR@FAR=1e-4: 96.21 -> 96.61 | Helped |
| Increase feature injection weight from λ = 0.15 to λ = 0.2 (ablation) | VPL-ArcFace (ResNet50) · CFP-FP | verification accuracy (%): 98.96 -> 98.82 | Hurt |
| Increase maximum accumulated iterations ∆t from 0 to 100 (ablation) | VPL-ArcFace (ResNet50) · IJB-C | TAR@FAR=1e-4: 96.21 -> 96.61 | Helped |
| Increase maximum accumulated iterations from ∆t = 100 to ∆t = 2000 (ablation) | VPL-ArcFace (ResNet50) · IJB-C | TAR@FAR=1e-4: 96.61 -> 96.22 | Hurt |
| Apply feature injection only to the Asian group (ablation) | ArcFace (ResNet50) · IFRT Asian | 1:1 verification accuracy (%): 37.27 -> 39.85 | Helped |
| Apply feature injection only to long-tail classes (ablation) | ArcFace (ResNet50) · IFRT All | 1:1 verification accuracy (%): 70.99 -> 72.78 | Helped |
Closest papers by meaning
- Robust Classification With Convolutional Prototype Learning (CVPR 2018 · similarity 0.86)
- Disentangled Variational Representation for Heterogeneous Face Recognition (AAAI 2019 · similarity 0.85)
- CosFace: Large Margin Cosine Loss for Deep Face Recognition (CVPR 2018 · similarity 0.85)
- Killing Two Birds With One Stone: Efficient and Robust Training of Face Recognition CNNs by Partial FC (CVPR 2022 · similarity 0.85)
- UIFace: Unleashing Inherent Model Capabilities to Enhance Intra-Class Diversity in Synthetic Face Recognition (ICLR 2025 · similarity 0.84)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 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.