Research map / Label-efficient and robust learning / Recognition and long-tail
How to Boost Face Recognition with StyleGAN?
Artem Sevastopolskiy, Yury Malkov, Nikita Durasov, Luisa Verdoliva, Matthias Nießner
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| pretrain encoder into StyleGAN latent space on unlabeled African+Asian faces before face recognition fine-tuning | ArcFace R-50 · RB-WebFace TPR@FPR=1e-3 (African) | TPR: 86.73 -> 91.51 | Improved |
| add StyleGAN pretraining | ArcFace R-50 · RFW average accuracy | accuracy: 94.39 -> 95.09 | Improved |
| add StyleGAN pretraining to other FR methods | SphereFace · RB-WebFace TPR@FPR=1e-3 (African) | TPR: 89.51 -> 90.52 | Improved |
| add StyleGAN pretraining | GAC · RB-WebFace TPR@FPR=1e-3 (African) | TPR: 89.13 -> 90.80 | Improved |
| increase unlabeled prior dataset from 1% to 100% (ablation) | StyleGAN pretraining · RB-WebFace TPR@FPR=1e-3 (African) | TPR: 90.44 -> 91.51 | Improved |
| use StyleGAN-encoder pretraining instead of autoencoder pretraining (ablation) | pretraining strategy · RB-WebFace TPR@FPR=1e-3 (African) | TPR: 90.83 -> 91.51 | Improved |
| use StyleGAN-encoder pretraining instead of VAE pretraining (ablation) | pretraining strategy · RB-WebFace TPR@FPR=1e-3 (African) | TPR: 90.39 -> 91.51 | Improved |
| add synthetic StyleGAN interpolation samples instead of pretraining (ablation) | ArcFace R-50 · RB-WebFace TPR@FPR=1e-3 (African) | TPR: 86.73 -> 86.85 with 6M interps | No clear effect |
| add pretraining on ResNet-100 backbone | ArcFace R-100 · RB-WebFace TPR@FPR=1e-3 (African) | TPR: 91.55 -> 93.00 | Improved |
| reduce labeled face recognition data (ablation) | StyleGAN pretraining · RFW and RB-WebFace | accuracy/TPR gain | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- ArcFace — ArcFace: Additive Angular Margin Loss for Deep Face Recognition (CVPR 2019) · better on 2 of 2 shared results
- SphereFace — SphereFace: Deep Hypersphere Embedding for Face Recognition (CVPR 2017) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- BroadFace: Looking at Tens of Thousands of People at Once for Face Recognition (ECCV 2020 · similarity 0.88)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.87)
- HyperFace: Generating Synthetic Face Recognition Datasets by Exploring Face Embedding Hypersphere (ICLR 2025 · similarity 0.87)
- UIFace: Unleashing Inherent Model Capabilities to Enhance Intra-Class Diversity in Synthetic Face Recognition (ICLR 2025 · similarity 0.86)
- CemiFace: Center-based Semi-hard Synthetic Face Generation for Face Recognition (NeurIPS 2024 · similarity 0.85)
- LAFS: Landmark-based Facial Self-supervised Learning for Face Recognition (CVPR 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.
