Research map / Image generation / Editing and style transfer
DiffFAS: Face Anti-Spoofing via Generative Diffusion Models
Xinxu Ge, Xin Liu, Zitong Yu, Jingang Shi, Chun Qi, Jie Li, Heikki Kälviäinen
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 |
|---|---|---|---|
| add generated spoof data from DiffFAS (ablation) | DiffFAS-V · MI->C | HTER: 15.50 -> 15.06 | Improved |
| add generated spoof data from DiffFAS (ablation) | DiffFAS-V · PADISI->WMCA | HTER: 15.15 -> 12.67 | Improved |
| add Relative Quality (RQ) loss (ablation) | DiffFAS-V · MI->C | HTER: 16.91 -> 15.06 | Improved |
| add RQ loss (ablation) | DiffFAS-V · WMCA->PADISI | HTER: 8.13 -> 7.09 | Improved |
| replace cross-entropy with RQ loss (ablation) | ResNet-18 · OCI->M | HTER: 7.71 -> 5.90 | Improved |
| use focal loss instead of cross-entropy (ablation) | ResNet-18 · OCI->M | HTER: 7.71 -> 11.48 | Hurt |
| set RQ live/spoof scaling to 0.4/0.2 instead of 0.4/0.1 (ablation) | RQ loss · OCI->M | HTER: 6.57 -> 3.10 | Improved |
| use DiffFAS generated data instead of DSDG | ResNet-18 · OCI->M | HTER: 8.57 -> 7.71 | Improved |
| oversample high-quality OULU data 3x (ablation) | ResNet-18 · OCI->M | HTER: 9.67 -> 8.29 | Improved |
| oversample high-quality OULU data 4x (ablation) | ResNet-18 · OCI->M | HTER: 9.67 -> 12.86 | Hurt |
| use patch-wise mean and variance in style fusion module (ablation) | DiffFAS generator · PADISI / OCIM generation | identity preservation (qualitative) | Improved |
Closest papers by meaning
- Towards Data-Centric Face Anti-spoofing: Improving Cross-Domain Generalization via Physics-Based Data Synthesis (IJCV 2024 · similarity 0.90)
- Test-Time Domain Generalization for Face Anti-Spoofing (CVPR 2024 · similarity 0.89)
- SLIP: Spoof-Aware One-Class Face Anti-Spoofing with Language Image Pretraining (AAAI 2025 · similarity 0.89)
- FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models (AAAI 2026 · similarity 0.89)
- mmFAS: Multimodal Face Anti-Spoofing Using Multi-Level Alignment and Switch-Attention Fusion (AAAI 2025 · similarity 0.89)
- Gradient Alignment for Cross-Domain Face Anti-Spoofing (CVPR 2024 · similarity 0.88)
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.
