Face Anti-Spoofing on MSU-MFSD
4.58HTER (%)Ours-100K
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Ours-100KTraining Dataset=FaceCoT-Gold100K2025.06 | 4.58 | — | — | — | 99.56 | |
| Ours-AllTraining Dataset=Full FaceCoT2025.06 | 5 | — | — | — | 99.35 | |
| I-FASTraining Dataset=CelebA-Spoof2025.06 | 5.63 | — | — | — | 98.73 | |
| I-FASEvaluation Protocol=Protocol 22026.03 | 5.63 | — | — | — | 98.73 | |
| I-FASProtocol=2, Training Domain=CelebA-Spoof2026.03 | 5.63 | — | — | — | 98.73 | |
| TAR-FASEvaluation Protocol=Protocol 22026.03 | 5.71 | — | — | — | 98.29 | |
| TAR-FASProtocol=2, Training Domain=CelebA-Spoof2026.03 | 5.71 | — | — | — | 98.29 | |
| Ours-CelebATraining Dataset=CelebA-Spoof (annotated by FAS caption model)2025.06 | 8.33 | — | — | — | 98.29 | |
| ViTAFTraining Dataset=CelebA-Spoof2025.06 | 12.86 | — | — | — | 93.14 | |
| ViTAFEvaluation Protocol=Protocol 22026.03 | 12.86 | — | — | — | 93.14 | |
| ViTAFProtocol=2, Training Domain=CelebA-Spoof2026.03 | 12.86 | — | — | — | 93.14 | |
| ViT-BTraining Dataset=CelebA-Spoof2025.06 | 16.67 | — | — | — | 89.89 | |
| FLIPTraining Dataset=CelebA-Spoof2025.06 | 19.37 | — | — | — | 89.98 | |
| FLIPEvaluation Protocol=Protocol 22026.03 | 19.37 | — | — | — | 89.95 | |
| FLIPProtocol=2, Training Domain=CelebA-Spoof2026.03 | 19.37 | — | — | — | 89.95 | |
| ViT-LTraining Dataset=CelebA-Spoof2025.06 | 20.87 | — | — | — | 85.65 | |
| ViT-LEvaluation Protocol=Protocol 22026.03 | 20.87 | — | — | — | 85.65 | |
| ViT-LProtocol=2, Training Domain=CelebA-Spoof2026.03 | 20.87 | — | — | — | 85.65 | |
| CDCN2020.03 | — | 70.82 | 100 | 99.99 | — | |
| CDCN++2020.03 | — | 72.29 | 100 | 99.98 | — | |
| DTNlearning paradigm=zero-shot learning2020.03 | — | 81.6 | 99.9 | 97.5 | — | |
| NN+LBP2020.03 | — | 50.57 | 99.93 | 93.54 | — | |
| OC-SVM_RBF+BSIF2020.03 | — | 64.81 | 87.44 | 74.69 | — | |
| SVM_RBF+LBP2020.03 | — | 47.68 | 99.5 | 97.61 | — |