Deepfake Detection on LAV-DF clean
100APVisual-only (WAM)
Evaluation results
| Method | Links | |
|---|---|---|
| Visual-only (WAM)Type=Semi-fragile WM, Modality=Visual2026.04 | 100 | |
| LAVAType=Multi-modal WM, Modality=A + V2026.04 | 100 | |
| Audio-only (AudioSeal)Type=Semi-fragile WM, Modality=Audio2026.04 | 99.9 | |
| EditGuardType=Semi-fragile WM, Modality=Visual2026.04 | 54.9 | |
| WavMarkType=Neural audio WM, Modality=Audio2026.04 | 35.4 | |
| EfficientNet-B4Type=Learned, Modality=Visual, Zero-shot=true, Training dataset=FF++ c232026.04 | 5 | |
| XceptionNetType=Learned, Modality=Visual, Zero-shot=true, Training dataset=FF++ c232026.04 | 3.9 |