Deepfake Detection on FaceForensics++ (Artifact Quality Metrics)
99.96Saturation ScoreLAA-Net
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| LAA-NetTraining set (Real)=true, Training set (Fake)=false2026.04 | 99.96 | 99.96 | 99.96 | 53.9 | 99.8 | 90.72 | |
| LipForensicsTraining set (Real)=true, Training set (Fake)=true2026.04 | 99.9 | 99.6 | 87.4 | 73.8 | 95.6 | 91.26 | |
| LAA-SwinTraining set (Real)=true, Training set (Fake)=false2026.04 | 99.79 | 99.77 | 99.92 | 81.6 | 94.88 | 95.19 | |
| XceptionTraining set (Real)=true, Training set (Fake)=true2026.04 | 99.3 | 98.6 | 99.7 | 53.8 | 74.2 | 85.12 | |
| CNN-augTraining set (Real)=true, Training set (Fake)=true2026.04 | 99.3 | 99.1 | 95.2 | 54.7 | 91.2 | 87.9 | |
| LSDATraining set (Real)=true, Training set (Fake)=true2026.04 | 98.7 | 94.4 | 98.3 | 66.4 | 90.7 | 89.7 | |
| LAA-FormerTraining set (Real)=true, Training set (Fake)=false2026.04 | 98.04 | 95.96 | 97.02 | 75.28 | 91.14 | 91.49 | |
| FaceXrayTraining set (Real)=true, Training set (Fake)=true2026.04 | 97.6 | 88.5 | 99.1 | 49.8 | 88.6 | 84.72 | |
| SBITraining set (Real)=true, Training set (Fake)=false2026.04 | 92 | 92.3 | 92.2 | 62.2 | 79.1 | 83.56 |