Deepfake Detection on Deepfake Detection Suite (NT, DF, F2F, FS, FSH, CDFv2, FFIW10K) (test)
92.25NT AccuracyQAD-E
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| QAD-Ew/ prior infor.=N, #params=6.5M x 1, backbone=EfficientNet-B1, training_modalities=12023.09 | 92.25 | 99.46 | 98.3 | 99.08 | 98.9 | 97.5 | 99.01 | 97.79 | |
| BZNetw/ prior infor.=Y, #params=22M x 3, training_modalities=3 (raw, c23, c40)2023.09 | 91.01 | 99.3 | 96.9 | 98.82 | — | — | — | 96.51 | |
| ADDw/ prior infor.=Y, #params=23.5M x 3, training_modalities=3 (raw, c23, c40)2023.09 | 89.08 | 99.25 | 96.53 | 98.21 | 98.25 | — | — | 96.26 | |
| RESNET50w/ prior infor.=Y, #params=23.5M x 3, training_modalities=3 (raw, c23, c40)2023.09 | 88.96 | 99.26 | 97.04 | 98.63 | 98.71 | 97.09 | 98.58 | 96.9 | |
| QAD-Rw/ prior infor.=N, #params=23.5M x 1, backbone=ResNet-50, training_modalities=12023.09 | 88.85 | 99.42 | 97.77 | 98.83 | 98.93 | 97.56 | 98.93 | 97.18 | |
| EFFICIENTNET-B1w/ prior infor.=Y, #params=6.5M x 3, training_modalities=3 (raw, c23, c40)2023.09 | 87.63 | 99.05 | 96.72 | 98.16 | 97.95 | 96.7 | 98.54 | 96.39 |