Deepfake Detection on Celeb-DF (test)
100AccuracyMARE
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MARE2026.01 | 100 | 100 | — | |
| MAREEvaluation Setup=intra-dataset2026.01 | 99.61 | 99.98 | — | |
| M2F2Venue=CVPR20252026.01 | 99.61 | 99.99 | — | |
| ExpDVenue=AAAI20242026.01 | 99.42 | 99.99 | — | |
| M2F2Venue=CVPR25, Evaluation Setup=intra-dataset2026.01 | 98.98 | 99.92 | — | |
| EfficientNetVenue=ICML20192026.01 | 98.65 | 99.87 | — | |
| RECCEVenue=CVPR22, Evaluation Setup=intra-dataset2026.01 | 98.59 | 99.94 | — | |
| MultiAttVenue=CVPR21, Evaluation Setup=intra-dataset2026.01 | 97.92 | 99.94 | — | |
| TALLVenue=ICCV23, Evaluation Setup=intra-dataset2026.01 | 97.57 | 98.55 | — | |
| RDBC2026.03 | 96.03 | — | — | |
| ResNet50# Params=25M, Layers=All2026.02 | 95.23 | — | — | |
| Mapping Networks# Params=2048, Layers=All2026.02 | 95.1 | — | — | |
| EffortVenue=ICML’25, Re-implemented=true2026.03 | 93.68 | — | — | |
| FIA-USAVenue=NeurIPS’25, Re-implemented=true2026.03 | 93.52 | — | — | |
| SBI*Venue=CVPR’22, Re-implemented=true, Focuses on augmenting real images=true2026.03 | 93.18 | — | — | |
| SFICVenue=TIFS20242026.01 | 92.86 | 97.76 | — | |
| FreqBlender*Venue=NeurIPS’24, Re-implemented=true, Focuses on augmenting real images=true2026.03 | 92.65 | — | — | |
| Mapping Networks# Params=1024, Layers=L-4, FC2026.02 | 92.1 | — | — | |
| CADDMVenue=CVPR’23, Re-implemented=true2026.03 | 91.63 | — | — | |
| ResNet50# Params=17M, Layers=L-4, FC2026.02 | 91.11 | — | — | |
| Mapping CNN (CNN1)# Params=2792, Training Strategy=Layer-wise training, Base Architecture=CNN12026.02 | 89.98 | — | — | |
| VLFFDVenue=CVPR’25, Re-implemented=true2026.03 | 89.8 | — | — | |
| GRUVenue=CVPR’242026.03 | 89 | — | — | |
| Mapping CNN (CNN1)# Params=2048, Training Strategy=Full Network training, Base Architecture=CNN12026.02 | 88.88 | — | — | |
| Mapping CNN (CNN1)# Params=1956, Training Strategy=Layer-wise training, Base Architecture=CNN12026.02 | 88.78 | — | — | |
| Mapping CNN (CNN2)# Params=2688, Training Strategy=Layer-wise training, Base Architecture=CNN22026.02 | 86.09 | — | — | |
| Mapping CNN (CNN2)# Params=2048, Training Strategy=Full Network training, Base Architecture=CNN22026.02 | 85.9 | — | — | |
| Mapping CNN (CNN2)# Params=1872, Training Strategy=Layer-wise training, Base Architecture=CNN22026.02 | 84.54 | — | — | |
| ProDetVenue=NeurIPS’242026.03 | 84.48 | — | — | |
| Mapping CNN (CNN1)# Params=1024, Training Strategy=Full Network training, Base Architecture=CNN12026.02 | 83.92 | — | — | |
| CNN1# Params=537,9942026.02 | 83.13 | — | — | |
| UCFVenue=ICCV’23, Re-implemented=true2026.03 | 81.9 | — | — | |
| COREVenue=CVPRW’22, Re-implemented=true2026.03 | 79.45 | — | — | |
| CNN2# Params=108,6182026.02 | 79.03 | — | — | |
| Mapping CNN (CNN2)# Params=1024, Training Strategy=Full Network training, Base Architecture=CNN22026.02 | 78.83 | — | — | |
| FoCusVenue=TIFS’242026.03 | 76.13 | — | — | |
| Face X-rayVenue=CVPR’20, Re-implemented=true2026.03 | 74.76 | — | — | |
| Qiao et al.Venue=TPAMI’242026.03 | 70 | — | — | |
| RECCEVenue=CVPR’22, Re-implemented=true2026.03 | 69.71 | — | — | |
| EfficientNetVenue=ICML’19, Re-implemented=true2026.03 | 64.59 | — | — | |
| CORECondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | — | 79.4 | 0.19 | |
| DFGazeCondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | — | 67.8 | 0.308 | |
| DFILCondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | — | 66.5 | 0.322 | |
| F3-netCondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | — | 75.7 | 0.229 | |
| LR-GCNCondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | — | 90.6 | 0.083 | |
| MaskRelationCondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | — | 95 | 0.023 | |
| RECCECondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | — | 68.7 | 0.292 | |
| TALL-SwinCondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | — | 81.9 | 0.062 | |
| TGNVenue=TIFS24, Evaluation Setup=intra-dataset2026.01 | — | 99.59 | — | |
| UCFCondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | — | 82.4 | 0.158 | |
| XceptionCondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | — | 73.7 | 0.235 |