Deepfake Detection on DFDC (test)
99.77AUCMARE
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| MARE2026.01 | 99.77 | — | 98.83 | — | — | — | |
| ExpDVenue=AAAI20242026.01 | 99.3 | — | 96.09 | — | — | — | |
| M2F2Venue=CVPR20252026.01 | 99.16 | — | 95.31 | — | — | — | |
| MAREEvaluation Setup=intra-dataset2026.01 | 98.3 | — | 93.75 | — | — | — | |
| ViT with distillationParameters=373M2021.07 | 97.8 | 91.9 | — | — | — | — | |
| Selim Efficient Net B7Parameters=462M, Ensemble=6 networks2021.07 | 97.2 | 90.6 | — | — | — | — | |
| SFAMTraining Dataset=UADFV2026.04 | 96.8 | — | — | — | — | — | |
| TGNVenue=TIFS24, Evaluation Setup=intra-dataset2026.01 | 96.24 | — | — | — | — | — | |
| Conv. Cross ViT Eff.Net B0 - VotingParameters=101M, Backbone=EfficientNet B0, Aggregation strategy=Voting2021.07 | 95.1 | 88 | — | — | — | — | |
| Conv. Cross ViT Eff.Net B0 - AvgParameters=101M, Backbone=EfficientNet B0, Aggregation strategy=Avg2021.07 | 94.7 | 85.6 | — | — | — | — | |
| EfficientNetVenue=ICML20192026.01 | 93.84 | — | 85.94 | — | — | — | |
| SFAMTraining Dataset=CDv12026.04 | 93.8 | — | — | — | — | — | |
| XceptionTraining Dataset=UADFV2026.04 | 93.6 | — | — | — | — | — | |
| Conv. Cross ViT Wodajo CNNParameters=142M, Backbone=Wodajo CNN2021.07 | 92.5 | 84.5 | — | — | — | — | |
| Efficient ViTParameters=109M2021.07 | 91.9 | 83.8 | — | — | — | — | |
| GenDFCategory=Fine-tuned2025.12 | 91.67 | — | 84.05 | — | — | — | |
| RECCEVenue=CVPR22, Evaluation Setup=intra-dataset2026.01 | 91.33 | — | 81.2 | — | — | — | |
| SFICVenue=TIFS20242026.01 | 91.28 | — | 84.38 | — | — | — | |
| RECCE*Category=Custom-made2025.12 | 90.58 | — | 81.2 | — | — | — | |
| DE-AdapterCategory=Fine-tuned2025.12 | 90.37 | — | 81.59 | — | — | — | |
| MultiAttCategory=Custom-made2025.12 | 90.32 | — | 76.81 | — | — | — | |
| MaskRelationCondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | 90.2 | — | — | — | 0.071 | — | |
| SFAMTraining Dataset=CDv22026.04 | 90 | — | — | — | — | — | |
| UIA-ViT*Category=Custom-made2025.12 | 89.98 | — | 80.23 | — | — | — | |
| Add-NetCategory=Custom-made2025.12 | 89.85 | — | 78.71 | — | — | — | |
| RFMCategory=Custom-made2025.12 | 89.75 | — | 80.83 | — | — | — | |
| XceptionCategory=Pre-trained2025.12 | 89.5 | — | 79.35 | — | — | — | |
| F3-NetCategory=Custom-made2025.12 | 88.39 | — | 76.17 | — | — | — | |
| Effort + TSRLRetrained=true, TSRL Integration=true2026.03 | 88.2 | — | — | — | — | — | |
| EffortRetrained=true, TSRL Integration=false2026.03 | 86.3 | — | — | — | — | — | |
| EFNB4 + SBIsInput Type=Frame, Training Set (Real)=true, Training Set (Fake)=false2022.04 | 86.15 | — | — | — | — | — | |
| Convolutional ViTParameters=89M2021.07 | 84.3 | 77 | — | — | — | — | |
| OursVenue=N/A2024.08 | 84.3 | — | — | — | — | — | |
| DFAAnalysis Level=Video-level2026.03 | 83.6 | — | — | — | — | 25.1 | |
| CDFARetrained=false, TSRL Integration=false2026.03 | 83 | — | — | — | — | — | |
| CAST-B0Year=2025, Training Dataset=FF++ (HQ)2025.06 | 81.25 | — | — | — | — | — | |
| Face X-ray + BIInput Type=Frame, Training Set (Real)=true, Training Set (Fake)=true, Setting=Variant 22022.04 | 80.92 | — | — | — | — | — | |
| Face-X-RayTraining Dataset=FF++2026.01 | 80.92 | — | — | — | — | — | |
| AVoiD-DF2023.10 | 80.6 | — | — | — | — | — | |
| UCFCondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | 80.5 | — | — | — | 0.177 | — | |
| SFAMTraining Dataset=DFDCP2026.04 | 80.5 | — | — | — | — | — | |
| LTTDtrained_on=FF++, evaluation_level=video-level2022.07 | 80.4 | — | — | — | — | — | |
| LTTDVenue=NeurIPS 20222024.08 | 80.4 | — | — | — | — | — | |
| XceptionTraining Dataset=CDv12026.04 | 80.3 | — | — | — | — | — | |
| FRDMInput Type=Frame, Training Set (Real)=true, Training Set (Fake)=true2022.04 | 79.7 | — | — | — | — | — | |
| GFFYear=2021, Training Dataset=FF++ (HQ)2025.06 | 79.7 | — | — | — | — | — | |
| Efficient-ViTAnalysis Level=Video-level2026.03 | 78.8 | — | — | — | — | 30 | |
| CAST-B5Year=2025, Training Dataset=FF++ (HQ)2025.06 | 78.39 | — | — | — | — | — | |
| ResNeXt-LSTMAnalysis Level=Video-level2026.03 | 77.6 | — | — | — | — | 31.9 | |
| Ours†Training Dataset=FF++, Feature Extracting Backbone=Swin Transformer2026.01 | 77.5 | — | — | — | — | — | |
| XceptionAnalysis Level=Video-level2026.03 | 77.4 | — | — | — | — | 32.5 | |
| SLADDRetrained=false, TSRL Integration=false2026.03 | 77.2 | — | — | — | — | — | |
| CViTAnalysis Level=Video-level2026.03 | 77 | — | — | — | — | 32.3 | |
| LR-GCNCondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | 76.9 | — | — | — | 0.22 | — | |
| TALL-SwinVenue=ICCV 20232024.08 | 76.8 | — | — | — | — | — | |
| CLIP + TSRLRetrained=true, TSRL Integration=true2026.03 | 76.8 | — | — | — | — | — | |
| TALLVenue=ICCV23, Evaluation Setup=intra-dataset2026.01 | 76.78 | — | — | — | — | — | |
| DCLYear=2022, Training Dataset=FF++ (HQ)2025.06 | 76.71 | — | — | — | — | — | |
| NACOVenue=ECCV 20242024.08 | 76.7 | — | — | — | — | — | |
| LRLInput Type=Frame, Training Set (Real)=true, Training Set (Fake)=true2022.04 | 76.53 | — | — | — | — | — | |
| LRLTraining Dataset=FF++2026.01 | 76.53 | — | — | — | — | — | |
| XceptionTraining Dataset=CDv22026.04 | 76.4 | — | — | — | — | — | |
| RealForensicsVenue=CVPR 20222024.08 | 75.9 | — | — | — | — | — | |
| CLIPRetrained=true, TSRL Integration=false2026.03 | 75.9 | — | — | — | — | — | |
| CORECondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | 75.7 | — | — | — | 0.227 | — | |
| Ours*Training Dataset=FF++, Feature Extracting Backbone=EfficientNet-B42026.01 | 75.27 | — | — | — | — | — | |
| SBI+EB4trained_on=FF++, evaluation_level=video-level2022.07 | 74.9 | — | — | — | — | — | |
| TALL-SwinCondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | 74.8 | — | — | — | 0.133 | — | |
| PCL + I2GInput Type=Frame, Training Set (Real)=true, Training Set (Fake)=true2022.04 | 74.37 | — | — | — | — | — | |
| FTCNInput Type=Video, Training Set (Real)=true, Training Set (Fake)=true2022.04 | 74 | — | — | — | — | — | |
| FTCN-TTtrained_on=FF++, evaluation_level=video-level2022.07 | 74 | — | — | — | — | — | |
| FTCNVenue=ICCV 20212024.08 | 74 | — | — | — | — | — | |
| UCFRetrained=true, TSRL Integration=false2026.03 | 74 | — | — | — | — | — | |
| MDS2023.10 | 73.8 | — | — | — | — | — | |
| UCF + TSRLRetrained=true, TSRL Integration=true2026.03 | 73.8 | — | — | — | — | — | |
| SFDGYear=2023, Training Dataset=FF++ (HQ)2025.06 | 73.64 | — | — | — | — | — | |
| LipForensicstrained_on=FF++, evaluation_level=video-level2022.07 | 73.5 | — | — | — | — | — | |
| LipForensicsVenue=CVPR 20212024.08 | 73.5 | — | — | — | — | — | |
| DAMInput Type=Video, Training Set (Real)=true, Training Set (Fake)=true2022.04 | 72.8 | — | — | — | — | — | |
| EFNB4 + SBIsInput Type=Frame, Training Set (Real)=true, Training Set (Fake)=false2022.04 | 72.42 | — | — | — | — | — | |
| SBIsTraining Dataset=FF++, Feature Extracting Backbone=EfficientNet-B42026.01 | 72.42 | — | — | — | — | — | |
| SPSLRetrained=false, TSRL Integration=false2026.03 | 72.4 | — | — | — | — | — | |
| MATTraining set=FF++2023.09 | 71.9 | — | 65.1 | — | — | — | |
| F3NetRetrained=false, TSRL Integration=false2026.03 | 71.8 | — | — | — | — | — | |
| SBIRetrained=false, TSRL Integration=false2026.03 | 71.7 | — | — | — | — | — | |
| MesoNetTraining set=FF++2023.09 | 71.4 | — | 60.3 | — | — | — | |
| CORE + TSRLRetrained=true, TSRL Integration=true2026.03 | 71.3 | — | — | — | — | — | |
| Face X-ray + BIInput Type=Frame, Training Set (Real)=true, Training Set (Fake)=true, Setting=Variant 12022.04 | 71.15 | — | — | — | — | — | |
| IIDRetrained=true, TSRL Integration=false2026.03 | 71.1 | — | — | — | — | — | |
| ProDetRetrained=true, TSRL Integration=false2026.03 | 71.1 | — | — | — | — | — | |
| SBIsTraining Dataset=FF++, Feature Extracting Backbone=Swin Transformer2026.01 | 71.08 | — | — | — | — | — | |
| FTCNInput Type=Video, Training Set (Real)=true, Training Set (Fake)=true, experiment_type=official code2022.04 | 71 | — | — | — | — | — | |
| SBIsTraining set=FF++2023.09 | 70.9 | — | 60.5 | — | — | — | |
| XceptionCondition=clean (no masking), Evaluation=cross-dataset, Training Dataset=FF++ (C23)2025.12 | 70.9 | — | — | — | 0.263 | — | |
| ProDet + TSRLRetrained=true, TSRL Integration=true2026.03 | 70.6 | — | — | — | — | — | |
| IID + TSRLRetrained=true, TSRL Integration=true2026.03 | 70.4 | — | — | — | — | — | |
| Fang & LinTraining set=FF++2023.09 | 70.2 | — | 63.2 | — | — | — | |
| LSDARetrained=false, TSRL Integration=false2026.03 | 70.1 | — | — | — | — | — | |
| Face X-rayYear=2020, Training Dataset=FF++ (HQ)2025.06 | 70 | — | — | — | — | — | |
| CNNAnalysis Level=Video-level2026.03 | 69.8 | — | — | — | — | 37.5 |