Deepfake Detection on CelebDF v2
0.999AUCDefakerOne
Evaluation results
| Method | Links | |
|---|---|---|
| DefakerOneModel Category=MLLMs2026.05 | 0.999 | |
| VLAForgeVenue=-2026.03 | 0.968 | |
| LAKANInput Type=Frame, Training Set (Real)=true, Training Set (Fake)=false2025.10 | 0.9663 | |
| SePLTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.96 | |
| ForAdaVenue=CVPR’252026.03 | 0.957 | |
| ForAdaTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.957 | |
| Forensics Adapter2026.04 | 0.957 | |
| EffortVenue=ICML’252026.03 | 0.956 | |
| EffortTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.956 | |
| RAEVenue=ECCV 2024, Input Type=Frame, Training Set (Real)=true, Training Set (Fake)=false2025.10 | 0.955 | |
| LAA-NetVenue=CVPR 2024, Input Type=Frame, Training Set (Real)=true, Training Set (Fake)=false2025.10 | 0.954 | |
| LAA-NETVenue=CVPR’242026.03 | 0.954 | |
| DFD-FCGVenue=CVPR 2025, Input Type=Video, Training Set (Real)=true, Training Set (Fake)=true2025.10 | 0.95 | |
| FCGVenue=CVPR’252026.03 | 0.95 | |
| SFAM2026.04 | 0.95 | |
| DFD-FCGProcessing level=Video-level2026.05 | 0.949 | |
| SpInShieldProcessing level=Video-level2026.05 | 0.948 | |
| Yan et al.Venue=CVPR’252026.03 | 0.947 | |
| FreqBlenderVenue=NeurIPS 2024, Input Type=Frame, Training Set (Real)=true, Training Set (Fake)=false2025.10 | 0.9459 | |
| HybridTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.945 | |
| CADDMVenue=CVPR’232026.03 | 0.939 | |
| CDFAVenue=ECCV’242026.03 | 0.938 | |
| CDFATraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.938 | |
| CDFAProcessing level=Frame-level2026.05 | 0.938 | |
| SBIVenue=CVPR’222026.03 | 0.932 | |
| SBIVenue=CVPR 2022, Input Type=Frame, Training Set (Real)=true, Training Set (Fake)=false2025.10 | 0.9318 | |
| UDDVenue=AAAI 2025, Input Type=Frame, Training Set (Real)=true, Training Set (Fake)=true2025.10 | 0.931 | |
| UDDTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.931 | |
| AUNetVenue=CVPR 2023, Input Type=Frame, Training Set (Real)=true, Training Set (Fake)=false2025.10 | 0.9277 | |
| ProDetVenue=Neurips’242026.03 | 0.926 | |
| ProDetTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.926 | |
| ProDetProcessing level=Frame-level2026.05 | 0.926 | |
| FakeSTormerProcessing level=Video-level2026.05 | 0.924 | |
| TALL++Venue=IJCV 2024, Input Type=Video, Training Set (Real)=true, Training Set (Fake)=true2025.10 | 0.9196 | |
| PTFDProcessing level=Video-level2026.05 | 0.914 | |
| CDFAModel Category=Vision Models2026.05 | 0.914 | |
| DCA EMNBackbone=DINOv3 ViT-L/16, Feature dimensionality=3,072 dims, Region=EMN, Evaluation protocol=zero-shot2026.05 | 0.9106 | |
| Ours (Mouth)Backbone=Frozen2026.05 | 0.905 | |
| A1 multi-region: mouth + skinBackbone=DINOv3 ViT-L/16, Region=mouth + skin, Evaluation protocol=zero-shot2026.05 | 0.8991 | |
| RepDFDVenue=AAAI’252026.03 | 0.899 | |
| RepDFDTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.899 | |
| LSDAVenue=CVPR’242026.03 | 0.898 | |
| NACOVenue=ECCV 2024, Input Type=Video, Training Set (Real)=true, Training Set (Fake)=true2025.10 | 0.895 | |
| NACOProcessing level=Video-level2026.05 | 0.895 | |
| SAMVenue=CVPR’242026.03 | 0.89 | |
| SLFProcessing level=Video-level2026.05 | 0.89 | |
| AltFreezingProcessing level=Video-level2026.05 | 0.89 | |
| A1 single region: noseBackbone=DINOv3 ViT-L/16, Region=nose, Evaluation protocol=zero-shot2026.05 | 0.8886 | |
| SBITraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.886 | |
| SBIProcessing level=Frame-level2026.05 | 0.886 | |
| Effort2026.04 | 0.885 | |
| EffortModel Category=Vision Models2026.05 | 0.882 | |
| SRI-NetEvaluation Granularity=Frame-level2026.02 | 0.875 | |
| LSDATraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.875 | |
| LSDAProcessing level=Frame-level2026.05 | 0.875 | |
| SeeABLEVenue=ICCV 2023, Input Type=Frame, Training Set (Real)=true, Training Set (Fake)=false2025.10 | 0.873 | |
| SeeABLEVenue=ICCV’232026.03 | 0.873 | |
| TALLProcessing level=Video-level2026.05 | 0.871 | |
| Ours (Nose)Backbone=Frozen2026.05 | 0.87 | |
| FTCNTraining Dataset=FaceForensics++, Evaluation Protocol=Cross-dataset generalization2022.01 | 0.869 | |
| RealForensicsTraining Dataset=FaceForensics++, Evaluation Protocol=Cross-dataset generalization2022.01 | 0.869 | |
| FTCNProcessing level=Video-level2026.05 | 0.869 | |
| FIA-USAVenue=ARXIV’25, Evaluation Granularity=Frame-level2026.02 | 0.867 | |
| EDVenue=AAAI’24, Evaluation Granularity=Frame-level2026.02 | 0.864 | |
| A1 single region: mouthBackbone=DINOv3 ViT-L/16, Region=mouth, Evaluation protocol=zero-shot2026.05 | 0.858 | |
| Face2PartsBackbone=Fine-tuned2026.05 | 0.853 | |
| ProDetVenue=NIPS’24, Evaluation Granularity=Frame-level2026.02 | 0.842 | |
| SRMTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.84 | |
| IIDVenue=CVPR’232026.03 | 0.838 | |
| IIDTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.838 | |
| IIDProcessing level=Frame-level2026.05 | 0.838 | |
| FadaVenue=CVPR’25, Evaluation Granularity=Frame-level2026.02 | 0.837 | |
| SLADDTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.837 | |
| UCFTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.837 | |
| UCF2026.04 | 0.837 | |
| SLADDProcessing level=Frame-level2026.05 | 0.837 | |
| UCFProcessing level=Frame-level2026.05 | 0.837 | |
| TALLVenue=ICCV’232026.03 | 0.831 | |
| A1 single region: eyesBackbone=DINOv3 ViT-L/16, Region=eyes, Evaluation protocol=zero-shot2026.05 | 0.831 | |
| LSDAVenue=CVPR’24, Evaluation Granularity=Frame-level2026.02 | 0.83 | |
| LipForensicsTraining Dataset=FaceForensics++, Evaluation Protocol=Cross-dataset generalization2022.01 | 0.824 | |
| AFWProcessing level=Video-level2026.05 | 0.824 | |
| LipForensicsBackbone=Task-specific2026.05 | 0.824 | |
| RECCETraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.823 | |
| RECCE2026.04 | 0.823 | |
| FakeVLMModel Category=MLLMs2026.05 | 0.819 | |
| Xception2026.04 | 0.816 | |
| CLIP2026.04 | 0.814 | |
| SBIVenue=CVPR’22, Evaluation Granularity=Frame-level2026.02 | 0.813 | |
| CORETraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.809 | |
| Efficientnet-B42026.04 | 0.808 | |
| SPSLTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.799 | |
| F3NetVenue=ECCV’20, Evaluation Granularity=Frame-level2026.02 | 0.798 | |
| Face X-rayTraining Dataset=FaceForensics++, Evaluation Protocol=Cross-dataset generalization2022.01 | 0.795 | |
| F3NetTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.04 | 0.789 | |
| F3Net2026.04 | 0.789 | |
| DINOv2 ViT-L/14 DCA EMNBackbone=DINOv2 ViT-L/14, Region=EMN, Evaluation protocol=zero-shot2026.05 | 0.785 | |
| SPSLVenue=CVPR’20, Evaluation Granularity=Frame-level2026.02 | 0.765 | |
| SFDGVenue=CVPR’232026.03 | 0.758 | |
| Multi-taskTraining Dataset=FaceForensics++, Evaluation Protocol=Cross-dataset generalization2022.01 | 0.757 |