Deepfake Detection on FF++
99.9AUCLRNet
Evaluation Results
| Method | Links | |
|---|---|---|
| LRNetSize=0.18 M, Data Augmentation=false, Training Dataset=FF++2021.04 | 99.9 | |
| TALLBackbone=Swin Transformer, Input Type=Video2025.12 | 99.9 | |
| PCL+I2GBackbone=Resnet, Input Type=Frame2025.12 | 99.8 | |
| FTCNBackbone=3D ResNet, Input Type=Video2025.12 | 99.8 | |
| RealForensicsBackbone=3D ResNet, Input Type=Video2025.12 | 99.8 | |
| NACOBackbone=Vision Transformer, Input Type=Video2025.12 | 99.8 | |
| XceptionSize=20.8 M, Data Augmentation=false, Training Dataset=FF++2021.04 | 99.7 | |
| Nirkin et al.Evaluation Protocol=Close-set2022.03 | 99.7 | |
| AltFreezingBackbone=3D ResNet, Input Type=Video2025.12 | 99.7 | |
| LTTDBackbone=Vision Transformer, Input Type=Video2025.12 | 99.5 | |
| XceptionBackbone=Xception, Input Type=Frame2025.12 | 99.3 | |
| UIA-ViTBackbone=Vision Transformer, Input Type=Frame2025.12 | 99.3 | |
| Xcep-c0Evaluation Protocol=Close-set2022.03 | 99.26 | |
| StyleFlowBackbone=3D ResNet, Input Type=Video2025.12 | 99.1 | |
| FakeRadarBackbone=Vision Transformer, Input Type=Video2025.12 | 99.1 | |
| ICT-RefEvaluation Protocol=Open-set2022.03 | 98.56 | |
| Xcep-c23Evaluation Protocol=Close-set2022.03 | 98.54 | |
| Face X-rayEvaluation Protocol=Close-set2022.03 | 98.44 | |
| SLADDBackbone=Xception, Input Type=Frame2025.12 | 98.4 | |
| CNN+RNNSize=24.3 M, Data Augmentation=false, Training Dataset=FF++2021.04 | 98.3 | |
| Face X-RayBackbone=HRNet, Input Type=Frame2025.12 | 97.8 | |
| CapsuleSize=15 M, Data Augmentation=false, Training Dataset=FF++2021.04 | 96.6 | |
| CapsuleEvaluation Protocol=Close-set2022.03 | 96.5 | |
| DSP-FWASize=28 M, Data Augmentation=true, Training Dataset=Unpub.2021.04 | 93 | |
| FFDEvaluation Protocol=Open-set2022.03 | 92.32 | |
| ICTEvaluation Protocol=Open-set2022.03 | 90.22 | |
| Meso4Size=0.03 M, Data Augmentation=false, Training Dataset=Unpub.2021.04 | 84.7 | |
| DSP-FWAEvaluation Protocol=Open-set2022.03 | 81.9 | |
| FWASize=26 M, Data Augmentation=true, Training Dataset=Unpub.2021.04 | 80.1 | |
| FWAEvaluation Protocol=Open-set2022.03 | 74.82 | |
| Patch-ForenEvaluation Protocol=Open-set2022.03 | 73.75 | |
| Multi-taskEvaluation Protocol=Close-set2022.03 | 72.23 | |
| CNNDetectEvaluation Protocol=Open-set2022.03 | 71.08 | |
| MesoInc4Evaluation Protocol=Close-set2022.03 | 63.41 |