List your company
ResearchFormerly WizWand

Deepfake Detection on DFD

0.9922AUCAUNet
0.8755120.9058060.93610.966394Mar 2, 2022Nov 14, 2022Jul 30, 2023Apr 12, 2024Dec 26, 2024Sep 9, 2025May 25, 2026
Updated 5mo ago

Evaluation results

MethodLinks
AUNetVenue=CVPR23, Training set (Real)=true, Training set (Fake)=false2025.030.9922
AUNetInput=Video, Backbone=Xception+ART, Training Dataset=FF++2025.080.992
PCL + I2GInput Type=Frame, Training Set (Real)=true, Training Set (Fake)=true2022.040.9907
CADDMVenue=CVPR23, Training set (Real)=true, Training set (Fake)=false2025.030.9903
RAEInput=Frame, Backbone=ViT-B/16, Training Dataset=FF++2025.080.99
FakeSTormerProcessing level=Video-level2026.050.985
AltFreezingInput=Video, Backbone=3D ResNet-50, Training Dataset=FF++2025.080.985
LAA-Net (w/ SBI)Venue=CVPR24, Training set (Real)=true, Training set (Fake)=false2025.030.9843
LAA-NetInput=Frame, Backbone=EFNB4, Training Dataset=FF++2025.080.984
SpInShieldProcessing level=Video-level2026.050.981
M2F2-Det (w/ SBI)Training set (Real)=true, Training set (Fake)=false2025.030.977
EFNB4 + SBIsInput Type=Frame, Training Set (Real)=true, Training Set (Fake)=false2022.040.9756
SBIVenue=CVPR22, Training set (Real)=true, Training set (Fake)=false2025.030.9756
HAADTraining Dataset=FF++, Evaluation Protocol=Cross-dataset Evaluation2026.050.975
ForAdaTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.040.972
SePLTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.040.972
ForAdaInput=Frame, Backbone=CLIP ViT-L/14, Training Dataset=FF++2025.080.972
NACOProcessing level=Video-level2026.050.97
GenD (CLIP)Input=Frame, Backbone=CLIP ViT-L/14, Training Dataset=FF++2025.080.97
HybridTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.040.967
GenD (DINO)Input=Frame, Backbone=DINOv3 ViT-L/16, Training Dataset=FF++2025.080.966
OursTraining Dataset=FF++ c23, Evaluation Protocol=Protocol-12024.080.965
EffortTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.040.965
P&PInput=Video, Backbone=CLIP ViT-L/14, Training Dataset=FF++2025.080.965
EffortInput=Frame, Backbone=CLIP ViT-L/14, Training Dataset=FF++2025.080.965
GenD (PE)Input=Frame, Backbone=PEcoreL, Training Dataset=FF++2025.080.965
FACTOREvaluation Protocol=Ref. Set, Reference Set Source=50% subset of training videos2023.110.963
FakeRadarBackbone=Vision Transformer, Input Type=Video2025.120.962
StyleFlowBackbone=3D ResNet, Input Type=Video2025.120.961
SLFProcessing level=Video-level2026.050.961
StyleDFDInput=Video, Backbone=3D ResNet-50, Training Dataset=FF++2025.080.961
X^2-DFDTraining Dataset=FF++, Evaluation Protocol=Cross-dataset Evaluation, Re-implemented=true2026.050.959
Xcep-c23Evaluation Protocol=Open-set2022.030.956
LSDAVenues=CVPR 2024, Training Dataset=FF++ c23, Evaluation Protocol=Protocol-12024.080.956
LSDABackbone=EfficientNet, Input Type=Frame2025.120.956
UDDInput=Frame, Backbone=CLIP ViT-B/16, Training Dataset=FF++2025.080.955
Face X-ray + BIInput Type=Frame, Training Set (Real)=true, Training Set (Fake)=true, Setting=Variant 22022.040.954
Face X-RayBackbone=HRNet, Input Type=Frame2025.120.954
CDFATraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.040.954
CDFAProcessing level=Frame-level2026.050.954
CFMVenues=TIFS 2023, Training Dataset=FF++ c23, Evaluation Protocol=Protocol-12024.080.952
EffortTraining Dataset=FF++, Evaluation Protocol=Cross-dataset Evaluation, Re-implemented=true2026.050.951
UIA-ViTBackbone=Vision Transformer, Input Type=Frame2025.120.947
UCFVenue=ICCV23, Training set (Real)=true, Training set (Fake)=true2025.030.945
UCFBackbone=Xception, Input Type=Frame2025.120.945
FTCNInput Type=Video, Training Set (Real)=true, Training Set (Fake)=true, experiment_type=official code2022.040.944
Face X-rayEvaluation Protocol=Open-set2022.030.9414
VLFFDTraining Dataset=FF++, Evaluation Protocol=Cross-dataset Evaluation, Re-implemented=true2026.050.94
A1 multi-region: mouth + skinBackbone=DINOv3 ViT-L/16, Region=mouth + skin, Evaluation protocol=zero-shot2026.050.9391
IIDVenues=CVPR 2023, Training Dataset=FF++ c23, Evaluation Protocol=Protocol-12024.080.939
IIDTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.040.939
IIDTraining Dataset=FF++, Evaluation Protocol=Cross-dataset Evaluation2026.050.939
IIDProcessing level=Frame-level2026.050.939
AltFreezingBackbone=3D ResNet, Input Type=Video2025.120.937
Face X-ray + BIInput Type=Frame, Training Set (Real)=true, Training Set (Fake)=true, Setting=Variant 12022.040.9347
ICT-RefEvaluation Protocol=Open-set2022.030.9317
Ours (Mouth)Backbone=Frozen2026.050.93
A1 single region: noseBackbone=DINOv3 ViT-L/16, Region=nose, Evaluation protocol=zero-shot2026.050.9299
A1 single region: mouthBackbone=DINOv3 ViT-L/16, Region=mouth, Evaluation protocol=zero-shot2026.050.9299
DCA EMNBackbone=DINOv3 ViT-L/16, Feature dimensionality=3,072 dims, Region=EMN, Evaluation protocol=zero-shot2026.050.9289
DFD-FCGProcessing level=Video-level2026.050.928
FIA-USATraining Dataset=FF++, Evaluation Protocol=Cross-dataset Evaluation, Re-implemented=true2026.050.926
DAIDTraining Dataset=FF++, Evaluation Protocol=Cross-dataset Evaluation, Re-implemented=true2026.050.925
DCLVenues=AAAI 2022, Training Dataset=FF++ c23, Evaluation Protocol=Protocol-12024.080.921
PTFDProcessing level=Video-level2026.050.92
FRDMInput Type=Frame, Training Set (Real)=true, Training Set (Fake)=true2022.040.919
DSP-FWAEvaluation Protocol=Open-set2022.030.9099
MSBA-CLIPTraining Dataset=FF++ (C23), Evaluation Level=Frame-level2026.020.9093
SLADDTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.040.904
SLADDProcessing level=Frame-level2026.050.904
AltFreezingProcessing level=Video-level2026.050.902
SpInShieldAttack Type=Spectral Tilt2026.050.9018
ProDetTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.040.901
ProDetProcessing level=Frame-level2026.050.901
SpInShieldAttack Type=SNR-based Noise2026.050.8948
HFRTraining Dataset=FF++ [7]2026.030.8941
Face2PartsBackbone=Fine-tuned2026.050.894
SRI-NetEvaluation Granularity=Frame-level2026.020.893
LRLInput Type=Frame, Training Set (Real)=true, Training Set (Fake)=true2022.040.8924
LocalRLVenue=AAAI21, Training set (Real)=true, Training set (Fake)=false2025.030.8924
TALLProcessing level=Video-level2026.050.892
RECCEVenues=CVPR 2022, Training Dataset=FF++ c23, Evaluation Protocol=Protocol-12024.080.891
RECCETraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.040.891
RECCETraining Dataset=FF++, Evaluation Protocol=Cross-dataset Evaluation2026.050.891
Xcep-c0Evaluation Protocol=Open-set2022.030.8905
A1 single region: eyesBackbone=DINOv3 ViT-L/16, Region=eyes, Evaluation protocol=zero-shot2026.050.887
Ours (Nose)Backbone=Frozen2026.050.887
SRMVenues=CVPR 2021, Training Dataset=FF++ c23, Evaluation Protocol=Protocol-12024.080.885
SRMTraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.040.885
SRMTraining Dataset=FF++, Evaluation Protocol=Cross-dataset Evaluation2026.050.885
COREVenues=CVPRW 2022, Training Dataset=FF++ c23, Evaluation Protocol=Protocol-12024.080.882
CORETraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.040.882
CORETraining Dataset=FF++, Evaluation Protocol=Cross-dataset Evaluation2026.050.882
DSMTraining Dataset=FF++ [7], Publication Year=20252026.030.8814
LSDATraining Dataset=FaceForensics++ (c23), Evaluation Protocol=Cross-dataset2026.040.881
LSDATraining Dataset=FF++, Evaluation Protocol=Cross-dataset Evaluation2026.050.881
LSDAProcessing level=Frame-level2026.050.881
LSDAVenue=CVPR’24, Evaluation Granularity=Frame-level2026.020.88
SFGDTraining Dataset=FF++ [7], Publication Year=20232026.030.88
LSDATraining Dataset=FF++ [7], Publication Year=20242026.030.88