List your company
ResearchFormerly WizWand

Deepfake Detection on CDF v2 (video-level AUC)

96Video-level AUCGenD (CLIP)
81.85685.52889.292.872Aug 8, 2025Sep 23, 2025Nov 8, 2025Dec 24, 2025Feb 8, 2026Mar 26, 2026May 11, 2026
Updated 6mo ago

Evaluation results

MethodLinks
GenD (CLIP)Input=Frame, Backbone=CLIP ViT-L/14, Training Dataset=FF++2025.0896
LOGERInput=Frame, Backbone=DINOv3+MetaCLIP22026.0495.9
GenD (PE)Input=Frame, Backbone=PEcoreL, Training Dataset=FF++2025.0895.8
GenD-PETraining Data=FF++, Evaluation Protocol=Video-level2026.0595.8
ForAdaInput=Frame, Backbone=CLIP ViT-L/142026.0495.7
ForAdaInput=Frame, Backbone=CLIP ViT-L/14, Training Dataset=FF++2025.0895.7
ForAdaTraining Data=FF++, Evaluation Protocol=Video-level2026.0595.7
EffortInput=Frame, Backbone=CLIP ViT-L/142026.0495.6
EffortInput=Frame, Backbone=CLIP ViT-L/14, Training Dataset=FF++2025.0895.6
RAEInput=Frame, Backbone=ViT-B/162026.0495.5
RAEInput=Frame, Backbone=ViT-B/16, Training Dataset=FF++2025.0895.5
LAA-NetInput=Frame, Backbone=EfficientNet-B42026.0495.4
LAA-NetInput=Frame, Backbone=EFNB4, Training Dataset=FF++2025.0895.4
FS-VFMTraining Data=FF++, Evaluation Protocol=Video-level2026.0595.4
DFD-FCGInput=Video, Backbone=CLIP ViT-L/142026.0495
DFD-FCGInput=Video, Backbone=CLIP ViT-L/14, Training Dataset=FF++2025.0895
P&PInput=Video, Backbone=CLIP ViT-L/142026.0494.7
P&PInput=Video, Backbone=CLIP ViT-L/14, Training Dataset=FF++2025.0894.7
SBIInput=Frame, Backbone=EfficientNet-B42026.0493.2
SBIInput=Frame, Backbone=EFNB4, Training Dataset=FF++2025.0893.2
EffortTraining Data=FF++, Evaluation Protocol=Video-level2026.0593.2
UDDInput=Frame, Backbone=CLIP ViT-B/162026.0493.1
UDDInput=Frame, Backbone=CLIP ViT-B/16, Training Dataset=FF++2025.0893.1
AUNetInput=Video, Backbone=Xception2026.0492.8
AUNetInput=Video, Backbone=Xception+ART, Training Dataset=FF++2025.0892.8
ProDetInput=Frame, Backbone=EfficientNet-B42026.0492.5
ProDetInput=Frame, Backbone=EFNB4, Training Dataset=FF++2025.0892.5
GenDInput=Frame, Backbone=DINOv3 ViT-L/162026.0492.2
GenD (DINO)Input=Frame, Backbone=DINOv3 ViT-L/16, Training Dataset=FF++2025.0892.2
TALL++Input=Video, Backbone=Swin-B2026.0492
TALL++Input=Video, Backbone=Swin-B, Training Dataset=FF++2025.0892
LSDAInput=Frame, Backbone=EfficientNet-B42026.0491.1
LSDAInput=Frame, Backbone=EFNB4, Training Dataset=FF++2025.0891.1
FSFMTraining Data=FF++, Evaluation Protocol=Video-level2026.0590.3
BlenDTraining Data=25k ScaleDF (real) + SBI fakes, Evaluation Protocol=Video-level2026.0590
AltFreezingInput=Video, Backbone=3D ResNet-502026.0489.5
NACOInput=Video, Backbone=ViT-B/162026.0489.5
AltFreezingInput=Video, Backbone=3D ResNet-50, Training Dataset=FF++2025.0889.5
NACOInput=Video, Backbone=ViT-B/16, Training Dataset=FF++2025.0889.5
StyleDFDInput=Video, Backbone=3D ResNet-50, Training Dataset=FF++2025.0889
FSBITraining Data=FF++ with pseudo-fakes, Evaluation Protocol=Video-level2026.0588.2
FTCNInput=Video, Backbone=3D ResNet-502026.0486.9
RealForensicsInput=Video, Backbone=Modified CSN2026.0486.9
FTCNInput=Video, Backbone=3D ResNet-50, Training Dataset=FF++2025.0886.9
RealForensicsInput=Video, Backbone=Modified CSN, Training Dataset=FF++2025.0886.9
SBITraining Data=FF++ with pseudo-fakes, Evaluation Protocol=Video-level2026.0586.2
LipForensicsInput=Video, Backbone=ResNet-182026.0482.4
LipForensicsInput=Video, Backbone=ResNet-18, Training Dataset=FF++2025.0882.4