List your company
ResearchFormerly WizWand

Frame-level Deepfake Detection on Average (Celeb-DF, Wild Deepfake, DFDC-P, DFD, DiffSwap)

83.78AUCDiffusionFake (ViT-B)
72.225675.225378.22581.2247Oct 6, 2024
Updated 8mo ago

Evaluation results

MethodLinks
DiffusionFake (ViT-B)Backbone=ViT-B, Training Dataset=FF++(HQ)2024.1083.7825.17
DiffusionFake (En-b4)Backbone=EfficientNet-B4, Training Dataset=FF++(HQ)2024.1081.8825.97
DCLTraining Dataset=FF++(HQ), Reproduced Status=false2024.1080.427.73
SBITraining Dataset=FF++(HQ), Reproduced Status=true2024.1077.7728.81
ViT-BBackbone=ViT-B, Training Dataset=FF++(HQ), Reproduced Status=true2024.1077.4430.24
LTWTraining Dataset=FF++(HQ), Reproduced Status=false2024.1077.0730.39
RECCETraining Dataset=FF++(HQ), Reproduced Status=true2024.1076.3631.38
CADDMTraining Dataset=FF++(HQ), Reproduced Status=true2024.1076.2130.81
EfficientNet-B4 (En-b4)Backbone=EfficientNet-B4, Training Dataset=FF++(HQ), Reproduced Status=true2024.1075.831.19
GFFTraining Dataset=FF++(HQ), Reproduced Status=true2024.1075.4632.51
MATTraining Dataset=FF++(HQ), Reproduced Status=true2024.1075.1332.03
F3-NetTraining Dataset=FF++(HQ), Reproduced Status=true2024.1074.9632.92
XceptionBackbone=Xception, Training Dataset=FF++(HQ), Reproduced Status=false2024.1072.6733.48