FD²Foremer: Thinking Face Forgery Detection in Midfrequency Geometry Details

Table 4

Results (%) of distillation for a sequence of selected . All trained with the image and the displacement map as inputs on low-quality FF++. The metric is ACC.

Teacher	Student		Testing data (ACC)
			FF++(LQ)	DFDC
DF^2Former(swin)	DF^2Former(meta)	0	81.85	85.43
DF^2Former(swin)	DF^2Former(meta)	0.2	82.73	86.72
DF^2Former(swin)	DF^2Former(meta)	0.4	82.86	85.77
DF^2Former(swin)	DF^2Former(meta)	0.6	82.83	86.58
DF^2Former(swin)	DF^2Former(meta)	0.8	82.51	86.27
DF^2Former(swin)	DF^2Former(meta)	1	82.37	86.08

The metric is ACC. Best results are shown in bold.