FD²Foremer: Thinking Face Forgery Detection in Midfrequency Geometry Details

Table 3

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)	0.5	81.76	86.17
DF^2Former(swin)	DF^2Former(meta)	1	82.73	86.72
DF^2Former(swin)	DF^2Former(meta)	1.5	82.42	86.47
DF^2Former(swin)	DF^2Former(meta)	2	82.19	85.74
DF^2Former(swin)	DF^2Former(meta)	2.5	81.71	85.86

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