Table 4.

Result 2.2: cross domain performance.

Method		DCGAN	DeepFake	VQ-VAE2.0
Ours	Accuracy	0.931	0.897	0.809
	Precision	0.935	0.906	0.839
	Recall	0.994	0.994	0.994
	F1-Score	0.964	0.948	0.910
CNN+Self-attention (2020)	Accuracy	0.412	0.532	0.461
	Precision	0.628	0.679	0.648
	Recall	0.991	0.991	0.991
	F1-Score	0.769	0.806	0.783
Pupil regular recognition+ Boundary IoU score (2022)	Accuracy	0.921	0.874	0.722
	Precision	0.926	0.888	0.782
	Recall	0.995	0.995	0.995
	F1-Score	0.959	0.938	0.875
Dual-color spaces+improved Xception model (2021)	Accuracy	0.834	0.787	0.614
	Precision	0.859	0.824	0.720
	Recall	0.994	0.994	0.994
	F1-Score	0.920	0.901	0.835
MaskCNN+RAN (2022)	Accuracy	0.745	0.671	0.719
	Precision	0.795	0.750	0.779
	Recall	0.989	0.989	0.989
	F1-Score	0.881	0.853	0.871