Table 3. Accuracies of FaceForensics++ (HQ) (ACC).

Model	Deepfake (%)	Face2Face (%)	FaceSwap (%)	NeuralTextures (%)
SPSL [26]	93.5	86.0	92.3	76.8
CAG-MobileViT [18]	93.7	94.1	96.3	87.9
XceptionNet [4]	96.4	86.9	90.3	80.7
Fre-Xception [22]	95.5	93.5	95.2	90.4
IR-Capsule [27]	97.3	81.8	94.2	79.3
TSFF-Net (ours)	97.7	91.0	98.9	90.0