Table 1 Comparison of detection...

Table 1

Comparison of detection performance of several methods on the first genreation datasets with AUC as the performance metric. The AUC results of DefakeHop++ in both frame-level and video-level are given. The best and the second-best results are shown in boldface and underbared, respectively. The AUC results of benchmarking methods are taken from [21] and the number of parameters are from https://keras.io/api/applications. Also, we use ^a to denote deep learning methods and ^b to denote non-deep-learning methods.

		1st Generation
Method	Model	UADFV	FF++	#param
Two-stream [44]	InceptionV3^a[33]	85.1%	70.1%	23.9M
Meso4 [1]	Designed CNN^a	84.3%	84.7%	28.0K
MesoInception4 [1]	Designed CNN^a	82.1%	83.0%	28.6K
HeadPose [37]	SVM^b	89.0%	47.3%	−
FWA [20]	ResNet-50^a[12]	9.4%	80.1%	25.6M
VA-MLP [23]	Designed CNN^a	0.2%	66.4%	−
VA-LogReg [23]	Logistic Regression^b	54.0%	78.0%	−
Xception-raw [27]	XceptionNet^a[ ]	80.4%	99.7%	22.9M
Xception-c23 [27]	XceptionNet^a[ ]	91.2%	99.7%	22.9M
Xception-c40 [27]	XceptionNet^a[ ]	83.6%	95.5%	22.9M
Multi-task [25]	Designed CNN^a	65.8%	76.3%	−
Capsule [26]	CapsuleNet^a[30]	61.3%	96.6%	3.9M
DSP-FWA [19]	SPPNet^a[13]	97.7%	93.0%	−
Multi-attentional [43]	Efficient-B4^a[34]	−	99.8%	19.5M
DefakeHop [2]	DefakeHop^b	100%	96.0%	42.8K
Ours (Frame Level)	DefakeHop++^b	100%	98.4%	238K
Ours (Video Level)	DefakeHop++^b	100%	99.3%	238K

		1st Generation
Method	Model	UADFV	FF++	#param
Two-stream [44]	InceptionV3^a[33]	85.1%	70.1%	23.9M
Meso4 [1]	Designed CNN^a	84.3%	84.7%	28.0K
MesoInception4 [1]	Designed CNN^a	82.1%	83.0%	28.6K
HeadPose [37]	SVM^b	89.0%	47.3%	−
FWA [20]	ResNet-50^a[12]	9.4%	80.1%	25.6M
VA-MLP [23]	Designed CNN^a	0.2%	66.4%	−
VA-LogReg [23]	Logistic Regression^b	54.0%	78.0%	−
Xception-raw [27]	XceptionNet^a[ ]	80.4%	99.7%	22.9M
Xception-c23 [27]	XceptionNet^a[ ]	91.2%	99.7%	22.9M
Xception-c40 [27]	XceptionNet^a[ ]	83.6%	95.5%	22.9M
Multi-task [25]	Designed CNN^a	65.8%	76.3%	−
Capsule [26]	CapsuleNet^a[30]	61.3%	96.6%	3.9M
DSP-FWA [19]	SPPNet^a[13]	97.7%	93.0%	−
Multi-attentional [43]	Efficient-B4^a[34]	−	99.8%	19.5M
DefakeHop [2]	DefakeHop^b	100%	96.0%	42.8K
Ours (Frame Level)	DefakeHop++^b	100%	98.4%	238K
Ours (Video Level)	DefakeHop++^b	100%	99.3%	238K

[ViewLarge]

Sharing Unavailable