Table 4.

Comparison performance results of models without GAN and with GAN on internal dataset

Model	Method	Accuracy	Precision	Recall	Specificity	F1-score
VGG16	without GAN	0.9814	0.9898	0.9720	0.9904	0.9808
	with GAN	0.9872	0.9949	0.9788	0.9952	0.9868
VGG19	without GAN	0.9841	0.9924	0.9751	0.9928	0.9837
	with GAN	0.9875	0.9894	0.9850	0.9898	0.9872
Xception	without GAN	0.9902	0.9900	0.9900	0.9904	0.9900
	with GAN	0.9927	0.9913	0.9938	0.9916	0.9925
ResNet50	without GAN	0.9798	0.9819	0.9769	0.9826	0.9794
	with GAN	0.9911	0.9969	0.9850	0.9970	0.9909
ResNet50v2	without GAN	0.9860	0.9931	0.9782	0.9934	0.9856
	with GAN	0.9942	0.9975	0.9907	0.9976	0.9941
InceptionV3	without GAN	0.9881	0.9949	0.9807	0.9952	0.9878
	with GAN	0.9951	0.9949	0.9913	0.9952	0.9950
InceptionResNetV2	without GAN	0.9893	0.9931	0.9850	0.9934	0.9890
	with GAN	0.9893	0.9962	0.9913	0.9964	0.9938
DenseNet121	without GAN	0.9893	0.9919	0.9863	0.9922	0.9891
	with GAN	0.9921	0.9919	0.9919	0.9922	0.9919
DenseNet169	withoutGAN	0.9911	0.9944	0.9875	0.9946	0.9909
	with GAN	0.9927	0.9956	0.9894	0.9958	0.9925