Table 5.

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

Model	Method	Accuracy	Precision	Recall	Specificity	F1-score
VGG16	without GAN	0.8670	0.7447	0.8364	0.8799	0.7879
	with GAN	0.9023	0.8024	0.8877	0.9083	0.8429
VGG19	without GAN	0.8836	0.7646	0.8758	0.8868	0.8194
	with GAN	0.9086	0.8086	0.9048	0.9102	0.8540
Xception	without GAN	0.9145	0.8039	0.9399	0.9039	0.8666
	with GAN	0.9486	0.8779	0.9593	0.9440	0.9168
ResNet50	without GAN	0.8978	0.7900	0.8906	0.9007	0.8373
	with GAN	0.9303	0.8644	0.9061	0.9404	0.8847
ResNet50v2	without GAN	0.8914	0.7755	0.8902	0.8919	0.8289
	with GAN	0.9167	0.8272	0.9077	0.9205	0.8656
InceptionV3	without GAN	0.9061	0.8075	0.8955	0.9105	0.8492
	with GAN	0.9498	0.8936	0.9423	0.9530	0.9173
InceptionResNetV2	without GAN	0.8983	0.7928	0.8875	0.9028	0.8374
	with GAN	0.9373	0.8682	0.9287	0.9409	0.8974
DenseNet121	without GAN	0.9000	0.7953	0.8907	0.9039	0.8403
	with GAN	0.9346	0.8673	0.9203	0.9407	0.8930
DenseNet169	withoutGAN	0.9032	0.7982	0.8997	0.9046	0.8459
	with GAN	0.9319	0.8624	0.9156	0.9388	0.8882