Table 3.

Comparison of experimental results using different well-known CNN architectures with their trained weights.

	CNN Model	Accuracy	Precision	Recall	Specificity	F1_score
Non-Normalized	GoogleNet	83.87%	0.8373	0.8404	0.9677	0.8379
	VGG16	88.71%	0.8885	0.8889	0.9774	0.8835
	VGG19	87.10%	0.8674	0.8725	0.9742	0.8681
	DenseNet201	89.86%	0.9005	0.9005	0.9797	0.8986
	AlexNet	86.18%	0.8764	0.8629	0.9722	0.8554
Normalized Augmented	GoogleNet	85.24%	0.8492	0.8524	0.9705	0.8480
	VGG16	87.14%	0.8723	0.8714	0.9743	0.8677
	VGG19	85.00%	0.8465	0.8500	0.9700	0.8454
	DenseNet201	88.33%	0.8795	0.8833	0.9767	0.8797
	AlexNet	82.38%	0.8588	0.8238	0.9648	0.7975
Non-Normalized Augmented	GoogleNet	82.03%	0.8235	0.8219	0.9640	0.8158
	VGG16	82.72%	0.8515	0.8279	0.9653	0.8202
	VGG19	81.11%	0.8128	0.8120	0.9622	0.7920
	DenseNet201	83.41%	0.8460	0.8364	0.9668	0.8368
	AlexNet	79.72%	0.8100	0.8004	0.9594	0.7899