Table 5. Complexities of the CNN models.

Model	Learnable parameters (millions)	Prediction time per image (ms)		Training time per epoch (s)
		Original dataset	Augmented dataset	Original dataset	Augmented dataset
AlexNet	44.56	0.3331	0.3385	19	58
DiatomNet	1.85	0.3213	0.3223	15	54
GoogleNet	6.04	0.5471	0.5507	27	96
Inceptionv3	21.91	0.7072	0.6860	94	351
ResNet18	11.21	0.4153	0.4190	34	128
VGG16	113.57	2.2139	2.3018	160	584
Xception	20.94	1.3322	1.1116	101	382