Table 5.

This table summarises the seven CNN architectures adopted for the comparative analysis in terms of their main contributions, datasets involved, FLOPs, parameters and input shapes.

Architectures	Main contributions	Datasets	FLOPs (B)	Parameters (M)	Input shapes
EfficientNetB028	Compound scaling	ImageNet-ILSVRC-2012-CLS)47	0.39	5.3	224 × 224
EfficientNetV2B029	Progressive learning	ImageNet-ILSVRC-2012-CLS	0.72	7.1	224 × 224
EfficientNetV2B0-21k29	Progressive learning	ImageNet-21k48	0.72	7.1	224 × 224
ResNetV1-5049	Residual learning	ImageNet-ILSVRC-2012-CLS	4.1	25.6	224 × 224
ResNetV2-5031	Identity mapping	ImageNet-ILSVRC-2012-CLS	4.1	25.6	224 × 224
MobileNetV132	Depth-wise separable convolutions	ImageNet-ILSVRC-2012-CLS	0.6	4.2	224 × 224
MobileNetV233	Inverted residuals and linear bottlenecks	ImageNet-ILSVRC-2012-CLS	0.3	3.4	224 × 224