Table 2.

Convolutional neural network-based models in the Keras Application.

Model	Top 1 accuracy	Top 5 accuracy	Depth	Size (MB)	Parameters	Reference
VGG16	0.713	0.901	23	528	138,357,544	[33]
VGG19	0.713	0.900	26	549	143,667,240
ResNet50	0.749	0.921	—	98	25,636,712	[34]
ResNet101	0.764	0.928	—	171	44,707,176
ResNet152	0.766	0.931	—	232	60,419,944
ResNet50V2	0.760	0.930	—	98	25,613,800	[35]
ResNet101V2	0.772	0.938	—	171	44,675,560
ResNet152V2	0.780	0.942	—	232	60,380,648
InceptionV3	0.779	0.937	159	92	23,851,784	[36]
InceptionResNetV2	0.803	0.953	572	215	55,873,736	[37]
Xception	0.790	0.945	126	88	22,910,480	[38]
MobileNet	0.704	0.895	88	16	4,253,864	[39]
MobileNetV2	0.713	0.901	88	14	3,538,984	[40]
DenseNet121	0.750	0.923	121	33	8,062,504	[41]
DenseNet169	0.762	0.932	169	57	14,307,880
DenseNet201	0.773	0.936	201	80	20,242,984
NASNetMobile	0.744	0.919	—	23	5,326,716	[42]
NASNetLarge	0.825	0.960	—	343	88,949,818
EfficientNetB0	—	—	—	29	5,330,571	[43]
EfficientNetB1	—	—	—	31	7,856,239
EfficientNetB2	—	—	—	36	9,177,569
EfficientNetB3	—	—	—	48	12,320,535
EfficientNetB4	—	—	—	75	19,466,823
EfficientNetB5	—	—	—	118	30,562,527
EfficientNetB6	—	—	—	166	43,265,143
EfficientNetB7	—	—	—	256	66,658,687