Table 1.

Properties of the 8 pre-trained models selected for the experiments.

Model	Depth	Size	Parameters (Millions)	Image Input Size
ResNet50	50	96 MB	25.6	224 × 224
ResNet101	101	167 MB	44.6	224 × 224
VGG16	16	515 MB	138	224 × 224
VGG19	19	535 MB	144	224 × 224
GoogleNet	22	27 MB	7	224 × 224
MobileNet V2	53	13 MB	3.5	224 × 224
AlexNet	8	227 MB	61	227 × 227
DenseNet201	201	77 MB	20	224 × 224

Note: These 8 pre-trained networks have different network depths, network sizes, network parameter scales, and input image sizes, and they cover almost all forms of current deep learning models. The results obtained from the validation experiments on these 8 pre-trained models with small samples are generalizable.