Table 2. Same model recognition accuracy.

Model	Evaluation index
	Training set accuracy (%)	Verification set accuracy (%)	Average processing time of a single image (s)	Floating-point arithmetic
ResNet50	97.17	92.13	0.243	3.89×109
ResNext	97.64	92.73	0.301	4.82×109
MoblieViTv3	96.96	91.34	0.284	2.54×109
ResNet101	97.10	91.81	0.457	5.02×109
DenseNet	97.89	92.84	0.451	5.31×109
EfficientNet	97.62	93.47	0.456	5.36×109
FS-ResNet50	98.31	95.87	0.275	4.17×109