An Improved Image Classification Method for Cervical Precancerous Lesions Based on ShuffleNet

Table 2

Network comparison experimental data.

Method	Accuracy (%)	Precision (%)	Recall (%)	F1-score (%)
VGG-16	50.72 ± 2.12	45.63 ± 3.25	45.67 ± 1.02	45.07 ± 1.59
ResNet34	83.95 ± 4.02	84.88 ± 3.18	81.28 ± 4.51	82.81 ± 3.44
GoogleNet	53.72 ± 5.42	47.43 ± 4.77	51.73 ± 4.82	45.09 ± 5.03
DenseNet121	86.39 ± 1.45	87.00 ± 1.91	83.95 ± 2.62	85.17 ± 1.98
MobileNet	54.30 ± 1.57	65.12 ± 2.18	44.60 ± 1.69	43.45 ± 2.03
ShuffleNet	80.37 ± 2.06	79.90 ± 1.89	79.42 ± 1.58	79.60 ± 1.95
ShuffleNet_SK	81.23 ± 2.03	81.65 ± 1.64	79.88 ± 2.25	80.67 ± 1.83
ShuffleNet_SE	81.38 ± 1.95	81.76 ± 2.32	80.74 ± 1.87	81.16 ± 2.26