Table 1 Performance of five models on test set.

From: Optimizing vitiligo diagnosis with ResNet and Swin transformer deep learning models: a study on performance and interpretability

Models

ACC (%)

SEN (%)

SPE (%)

AUC

PRE (%)

F1-score (%)

ResNet34

89.26

90.04

88.01

0.90

92.46

91.23

ResNet50

88.49

88.54

88.41

0.88

92.58

90.51

ResNet101

87.18

87.92

85.98

0.88

91.10

89.48

Swin transformer base

92.74

92.53

93.09

0.92

95.62

93.75

Swin transformer large

93.82

94.02

93.50

0.94

95.93

94.97

Back to article page