Skip to main content
. 2021 Oct 18;8:754202. doi: 10.3389/fmed.2021.754202

Table 2.

Quantitative results of CNN based vitiligo classification.

F1 AUC SP SE
(a) Test Set A
VGG 0.8864 ± 0.005 0.8506 ± 0.005 0.9129 ± 0.012 0.7905 ± 0.023
ResNet 0.8732 ± 0.006 0.8397 ± 0.007 0.9019 ± 0.014 0.7753 ± 0.025
DenseNet 0.8808 ± 0.005 0.8465 ± 0.011 0.9064 ± 0.013 0.7838 ± 0.037
(b) Test Set B
VGG 0.7649 ± 0.015 0.7302 ± 0.008 0.7905 ± 0.021 0.6699 ± 0.014
ResNet 0.7363 ± 0.031 0.7024 ± 0.023 0.7614 ± 0.042 0.6434 ± 0.039
DenseNet 0.7562 ± 0.024 0.7401 ± 0.028 0.7682 ± 0.021 0.7119 ± 0.038
(c) Test Set C
VGG 0.9684 ± 0.005 0.9946 ± 0.005 0.9629 ± 0.012 0.9721 ± 0.002
ResNet 0.9584 ± 0.004 0.9848 ± 0.007 0.9567 ± 0.016 0.9515 ± 0.010
DenseNet 0.9618 ± 0.030 0.9819 ± 0.027 0.9609 ± 0.027 0.9623 ± 0.032