Table 2 Experiment results of skin segmentation for the PH2 dataset
Family | Methods | Year | Dice(%) | IoU(%) | Precision(%) | Recall(%) |
|---|---|---|---|---|---|---|
CNN | UNet [1]\(^*\) | 2015 | 88.68±7.95 | 81.85±8.50 | 83.73±5.94 | 95.15±5.75 |
UNet++ [2]\(^*\) | 2018 | 91.20 | 84.35 | 86.86 | 96.69 | |
Atten_UNet [3]\(^*\) | 2018 | 90.37±8.96 | 82.21±9.23 | 85.25±5.85 | 95.98±5.53 | |
CENet [4]\(^*\) | 2019 | 91.75±7.42 | 85.06±9.74 | 85.27±5.46 | 96.70±5.18 | |
XlSor [37]\(^*\) | 2019 | 92.95±3.63 | 87.36±5.66 | 95.91±2.61 | 96.58±2.58 | |
CA-Net [5]\(^*\) | 2020 | 90.45±8.67 | - | - | - | |
Transformer | SwinUnet [26]\(^*\) | 2021 | 92.88 | 87.16 | 91.58 | 95.33 |
TransUNet(ViT) [10]\(^*\) | 2021 | 90.85 | 83.97 | 86.64 | 97.14 | |
TransUNet(R50) [10]\(^*\) | 2021 | 92.59 | 86.76 | 91.31 | 95.12 | |
iU-Net(Ours) | 2022 | 93.80 | 88.74 | 91.57 | 96.93 |