Table 4 Performance comparison between the proposed segmentation and other state-of-the-art methods on ISBI 2016 challenge test set
From: DermoNet: densely linked convolutional neural network for efficient skin lesion segmentation
Method | JC (%) | DSC (%) |
|---|---|---|
HED [24] | 79.3 | 88.4 |
SegNet [4] | 70.0 | 82.3 |
UiT-Seg [25] | 80.6 | NA |
IHPC-CS [20] | 79.9 | NA |
CNN-CRF [20] | 79.7 | NA |
FCN - (JC) trained with DermoSafe data | 72.6 | 81.6 |
U-Net - (JC) trained with DermoSafe data | 80.3 | 87.6 |
DermoNet (JC) trained without DermoSafe data | 81.1 | 88.2 |
DermoNet (JC) trained with DermoSafe data | 82.5 | 89.4 |