Table 3 Performance evaluation of different segmentation methods on PH2 dataset
From: Attention-based dual-path feature fusion network for automatic skin lesion segmentation
Methods | F1-score | SE | SP | AC | PC | JS |
|---|---|---|---|---|---|---|
UNet [18] | 0.789 | 0.781 | 0.851 | 0.859 | 0.787 | 0.651 |
Attention UNet [47] | 0.823 | 0.796 | 0.934 | 0.889 | 0.852 | 0.700 |
R2UNet [48] | 0.834 | 0.812 | 0.923 | 0.896 | 0.857 | 0.715 |
BCDUNet [24] | 0.866 | 0.864 | 0.938 | 0.914 | 0.869 | 0.764 |
HiFormeS [25] | 0.922 | 0.982 | 0.903 | 0.937 | 0.871 | 0.856 |
Ours(VGG) | 0.897 | 0.893 | 0.954 | 0.934 | 0.902 | 0.814 |
Ours(ResNet101) | 0.925 | 0.954 | 0.922 | 0.945 | 0.915 | 0.872 |