TABLE 2.

Skin lesion segmentation performance of our SL-HarDNet and several popular segmentation methods on the ISIC-2016&PH2 test set and ISIC2018 dataset.

Datasets	Methods	DIC	JAC	ACC	SEN	SPE
ISIC-2016&PH2	FCN	0.889	0.811	0.932	0.967	0.922
	U-Net++	0.910	0.844	0.937	0.925	0.960
	CA-Net	0.894	0.819	0.936	0.938	0.947
	TransFuse	0.914	0.850	0.945	0.972	0.919
	TransUNet	0.917	0.853	0.942	0.968	0.915
	SL-HarDNet (Ours)	0.927	0.871	0.953	0.975	0.926
ISIC-2018	U-Net	0.848	0.769	0.945	0.881	0.964
	DeepLabv3	0.894	0.825	0.962	0.910	0.967
	CE-Net	0.906	0.839	0.969	0.916	0.976
	UCTransNet	0.910	0.849	0.971	0.920	0.976
	BAT	0.911	0.848	0.971	0.925	0.974
	SL-HarDNet (Ours)	0.915	0.853	0.972	0.926	0.980

The bold value is to emphasize that this value is optimal.