Table 3.

Results for the Kvasir-SEG dataset. Boldface numbers denote the highest values for each metrics.

Model	mDSC	mIoU	Recall	Precision
U-Net [23]	0.715	0.433	0.631	0.922
Double U-Net [27]	0.813	0.733	0.840	0.861
FCN8 (VGG16 backbone) [20]	0.831	0.737	0.835	0.882
PSPNet (ResNet50 backbone) [75]	0.841	0.744	0.836	0.890
HRNet [76]	0.845	0.759	0.859	0.878
ResUNet++ with CRF [26]	0.851	0.833	0.876	0.823
DeepLabv3+ (ResNet101 backbone) [77]	0.864	0.786	0.859	0.906
U-Net (ResNet34 backbone) [23]	0.876	0.810	0.944	0.862
FANet [34]	0.880	0.81	0.906	0.901
PolypSegNet [28]	0.887	0.826	0.925	0.917
HarDNet-MSEG [31]	0.904	0.848	0.923	0.907
Focus U-Net	0.910	0.845	0.916	0.917