TABLE 5. Baseline Methods for polyp Segmentation on the Kvasir-SEG Dataset. Two Best Scores are Highlighted in Bold. “−” Shows That There is no Backbone Used in the Network.

Method	Backbone	Jaccard C.	DSC	F2-score	Precision	Recall	Overall Acc.	FPS
UNet [58]	–	0.4713	0.5969	0.5980	0.6722	0.6171	0.8936	11.0161
ResUNet [61]	–	0.5721	0.6902	0.6986	0.7454	0.7248	0.9169	14.8204
ResUNet++ [3]	–	0.6126	0.7143	0.7198	0.7836	0.7419	0.9172	7.0193
FCN8 [57]	VGG 16	0.7365	0.8310	0.8248	0.8817	0.8346	0.9524	24.9100
HRNet [64]	–	0.7592	0.8446	0.8467	0.8778	0.8588	0.9524	11.6970
DoubleUNet [42]	VGG 19	0.7332	0.8129	0.8207	0.8611	0.8402	0.9489	7.4687
PSPNet [59]	ResNet50	0.7444	0.8406	0.8314	0.8901	0.8357	0.9525	16.8000
DeepLabv3+ [60]	ResNet50	0.7759	0.8572	0.8545	0.8907	0.8616	0.9614	27.9000
DeepLabv3+ [60]	ResNet101	0.7862	0.8643	0.8570	0.9064	0.8592	0.9608	16.7500
UNet [58]	ResNet34	0.8100	0.8757	0.8622	0.9435	0.8597	0.9681	35.0000
ColonSegNet (Proposed)	–	0.7239	0.8206	0.8206	0.8435	0.8496	0.9493	182.3812