TABLE 5.

Performance comparison with SOTA methods on TN-SCUI datasets. $\text{Red}$ , $\text{Green}$ , and $\text{Blue}$ indicate the best, second-best, and third-best performance.

Network	TN-SCUI datasets		#Params (M)	Model size (MB)
	IoU	Dice
U-Net Ronneberger et al. (2015)	0.718	0.806	29.59	118
SegNet Badrinarayanan et al. (2017)	0.726	0.819	17.94	71.8
FATNet Wu et al. (2022)	$\text{0.751}$	$\text{0.842}$	27.43	109
Swin-UNet Cao et al. (2021)	0.744	0.835	25.86	105
TransUNet Chen et al. (2021b)	$\text{0.746}$	0.837	88.87	401
EANet Wang et al. (2022)	$\text{0.751}$	$\text{0.839}$	47.07	118
UNeXt-L Valanarasu and Patel (2022)	0.693	0.794	$\text{3.99}$	$\text{15.2}$
PL-Net† (Our)	0.742	0.830	$\text{3.77}$	$\text{15.6}$
PL-Net (Our)	$\text{0.767}$	$\text{0.851}$	$\text{15.03}$	$\text{60.7}$

The bold values indicates the best performance.