Table 8 Comparison of computational complexity.

From: Dual-branch dynamic hierarchical U-Net with multi-layer space fusion attention for medical image segmentation

Methods

GFLOPs

Inference speed (ms)

Paras (M)

FCN

19.52

22.84

18.6

U-Net

40.23

17.78

17.27

ResU-Net

68.08

32.34

13.04

Attention U-Net

30.91

29.77

34.88

U-Net++

34.63

24.49

9.16

ResU-Net++

70.84

36.08

14.48

TransformU-Net

78.63

41.21

105.28

SwimU-Net

40.52

22.34

27.17

HiFormer

45.72

28.31

25.51

D2HU-Net

32.94

21.13

41.58