Table 8 This table compares the IoU performance of different backbone networks across three datasets, with units in percentage (%). The best-performing values are highlighted in bold, while the second-best values are underlined.

From: Heterogeneous dual-decoder network for road extraction in remote sensing images

Model

DeepGlobe

Ottawa

CHN6-CUG

FLOPs (G)

Params (M)

Unet59

63.24

84.78

59.35

121.56

13.40

HDD-Unet

66.47

90.07

64.65

176.26

16.15

SwinT60

66.38

88.56

64.22

32.36

28.26

HDD-SwinT

69.85

91.82

66.34

42.67

30.77

HDDNet

71.39

91.85

67.27

28.65

22.77

Back to article page