Table 3 Quantitative comparison results on the Vaihingen dataset with state-of-the-art methods.

From: Spatially adaptive interaction network for semantic segmentation of high-resolution remote sensing images

Method

Impervious surface

Building

Low vegetation

Tree

Car

mean \(\hbox {F}_1\)

OA

mIoU

Mem (GB)

DeepLabv3+26

91.4

94.7

79.6

87.6

85.8

87.8

89.9

79.0

7.36

PSPNet23

90.6

94.3

79.0

87.0

70.7

84.3

89.1

74.1

6.14

FSFNet75

91.9

94.8

82.5

89.0

85.4

88.7

89.8

-

6.89

CLANet76

91.5

94.6

82.9

89.5

87.3

89.2

89.8

80.7

7.62

CTFNet77

90.7

94.4

81.7

87.3

82.7

87.4

88.6

77.8

5.67

MFANet78

92.7

95.8

84.2

89.3

84.7

89.3

90.6

81.6

6.28

BGFNet79

92.9

95.5

84.4

89.6

86.7

89.8

90.6

81.7

5.99

HRNet74

85.8

91.6

71.1

79.9

74.2

88.7

90.1

72.6

2.76

Basaeline11

88.3

93.0

79.7

87.1

83.3

89.0

86.3

78.7

1.40

SAINet

92.5

96.5

85.9

89.3

89.4

90.7

91.9

82.5

1.87

  1. Best results are in bold.