Table 3 Quantitative comparisons with classical models on the INRIA dataset.

From: Dynamic atrous attention and dual branch context fusion for cross scale Building segmentation in high resolution remote sensing imagery

 

OA (%)

Precision (%)

Recall (%)

F1-score (%)

mIoU (%)

DeepLabV3+

93.28

87.35

86.4

86.87

76.8

PSPNet

94.24

90.81

83.7

87.11

77.99

HRNet

94.78

89.62

88.05

88.83

80.85

B-FGC-Net

94.70

87.82

88.12

87.97

79.31

Segformer

95.16

91.89

86.94

89.34

81.44

TDNet

95.11

91.62

87.02

89.26

81.34

DSymFuser

95.11

91.44

88.25

89.81

82.09

GDGNet

95.09

91.56

87.93

89.71

81.64

SparseFormer

95.12

91.35

88.14

89.71

81.98

SegTDformer

95.39

91.54

88.51

89.94

82.54

  1. The highest value for each metric is marked as bold.
Back to article page