Table 3 Performance comparison between CSA-Net and current methods on the Gothic architecture dataset

From: A coordinate-to-semantic attention network for multi-label ancient Chinese architecture image classification

Methods

mAP(%)

P(%)

R(%)

F1(%)

VGG19+GCN-A20

80.27

76.82

74.89

75.84

VGG19+GCN-B20

81.25

78.80

75.39

77.06

MSRN44

86.49

82.88

78.65

79.64

GATN43

88.27

83.25

83.00

82.13

NOAH(ResNet101)15

89.78

84.89

83.24

83.57

CSA-Net

91.18

86.73

83.26

84.91

  1. Best results are marked in bold.
Back to article page