Table 1 Comparison of performance (%) between AM-GCN and other recognition methods.

From: Dynamic graph convolutional network for assembly behavior recognition based on attention mechanism and multi-scale feature fusion

 

A

B

C

D

E

F

G

H

I

J

K

L

M

N

O

ResNet101

94.2

96.4

82.1

64.2

70.0

92.4

91.7

84.2

93.7

78.6

84.0

61.4

83.6

96.7

61.2

SE-ResNet101

97.8

95.6

89.4

74.8

77.0

96.4

96.4

89.6

98.0

89.1

89.5

77.8

89.2

97.6

72.2

ML-GCN

98.0

96.2

92.4

76.5

87.7

97.3

97.5

93.8

98.5

89.3

91.9

81.6

92.0

98.6

75.2

ADD-GCN

96.3

98.2

94.9

80.9

85.3

97.7

98.2

95.9

98.6

92.5

93.4

84.6

95.1

99.5

78.8

AM-GCN

99.0

98.1

95.4

81.1

85.4

97.8

98.5

96.7

97.6

94.7

93.8

84.8

96.0

98.2

79.6

Back to article page