Skip to main content

Table 12 Results of the CNN model with ImageDataGenerator with filters on different size of fully connected layer

From: Automatic classification of spread‐F types in ionogram images using support vector machine and convolutional neural network

FC layer

Filter applied rate (%)

Model accuracy with different image filters (%)

Image sharpening

Image thresholding

Median blur

Gamma correction

Fast-NLmeans

Bilateral

4 layers

20

90.5

89.5

89.3

88.2

88.1

89.2

40

90.1

90.1

89.3

88.7

86.8

89.2

60

89.1

91.8

88.6

88.5

89.8

89.6

80

90.2

89.2

90.3

88.2

88.5

89.1

100

88.6

88.4

86.4

90.2

88.8

89.5

5 layers

20

90.8

90.8

90.2

88.6

88.2

89.8

40

89.5

90.8

90.1

89.6

89.5

89.2

60

88.4

89.2

89.9

91.3

91.1

89.8

80

91.6

90.9

89.9

90.8

88.4

90.9

100

90.3

89.2

90.3

89.9

89.9

88.6

6 layers

20

91.9

88.9

90.2

88.5

89.2

89.1

40

89.3

90.8

90.2

90.8

88.4

88.4

60

90.6

90.2

90.1

89.5

88.9

90.1

80

90.6

89.9

89.2

89.9

90.8

90.3

100

89.9

91.8

90.5

91.1

90.3

87.9

7 layers

20

90.5

90.2

88.2

89.6

92.0

91.1

40

91.8

91.3

89.5

90.1

90.6

90.6

60

90.9

91.6

90.3

87.8

89.5

89.3

80

90.6

91.1

90.6

89.6

88.8

90.9

100

90.6

91.9

89.8

90.2

88.2

88.5

8 layers

20

90.2

90.2

90.9

89.1

90.6

92.0

40

91.3

89.2

88.8

88.1

91.1

92.0

60

90.6

89.9

88.6

91.5

88.9

90.5

80

90.5

90.9

91.2

81.8

89.8

90.9

100

90.8

91.1

88.2

90.1

90.3

88.6

9 layers

20

89.6

91.3

92.0

90.5

91.1

89.9

40

91.6

91.1

90.5

90.8

91.1

91.3

60

91.1

89.8

91.2

89.6

90.6

89.3

80

88.8

90.6

89.5

88.9

91.2

90.8

100

87.5

88.5

90.6

90.6

89.1

90.9