Table 5.

CNN architecture.

Layer Type	Number of Filters	Size of Feature Map (Width × Height × Channel)	Size of Filter	Stride	Padding
Image input layer		119 $\times$ 183 $\times$ 3
1st convolutional layer	96	55 $\times$ 87 $\times$ 96	11 $\times$ 11 $\times$ 3	2 $\times$ 2	0 $\times$ 0
Rectified linear unit (ReLU) layer		55 $\times$ 87 $\times$ 96
Local response normalization layer		55 $\times$ 87 $\times$ 96
Max pooling layer	1	27$\times$ 43$\times$ 96	3 $\times$ 3	2 $\times$ 2	0 $\times$ 0
2nd convolutional layer	128	27 $\times$43 $\times$128	5 $\times$ 5 $\times$ 96	1 $\times$ 1	2 $\times$ 2
ReLU layer		27 $\times$ 43$\times$ 128
Local response normalization layer		27 $\times$ 43 $\times$ 128
Max pooling layer	1	13 $\times$ 21 $\times$ 128	3 $\times$ 3	2 $\times$ 2	0 $\times$ 0
3rd convolutional layer	256	13 $\times$ 21 $\times$ 256	3 $\times$ 3 $\times$ 128	1 $\times$ 1	1 $\times$ 1
ReLU layer		13 $\times$ 21 $\times$ 256
4th convolutional layer	256	13 $\times$ 21 $\times$ 256	3 $\times$ 3 $\times$ 256	1 $\times$ 1	1 $\times$ 1
ReLU layer		13 $\times$ 21 $\times$ 256
5th convolutional layer	128	13 $\times$ 21 $\times$ 128	3 $\times$ 3 $\times$ 256	1 $\times$ 1	1 $\times$ 1
ReLU layer		13 $\times$ 21 $\times$ 128
Max pooling layer	1	6 $\times$ 10 $\times$ 128	3 $\times$ 3	2 $\times$ 2	0 $\times$ 0
1st fully connected layer		4096
ReLU layer		4096
2nd fully connected layer		1024
ReLU layer		1024
Dropout layer		1024
3rd fully connected layer		2
Softmax layer		2
Classification layer (output layer)		2