Table 2.

Detailed structure description of our proposed CNN method for the gender recognition problem.

Layer Name	Number of Filters	Filter Size	Stride Size	Padding Size	Window Channel Size	Dropout Probability	Output Size
Input Layer	n/a	n/a	n/a	n/a	n/a	n/a	183 × 119 × 1
Convolution Layer 1	96	11 × 11 × 1	2 × 2	0	n/a	n/a	87 × 55 × 96
Rectified Linear Unit	n/a	n/a	n/a	n/a	n/a	n/a	87 × 55 × 96
Cross-Channel Normalization Layer	n/a	n/a	n/a	n/a	5	n/a	87 × 55 × 96
MAX Pooling Layer 1	1	3 × 3	2 × 2	0	n/a	n/a	43 × 27 × 96
Convolution Layer 2	128	5 × 5 × 96	1 × 1	2 × 2	n/a	n/a	43 × 27 × 128
Rectified Linear Unit	n/a	n/a	n/a	n/a	n/a	n/a	43 × 27 × 128
Cross-Channel Normalization Layer	n/a	n/a	n/a	n/a	5	n/a	43 × 27 × 128
MAX Pooling Layer 2	1	3 × 3	2 × 2	0	n/a	n/a	21 × 13 × 128
Convolution Layer 3	256	3 × 3 × 128	1 × 1	1 × 1	n/a	n/a	21 × 13 × 256
Rectified Linear Unit	n/a	n/a	n/a	n/a	n/a	n/a	21 × 13 × 256
Convolution Layer 4	256	3 × 3 × 256	1 × 1	1 × 1	n/a	n/a	21 × 13 × 256
Rectified Linear Unit	n/a	n/a	n/a	n/a	n/a	n/a	21 × 13 × 256
Convolution Layer 5	128	3 × 3 × 256	1 × 1	1 × 1	n/a	n/a	21 × 13 × 128
Rectified Linear Unit	n/a	n/a	n/a	n/a	n/a	n/a	21 × 13 × 128
MAX Pooling Layer 5	1	3 × 3	2 × 2	0	n/a	n/a	10 × 6 × 128
Fully Connected Layer 1	n/a	n/a	n/a	n/a	n/a	n/a	4096
Rectified Linear Unit	n/a	n/a	n/a	n/a	n/a	n/a	4096
Fully Connected Layer 2	n/a	n/a	n/a	n/a	n/a	n/a	1024
Rectified Linear Unit	n/a	n/a	n/a	n/a	n/a	n/a	1024
Dropout Layer	n/a	n/a	n/a	n/a	n/a	50%	1024
Output Layer	n/a	n/a	n/a	n/a	n/a	n/a	2
Softmax Layer	n/a	n/a	n/a	n/a	n/a	n/a	2
Classification Layer	n/a	n/a	n/a	n/a	n/a	n/a	2