TABLE 3. The Architecture of Lighter CNN.

Layer	Number of filters, n	Size/stride	Activation function	Output size
Input	N/A	N/A	N/A
Convolutional	64	8/2	N/A
Batch normalization	N/A	N/A	Leaky Relu
Convolutional	128	8/1	N/A
Batch normalization	N/A	N/A	Relu
Average pool	N/A	4/2	N/A
Dropout	N/A	N/A	N/A
Convolutional	256	8/1	N/A
Batch normalization	N/A	N/A	Relu
Convolutional	128	8/1	N/A
Batch normalization	N/A	N/A	Relu
Average pool	N/A	4/2	N/A
Dropout	N/A	N/A	N/A
Convolutional	64	8/1	N/A
Batch normalization	N/A	N/A	Relu
Convolutional	32	5/2	N/A
Batch normalization	N/A	N/A	Relu
Convolutional	32	5/1	N/A
Batch normalization	N/A	N/A	Relu
Flatten	N/A	N/A	N/A	512
Fully connected	32		Relu	32
Fully connected	4		Relu	4
Output layer	N/A		Sigmoid	1