TABLE 4.

Structure and parameters of the self‐defined deep classification network Dscint

Layer	Configuration
Convolution	11 × 11, 16, S = 4, P = 2
Pooling	MaxPool(3), S = 2
	Attention module
Convolution	5 × 5, 16, S = 1, P = 2
	BatchNorm
Pooling	MaxPool(3), S = 2
Convolution	3 × 3, 24, S = 1, P = 1
Convolution	3 × 3, 24, S = 1, P = 1
Convolution	3 × 3, 24, S = 1, P = 1
	BatchNorm
Pooling	MaxPool(3), S = 1
Fully connected	1024
Fully connected	1024
Softmax	4

Abbreviations: MaxPool, max pooling; P, padding; S, stride.