Research Article

Fault Line Selection Method Based on Transfer Learning Depthwise Separable Convolutional Neural Network

Table 3

The specific structural parameters of DSCNN.

LayerConvolution kernel/sampling window sizeOutput feature map size

Input——150  150
SeparableConv2D_13  3148  148  32
MaxPooling2D_12  274  74  32
SeparableConv2D_23  372  72  64
MaxPooling2D_22  236  36  64
SeparableConv2D_33  334  34  128
MaxPooling2D_32  217  17  128
SeparableConv2D_43  315  15  128
MaxPooling2D_42  27  7  128
Flatten——6272  1
Dense——512  1
Output——6  1