Fig. 2.

CNN architecture of supervised deep learning. The CNN is composed of four encode and four decode processes. The total number of layers is 30. CNN convolutional neural network, Conv convolution layer, BN batch normalization layer, ReLU activation function layer