My REDCNN framework is the residual encoder-decoder convolutional neural network. The architecture of REDCNN includes 4 convolution down-sampling layers, 4 transposed convolution up-sampling layers, and 1 convolution full-connection layer. The electric field distribution is set as input and input image is output of REDCNN. The source data is MNIST dataset from ‘http://yann.lecun.com/exdb/mnist/’.