Is it possible to get the ouput values of a convolutional layer for a given input?

Question

Ian Walter 2019-3-3

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/447913-is-it-possible-to-get-the-ouput-values-of-a-convolutional-layer-for-a-given-input

回答： Jayanti 2025-8-22

I am currently in the process of implementing a convolutional-autoencoder using layers from the MATLAB Deep Learning toolbox. I have a series of convolutional layers after the encoder portion of the autoencoder to convert the output to an NxN grid. In order to test this network on real data, I need to feed the outputs of this layer into an NxN pixel grid, and in order to do this I need to be able to view the output of the first half of my network. Below is my layer structure. Does anyone have any recommendations as to how I might get the information that I want from this network?

layers = [
    imageInputLayer([inputSize 1], 'Normalization','none')
    fullyConnectedLayer(16*N^2)
    reluLayer
    convert1d2dLayer    % convert from fully connected to usable by convolution layers
    convolution2dLayer(3, inputSize, 'Padding', 1)
    batchNormalizationLayer
    reluLayer
    maxPooling2dLayer(2, 'Stride', 2)
    convolution2dLayer(3, 2*inputSize, 'Padding', 1)
    batchNormalizationLayer
    reluLayer
    maxPooling2dLayer(2, 'Stride', 2)
    convolution2dLayer(3, 1, 'Padding', 1)
    sigmoidLayer('sigmoid') % custom sigmoid activation layer
    % ---end of encoder--- %
    transposedConv2dLayer(16, inputSize, 'Stride', 16) % upsampling
    % ---start of decoder--- %
    convolution2dLayer(5, 2*inputSize, 'Padding', 2)
    batchNormalizationLayer
    reluLayer
    maxPooling2dLayer(2, 'Stride', 2)
    convolution2dLayer(3, inputSize, 'Padding', 1)
    batchNormalizationLayer
    reluLayer
    maxPooling2dLayer(2, 'Stride', 2)
    fullyConnectedLayer(inputSize)
    fullyConnectedLayer(inputSize)
    softmaxLayer
    classificationLayer];