MATLAB: CNN Layers: high pass filters and truncation layer

convolutional neural networksdeep learninghigh pass filter layertruncation layer

I would like to create two layers for my CNN design where the input image is high-pass filtered via convolution with 16 (4 × 4) high-pass DCT filters, and then a Truncation layer (Truncation Linear Unit) activation function.

These layers are not built in matlab deep learning toolbox. How can I make them?

Thank you in adavance!

Best Answer

You can write your own Custom Layer using Deep Learning Toolbox in MATLAB. To get started you can refer to Define Custom Deep Learning Layers & Deep Learning Custom Layers.

Related Solutions

MATLAB: How to pass images to a cnn for analysis

Hey Teshan,

After you train a neural network for deep learning using trainNetwork into net, follow this procedure:

Use the function augmentedImageDatastore for effective preprocessing of the images
Use the activations function to compute the DL Network layer activations i.e. the image features using the CNN net
Use the predict function to finally make a prediction using the classifier

The overall workflow that I mentioned is available in this example as well.

Hope this helps!

MATLAB: Monitor Deep Learning Training Progress

I tried reproducing the above issue using an example provided in the documentation -

digitDatasetPath = fullfile(matlabroot,'toolbox','nnet','nndemos','nndatasets','DigitDataset');
imds = imageDatastore(digitDatasetPath, 'IncludeSubfolders',true,'LabelSource','foldernames');
numTrainFiles = 750;
[imdsTrain,imdsValidation] = splitEachLabel(imds,numTrainFiles,'randomize');
inputSize = [28 28 1];
numClasses = 10;
layers = [
    imageInputLayer(inputSize)
    convolution2dLayer(5,20)
    batchNormalizationLayer
    reluLayer
    fullyConnectedLayer(numClasses)
    softmaxLayer
    classificationLayer];
options = trainingOptions('sgdm', ...
    'MaxEpochs',4, ...
    'ValidationData',imdsValidation, ...
    'ValidationFrequency',30, ...
    'Verbose',false, ...
    'Plots','training-progress');
net = trainNetwork(imdsTrain,layers,options);

I was able to view the training progress

Please crosscheck the spellings of Name Value pairs in trainingOptions

Best Answer

Related Solutions

MATLAB: How to pass images to a cnn for analysis

MATLAB: Monitor Deep Learning Training Progress

Related Question