My pictures have a format of [227 227 1] so I had the idea to triplicate my processed pictures and put them back into one to get a format of [227 227 3], how can I do this?

Hi!

I'm doing a project where I am writing a hand gesture detection program but i've run into some problems.

I'm getting my picture through a webcam and processing them to be black and white and have a format of [227 227 1]. Unfortunally its just not working with my self written network.

Now I want to try a pretraind network but this requires a format of [227 227 3] but my pictures are still black and withe so they have a format of [227 227 1] so I had the idea to triplicate my processed picture and put them back into one to get a format of [227 227 3]

Does someone have and idea how I could do that?

Or could help me to figure out why my selfwritten network isn't working?

Attached you can find the Network, many thanks in advance!!

allImages = imageDatastore('Handgesten_BW','IncludeSubfolders',true, 'LabelSource','foldernames');
[imdsTrain,imdsVal] = splitEachLabel(allImages,0.7,'randomized');
                       
imageSize = [227 227 1];
OutputSize = 7;

%Layers

 layers = [
    imageInputLayer(imageSize,"Name","imageinput")
    convolution2dLayer([11 11],96,"Name","conv1","BiasLearnRateFactor",2,"Stride",[4 4])
    reluLayer("Name","relu1")
    crossChannelNormalizationLayer(5,"Name","norm1","K",1)
    maxPooling2dLayer([3 3],"Name","pool1","Stride",[2 2])
    groupedConvolution2dLayer([5 5],128,2,"Name","conv2","BiasLearnRateFactor",2,"Padding",[2 2 2 2])
    reluLayer("Name","relu2")
    crossChannelNormalizationLayer(5,"Name","norm2","K",1)
    maxPooling2dLayer([3 3],"Name","pool2","Stride",[2 2])
    convolution2dLayer([3 3],384,"Name","conv3","BiasLearnRateFactor",2,"Padding",[1 1 1 1])
    reluLayer("Name","relu3")
    groupedConvolution2dLayer([3 3],192,2,"Name","conv4","BiasLearnRateFactor",2,"Padding",[1 1 1 1])
    reluLayer("Name","relu4")
    groupedConvolution2dLayer([3 3],128,2,"Name","conv5","BiasLearnRateFactor",2,"Padding",[1 1 1 1])
    reluLayer("Name","relu5")
    maxPooling2dLayer([3 3],"Name","pool5","Stride",[2 2])
    fullyConnectedLayer(4096,"Name","fc6","BiasLearnRateFactor",2)
    reluLayer("Name","relu6")
    dropoutLayer(0.5,"Name","drop6")
    fullyConnectedLayer(4096,"Name","fc7","BiasLearnRateFactor",2)
    reluLayer("Name","relu7")
    dropoutLayer(0.5,"Name","drop7")
    fullyConnectedLayer(OutputSize,"Name","fc8","BiasLearnRateFactor",2)
    softmaxLayer("Name","prob")
    classificationLayer("Name","classoutput")];

    
    %Train/Val
    augmenter = imageDataAugmenter( ...
    'RandRotation',[-45 45], ...
    'RandYReflection', true, ...
    'RandScale',[0.8,1.2]);
    augimdsTrain = augmentedImageDatastore(imageSize,imdsTrain,'DataAugmentation',augmenter);                        
    augimdsVal = augmentedImageDatastore(imageSize,imdsVal,'DataAugmentation',augmenter);                                                     
     
    %Options
  options = trainingOptions('sgdm',...
      'MiniBatchSize',32, ...
      'MaxEpochs',15,...
      'InitialLearnRate',0.001,...
      'ValidationData',augimdsVal, ...
      'ValidationFrequency',20,...
      'Verbose',false, ...
      'executionenvironment','gpu',...
      'Shuffle','every-epoch', ...
      'Plots','training-progress');
  
      %'LearnRateSchedule','piecewise',...                
      %'LearnRateDropFactor',0.5,...
      %'LearnRateDropPeriod',25,...
      % crossChannelNormalizationLayer(5,'K',1), ...
  Netzwerk = trainNetwork(augimdsTrain,layers,options);
    save Netzwerk

[Jan]

"format of [227 227 1]"

This is not possible in Matlab. Trailing dimensions of length 1 (except for the 2nd dimension) vanish automatically:

x = rand(277, 277, 1);
size(x)

What exactly does this mean: "not working with my self written network"? This detail is important, so please share it.

Answer 1

[Image Analyst]

Here is the way I usually use

rgbImage = cat(3, grayImage, grayImage, grayImage);

Answer 2

[DGM]

To rearrange a single-channel intensity image into a grayscale RGB image, you can just do

Argb = repmat(Aint,[1 1 3]);

I don't know if it would be possible/practical or more efficient to find a way to process the single-channel images. I don't know anything about the rest of your code or the training topic.