Unexpected image size: All images must have the same size.

Hi, I'm having some problems with a bench of chest xray images. I tryed to use the code from the link below, but it did not work.

https://www.mathworks.com/matlabcentral/answers/385472-error-in-image-size

Error using trainNetwork (line 165)

Unexpected image size: All images must have the same size.

Error in chestXray1 (line 49)

net = trainNetwork(imdsTrain,layers,options);

inputSize = [224 224 1];
numClasses = 2;

layers = [
    imageInputLayer(inputSize)
    convolution2dLayer(5,20)
    batchNormalizationLayer
    reluLayer
    fullyConnectedLayer(numClasses)
    softmaxLayer
    classificationLayer];

options = trainingOptions('sgdm', ...
    'MaxEpochs',3, ...
    'ValidationData',imdsValidation, ...
    'ValidationFrequency',30, ...
    'Verbose',false, ...
    'Plots','training-progress');

net = trainNetwork(imdsTrain,layers,options);

[Geoff Hayes]

Andre - please clarify what you mean by I tryed to use the code from the link below, but it did not work. Did you modify the code for your image size of 224x224x1? Were there any errors when you ran this code? What was the output? Please provide more details.

[Andre Brandao]

So, I'm using the example of mnist digits for the Chest Xray images. The out put is:

Error using trainNetwork (line 165)

Unexpected image size: All images must have the same size.

Error in chestXray1 (line 49)

net = trainNetwork(imdsTrain,layers,options);

Then I tryed to use the solutation that was gave in that link to adjust the images in an appropriate way, and didn`t work.

Code below:

ext = 'png';        %adjust as appropriate
projectdir = 'Some\Appropriate\Directory';
dinfo = dir(fullfile(projectdir, '**', ['.' ext]);
allfiles = fullfile({dinfo.folder}, {dinfo.name});
for K = 1 : length(allfiles)
  thisfile = allfiles{K};
  try
    thisinfo = imfinfo(thisfile);
    if ~isfield(thisinfo, 'Width') || ~isfield(thisinfo, 'Height') || ~isfield(thisinfo, 'SamplesPerPixel')
      fprintf('image information for file "%s" is missing width or height or #channels\n', thisfile);
    else
      if thisinfo.Width ~= 227
        fprintf('File "%s" width expected 227 got %d\n', thisfile, thisinfo.Width);
      end
      if thisinfo.Height ~= 227
        fprintf('File "%s" height expected 227 got %d\n', thisfile, thisinfo.Height);
      end
      if thisinfo.SamplesPerPixel ~= 3
        fprintf('File "%s" expected 3 color channels, got %d\n', thisfile, thisinfo.SamplesPerPixel);
      end
    end
  catch ME
    fprintf('Could not get image information for file "%s"', thisfile);
  end
end

[Geoff Hayes]

Andre - but what do you mean by it didn't work. Did you get any errors? Was there any output from running the above code? I'm guessing that there must have been since the code is checking for images of dimensions 227x227x3 and you are interested in images of 224x224x1. You will need to modify the code to check for your dimensions...and modify the code so that it looks in the appropriate directory assuming your folder of png (?) images is not 'Some\Appropriate\Directory'. At the very least, you will need to change the following in the code

projectdir = 'Some\Appropriate\Directory';

must be a valid folder/directory on your computer where your images are.

  if thisinfo.Width ~= 227
    fprintf('File "%s" width expected 227 got %d\n', thisfile, thisinfo.Width);
  end
  if thisinfo.Height ~= 227
    fprintf('File "%s" height expected 227 got %d\n', thisfile, thisinfo.Height);
  end
  if thisinfo.SamplesPerPixel ~= 3
    fprintf('File "%s" expected 3 color channels, got %d\n', thisfile, thisinfo.SamplesPerPixel);
  end

In the above, you need to change the dimensions to match those that you are interested in.

  ext = 'png';   

Are your chest xrays pngs or some other format?

[Gabija Marsalkaite]

Please clarify if you want 1 channel with dimmentions 224x224 or 3 channel with 227x227 images. The code above is made for checking if all files are 227x227x3 but from your code snippet inputSize = [224 224 1];

[Andre Brandao]

Hi, thank for reply. I want 1 channel with 224x224. I've already adjusted the images to 224x224x1. The images are in jpeg format, maybe I need to convert to png. But also need help to do that. When I try to compile the code it stops and gives me that message:

Error using trainNetwork (line 165)
Unexpected image size: All images must have the same size.

Error in chestXray1 (line 49)
net = trainNetwork(imdsTrain,layers,options);
----------------------------------------------------------------
below the code I''m using to:

%%  Carregar dados
dataChest = fullfile('/Users/andrebr4/Documents/MATLAB/chestXray/chest_xray');
imds = imageDatastore(dataChest, ...
    'IncludeSubfolders',true, ...
    'LabelSource','foldernames');

%%  Dividir o conjunto de dados em cada categoria
numTrainingFiles = 750;
[imdsTrain,imdsValidation] = splitEachLabel(imds,numTrainingFiles,'randomize');

%%  Configurar a rede neural
inputSize = [224 224 1];
numClasses = 2;

layers = [
    imageInputLayer(inputSize)
    convolution2dLayer(5,20)
    batchNormalizationLayer
    reluLayer
    fullyConnectedLayer(numClasses)
    softmaxLayer
    classificationLayer];

%%  Opções de treino
options = trainingOptions('sgdm', ...
    'MaxEpochs',5, ...
    'ValidationData',imdsValidation, ...
    'ValidationFrequency',30, ...
    'Verbose',false, ...
    'Plots','training-progress');

%% Treinar a rede neural
net = trainNetwork(imdsTrain,layers,options);

%%  Executar rede treinada no conjunto de teste
YPred = classify(net,imdsValidation);
YValidation = imdsValidation.Labels;

%% Calcular a precisão
accuracy = sum(YPred == YValidation)/numel(YValidation)

[Geoff Hayes]

Andre - what is the output when you run the following code

ext = 'jpeg';        %adjust as appropriate
projectdir = '/Users/andrebr4/Documents/MATLAB/chestXray/chest_xray';
dinfo = dir(fullfile(projectdir, '**', ['.' ext]);
allfiles = fullfile({dinfo.folder}, {dinfo.name});
for K = 1 : length(allfiles)
  thisfile = allfiles{K};
  try
    thisinfo = imfinfo(thisfile);
    if ~isfield(thisinfo, 'Width') || ~isfield(thisinfo, 'Height') || ~isfield(thisinfo, 'SamplesPerPixel')
      fprintf('image information for file "%s" is missing width or height or #channels\n', thisfile);
    else
      if thisinfo.Width ~= 224
        fprintf('File "%s" width expected 224 got %d\n', thisfile, thisinfo.Width);
      end
      if thisinfo.Height ~= 224
        fprintf('File "%s" height expected 224 got %d\n', thisfile, thisinfo.Height);
      end
      if thisinfo.SamplesPerPixel ~= 1
        fprintf('File "%s" expected 1 color channels, got %d\n', thisfile, thisinfo.SamplesPerPixel);
      end
    end
  catch ME
    fprintf('Could not get image information for file "%s"', thisfile);
  end
end

[Andre Brandao]

is missing width or height or #channels

[Geoff Hayes]

try putting a breakpoint at the line

allfiles = fullfile({dinfo.folder}, {dinfo.name});

and then run the code. When the debugger pauses at thisline, step through the subsequent lines. What is thisfile set to? What is thisinfo?

Answer 1

[Dheeraj Singh]

You can use augmentedImageDataStore to resize all images to same size.

Use the following code for your problem:

dataChest = fullfile('/Users/andrebr4/Documents/MATLAB/chestXray/chest_xray');
imds = imageDatastore(dataChest, ...
    'IncludeSubfolders',true, ...
    'LabelSource','foldernames');

%%  Dividir o conjunto de dados em cada categoria
numTrainingFiles = 750;
[imdsTrain,imdsValidation] = splitEachLabel(imds,numTrainingFiles,'randomize');


%%%%%%%code for resizing
inputSize=[224 224 1];
imdsTrain=augmentedImageDatastore(inputSize, imdsTrain,'ColorPreprocessing','rgb2gray');

imdsValidation=augmentedImageDatastore(inputSize, imdsValidation,'ColorPreprocessing','rgb2gray');



%%  Configurar a rede neural
inputSize = [224 224 1];
numClasses = 2;

layers = [
    imageInputLayer(inputSize)
    convolution2dLayer(5,20)
    batchNormalizationLayer
    reluLayer
    fullyConnectedLayer(numClasses)
    softmaxLayer
    classificationLayer];

%%  Opções de treino
options = trainingOptions('sgdm', ...
    'MaxEpochs',5, ...
    'ValidationData',imdsValidation, ...
    'ValidationFrequency',30, ...
    'Verbose',false, ...
    'Plots','training-progress');

%% Treinar a rede neural
net = trainNetwork(imdsTrain,layers,options);

%%  Executar rede treinada no conjunto de teste
YPred = classify(net,imdsValidation);
YValidation = imdsValidation.Labels;

%% Calcular a precisão
accuracy = sum(YPred == YValidation)/numel(YValidation)

[Andre Brandao]

it works... thanks.

[Mohamed Nasr]

Hi i have the same problem but with 3channels what i do ?

[Mohamed Nasr]

and make error in YPred = classify(net,imdsValidation);

YValidation = imdsValidation.Labels;