Error using trainNetwork Invalid training data. The output size ([256 256 2]) of the last layer does not match the response size ([501 401 2]).

direct = dir('*.jpg');

n = length(direct);

direct = natsortfiles(direct);

for i = 1:n

AAA = imread(direct(i).name);

BBB = imresize(AAA,[256 256]);

BBB = repmat(BBB,1,1,3);

imwrite(BBB,sprintf('im%d.jpg',i))

end

dataDir = fullfile('C:\Users\39371\OneDrive\Desktop\fedetesi');

imDir = fullfile(dataDir,'scala di grigi vetro');

imds = imageDatastore(imDir);

pxDir = fullfile(dataDir,'warp_vetro_dnn');

classNames = ["warp","background"];

labelIDs = [1 0];

pxds = pixelLabelDatastore(pxDir,classNames,labelIDs);

imageSize = [256 256];

numClasses = numel(classNames);

lgraph = deeplabv3plusLayers(imageSize,numClasses,'resnet18');

cds = combine(imds,pxds);

opts = trainingOptions('sgdm',...

'MiniBatchSize',8,...

'MaxEpochs',10);

net = trainNetwork(cds,lgraph,opts);

[Federico]

Sono poco pratico nell'utilizzo di matlab, come posso risolvere questo errore?

Answer 1

[Aman]

Hi,

I understand that you are trying to do image segmentation, for that you are using the DeepLab v3+ CNN model and getting the size mismatch error when putting the model into the training.

This error occurs when the size of the output layer differs from the size of the response, which is nothing but the size of the actual ground truth.

To solve the issue by fixing response size, kindly refer this:

https://www.mathworks.com/support/search.html/answers/782218-the-output-size-of-the-last-layer-does-not-match-the-response-size.html?fq%5B%5D=asset_type_name:answer&fq%5B%5D=category:deeplearning/image-data-workflows&page=1

Without the actual data, it is difficult to point out what is causing this issue, but generally it can be solved by either fixing the output layer size or by reshaping the response size.

To solve the issue by fixing output size, kindly refer this:

https://www.mathworks.com/support/search.html/answers/1958459-invalid-training-data-the-output-size-of-128-128-2-the-last-layer-does-not-match-the-response-siz.html?fq%5B%5D=asset_type_name:answer&fq%5B%5D=category:deeplearning/image-data-workflows&page=1

Please refer to the following documentation to learn more about Image Segmentation using DeepLab v3+ CNN model

https://www.mathworks.com/help/vision/ref/deeplabv3pluslayers.html

I hope it helps!