Semantic Segmentation Output & Response size

Hello,

when training a pretrained model (resnet18) for image segmentation, I always get the following error:

Error using trainNetwork (line 170)

Invalid training data. The output size ([250 1000 2]) of the last layer does not match the response size

([250 1000 1]).

Error in SemanticSegmentation (line 45)

[net, info] = trainNetwork(pximds,lgraph,options);

Already checke the network with the deep network designer and it says no errors.

Here you can see my code:

imds=imageDatastore('\\tuwzs9a-vm-02\homes\ga54saw\windows\dokumente\QuaBu\Geteilte Bilder\test')

classNames = ["noKnot", "Knot"];

pixelLabelID = [0; 1];

pxds = pixelLabelDatastore('\\tuwzs9a-vm-02\homes\ga54saw\windows\dokumente\QuaBu',classNames,pixelLabelID);

pximds = pixelLabelImageDatastore(imds,pxds);

[imdsTrain, imdsVal, imdsTest, pxdsTrain, pxdsVal, pxdsTest] = partitionData(imds,pxds);

% Specify the network image size. This is typically the same as the traing image sizes.

imageSize = [250 1000 3];

% Specify the number of classes.

numClasses = 2;

% Create DeepLab v3+.

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

%lgraph=segnetLayers(imageSize,numClasses,2);

tbl = countEachLabel(pxds);

imageFreq = tbl.PixelCount ./ tbl.ImagePixelCount;

classWeights = median(imageFreq) ./ imageFreq;

pxLayer = pixelClassificationLayer('Name','labels','Classes',tbl.Name,'ClassWeights',classWeights);

lgraph = replaceLayer(lgraph,"classification",pxLayer);

% Define validation data.

pximdsVal = pixelLabelImageDatastore(imdsVal,pxdsVal);

% Define training options.

options = trainingOptions('sgdm', ...

'LearnRateSchedule','piecewise',...

'LearnRateDropPeriod',10,...

'LearnRateDropFactor',0.3,...

'Momentum',0.9, ...

'InitialLearnRate',1e-3, ...

'L2Regularization',0.005, ...

'ValidationData',pximdsVal,...

'MaxEpochs',30, ...

'MiniBatchSize',2, ...

'Shuffle','every-epoch', ...

'CheckpointPath', tempdir, ...

'VerboseFrequency',2,...

'Plots','training-progress',...

'ValidationPatience', 4);

%Start Training

pximds = pixelLabelImageDatastore(imdsTrain,pxdsTrain);

net = trainNetwork(pximds,lgraph,options);

%Test

I = readimage(imdsTest,9);

C = semanticseg(I, net);

B = labeloverlay(I,C,'Colormap',cmap,'Transparency',0.4);

imshow(B)

pixelLabelColorbar(cmap, classes);

Would be really happy to get some help to solve the problem.

Thank you very much.

Answer 1

[Divya Gaddipati]

As it is evident from the error message, the size at the output layer should be same as the ground truth image size. Try changing the output layer size to [250 1000 1].

Here's a link to Semantic Segmentation example in MATLAB:

https://www.mathworks.com/help/vision/examples/semantic-segmentation-using-deep-learning.html