Semantic Segmentation Response size and output size mismatch

Hello everyone,

I am trying to do a segmentation of a mushroom dataset. But, after I ran my code, I got an error.

Invalid training data. The output size ([32 32 2]) of the last layer does not match the response size ([168 300 2]).

Can someone help me?

%% Cereate a Semantic Segmentation Network
inputSize = [32 32 3];%Here, an image size of [32 32 3] is used for the network to process 64x64 RGB images.
imgLayer = imageInputLayer(inputSize);
%Create Downsampling Network
filterSize = 3;
numFilters = 32;
conv = convolution2dLayer(filterSize,numFilters,'Padding',1);
relu = reluLayer();

poolSize = 2;
maxPoolDownsample2x = maxPooling2dLayer(poolSize,'Stride',2);

downsamplingLayers = [
    conv
    relu
    maxPoolDownsample2x
    conv
    relu
    maxPoolDownsample2x
    ];

%Create Upsampling Network
filterSize = 4;
transposedConvUpsample2x = transposedConv2dLayer(4,numFilters,'Stride',2,'Cropping',1);
upsamplingLayers = [
    transposedConvUpsample2x
    relu
    transposedConvUpsample2x
    relu
    ];

%Create A Pixel Classification Layer
numClasses = 3;
conv1x1 = convolution2dLayer(1,numClasses);
finalLayers = [
    conv1x1
    softmaxLayer()
    pixelClassificationLayer()
    ];

%Stack All Layers
net = [
    imgLayer    
    downsamplingLayers
    upsamplingLayers
    finalLayers
    ];
%Train A Semantic Segmentation Network
dataDir = 'C:\Users\cleme\OneDrive\Bureau\tp_matlab\mushroom';
dataDir1 = 'C:\Users\cleme\OneDrive\Bureau\tp_matlab\mushroom\images';
imDir = fullfile(dataDir1,'*.jpg');
pxDir = fullfile(dataDir,'PixelLabelData');

classNames = ["cap" "stalk"];
pixelLabelID = [1 2];
pxds = pixelLabelDatastore(pxDir,classNames,pixelLabelID);

imds = imageDatastore(imDir);

% resize image 
imds.ReadFcn = @customreader;
reset(imds);
I = read(imds);
C = read(pxds);

%Visualize training images and ground truth pixel labels.
I = imresize(I,5);
L = imresize(uint8(C{1}),5);
imshowpair(I,L,'montage')

%Create a semantic segmentation network. This network uses a simple semantic segmentation network based on a downsampling and upsampling design. 
numFilters = 64;
filterSize = 3;
numClasses =  2;
layers = [
    imageInputLayer([32 32 1])
    convolution2dLayer(filterSize,numFilters,'Padding',1)
    reluLayer()
    maxPooling2dLayer(2,'Stride',2)
    convolution2dLayer(filterSize,numFilters,'Padding',1)
    reluLayer()
    transposedConv2dLayer(4,numFilters,'Stride',2,'Cropping',1);
    convolution2dLayer(1,numClasses);
    softmaxLayer()
    pixelClassificationLayer()
    ];
opts = trainingOptions('sgdm', ...
    'InitialLearnRate',1e-3, ...
    'MaxEpochs',100, ...
    'MiniBatchSize',64);
%Create a pixel label image datastore that contains training data.

trainingData = pixelLabelImageDatastore(imds,pxds);
net = trainNetwork(trainingData,layers,opts);

%% Function 
function data = customreader(filename)
onState = warning('off', 'backtrace');
c = onCleanup(@() warning(onState));
data = imread(filename);
data = data(:,:,min(1:3, end)); 
data = imresize(data, [32 32]);
data = rgb2gray(data);
end

[Clémence Labarre]

Okay I found my mistake

I had to harmonize pxds too : imresize[32 32]

And if the mismatch remains, we can modulate the output size in "layers" with Cropping and Padding attributes.