index is out of range for deletion - error while training semantic segmentation network

Hi,

I'm having an issue while training a UNET network for semantic segmentation on 3D nifti images.

I'm getting the following error:

>> network = trainNetwork(ds,lgraph,options);
Error using trainNetwork (line 165)
Matrix index is out of range for deletion.

Caused by:
    Matrix index is out of range for deletion.

I'm not sure whether I built wrongly the network or that I didn't created the datastore for training as needed (I'm trying to load 3D nifti images in order to train a 2D semantic segmentation of each "slice" of the 3D image).

Could you please assist me with this issue?

below is my code:

%% Create U-Net Network with Custom Encoder-Decoder Depth
imageSize = [512 512 1];
numClasses = 2;
encoderDepth = 5;
lgraph = unetLayers(imageSize,numClasses,'EncoderDepth',encoderDepth);


dataSetDir = fullfile('c:\','Users','ya','Desktop','images_for_matlab_training');
imageDir = fullfile(dataSetDir,'images');
labelDir = fullfile(dataSetDir,'gTruth');

% Create an imageDatastore object to store the training images
imds = imageDatastore(imageDir,'FileExtensions','.gz','ReadFcn',@ReadNifti);

% Create an pixelDatastore object to store the training images
classNames = ["liver","background"];
labelIDs   = [255 0];
pxds = pixelLabelDatastore(labelDir,classNames,labelIDs,'FileExtensions','.gz','ReadFcn',@ReadNifti);

% Create a datastore for training the network
ds = pixelLabelImageDatastore(imds,pxds);

% Set training options
options = trainingOptions('sgdm', ...
'InitialLearnRate',1e-3, ...
'MaxEpochs',3, ...
'MiniBatchSize',10, ...
'Plots','training-progress',...
'Verbose',1,...
'Shuffle','once');

% Train the network
network = trainNetwork(ds,lgraph,options);

Thanks!!

[GABRIELZHU ZHU]

Hello, have you ever figured this problem out? I met the same problem as you. Could you please help me with this?

[Mario Malic]

@GABRIELZHU ZHU this happens when you try to delete a row or column on a matrix which doesn't exist.

[GABRIELZHU ZHU]

imds = imageDatastore(fullfile(home, 'MAT'),'FileExtensions','.mat','ReadFcn',@matRead);

classnames = 'LowCVR';

pixelLabelID = 1;

pxds = pixelLabelDatastore(fullfile(home, 'Label-3'), classnames, pixelLabelID, 'FileExtensions','.mat',...

'ReadFcn',@matRead);

pximds = pixelLabelImageDatastore(imds, pxds);

layers = [

imageInputLayer([128 128 18],"Name","imageinput")

convolution2dLayer([3 3],32,"Name","conv_1","Padding",[1 1 1 1])

reluLayer("Name","relu_1")

maxPooling2dLayer([2 2],"Name","maxpool_1","Stride",[2 2])

convolution2dLayer([3 3],32,"Name","conv_2","Padding",[1 1 1 1])

reluLayer("Name","relu_2")

maxPooling2dLayer([2 2],"Name","maxpool_2","Stride",[2 2])

transposedConv2dLayer([4 4],32,"Name","transposed-conv_1","Cropping",[1 1 1 1],"Stride",[2 2])

reluLayer("Name","relu_3")

transposedConv2dLayer([4 4],32,"Name","transposed-conv_2","Cropping",[1 1 1 1],"Stride",[2 2])

reluLayer("Name","relu_4")

convolution2dLayer([1 1],8,"Name","conv_3")

softmaxLayer("Name","softmax")

pixelClassificationLayer("Name","classoutput")];

opts = trainingOptions('adam', ...

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

'MaxEpochs',200, ...

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

'MiniBatchSize',16);

net = trainNetwork(pximds, layers, opts)

[GABRIELZHU ZHU]

Thank you Mario. But I don't know where the step is.

[Mario Malic]

It looks like it's inside your last line. The full error message should show where is the actual error. If error is somewhere within trainNetwork, then maybe your setup is wrong? I do not know anything about Deep Learning, so I can't help you. Verify that your inputs are correct, use debugging to find out where is the error.