Datastore array exceeds maximum array size preference

I am currently working on an application of transfer learning. It is the first time i work with MATLAB and i followed all the courses available. I need to build a network which has an image as input and output. I updated the net I am using (Googlenet, but i have to eventually use other networks too) and when i try to make the command trainNetwork i have the error i mentioned in the title.

Here is the code:

x_images = imageDatastore("dataset\JPEGImages","LabelSource","foldernames");
contour_images = imageDatastore("dataset\contorni", "LabelSource","foldernames");

[train_x_images, test_x_images] = splitEachLabel(x_images, 0.8);
[train_contour_images, test_contour_images] = splitEachLabel(contour_images, 0.8);


train_x_images = augmentedImageDatastore([224,224], train_x_images);
test_x_images = augmentedImageDatastore([224,224], test_x_images);

train_contour_images= augmentedImageDatastore([224,224], train_contour_images);
test_contour_images= augmentedImageDatastore([224,224], test_contour_images);


net = googlenet;
lgraph = layerGraph(net);


layer_to_put = convolution2dLayer(224,224, "Name", "fc","Padding","same");
lgraph = replaceLayer(lgraph,"loss3-classifier", layer_to_put);
layer_to_put= convolution2dLayer(224,224,"Name","layer_conv","Padding","same");
lgraph = replaceLayer(lgraph, 'prob', layer_to_put);
layer_to_put = regressionLayer("Name", "output");
lgraph = replaceLayer(lgraph,"output",layer_to_put);


options = trainingOptions("sgdm","InitialLearnRate",0.001);
[myNet, info] = trainNetwork(train_x_images,lgraph,options);

I have another question which may seem stupid, when training the net how can i pass the correct output? (Which is stored in train_contour_images)

Answer 1

[Jayanti]

Hi Nicola,

I think the issue you are facing is due to maximum array size limitations in MATLAB. To resolve this try increasing the MATLAB's maximum array size preference.

Goto MATLAB Home tab-> Preferences -> Workspace -> Uncheck MATLAB array size limit checkbox

For your other question when training a neural network you can use “combine” function to create a combined datastore that pairs each input image with its corresponding output image.

Hope this will resolve the issue!