processImagesMNIST doesn't give the right data format for trainNetwork

Loading mnist:

According to the dataset for DL,

I load MNIST by this:

filenameImagesTrain = 'train-images-idx3-ubyte.gz';
filenameLabelsTrain = 'train-labels-idx1-ubyte.gz';
XTrain = processImagesMNIST(filenameImagesTrain);
YTrain = processLabelsMNIST(filenameLabelsTrain);

By running whos XTrain and YTrain,

XTrain size: 28 28 1 6000, class: dlarray, YTrain size: 60000 1, class: categorical.

Training network

I define a layers as the Convolutional neural network.

When I run the training,

net = trainNetwork(XTrain, YTrain, layers, options);

It throws the error:

Invalid 2-D image training data. Specify image data as a 3-D numeric array containing a single image, a 4-D numeric array containing multiple images, a datastore, or a table containing image file paths or images in the first column.

But XTrain is indeed 4-d array.

What's the problem here?

[Runcong Kuang]

I use

XTrain = extractdata(XTrain)

to convert dlarray to numerical array.

And then it works.

But I believe dlarray is advanced for doing DL. If I convert it back to do the DL, I think it's not efficient.

What's the best way to make use of the dlarray directly? Any toolbox?

Thanks.

Answer 1

[Aman]

Hi Runcong,

It is my understanding that you are having issues while loading the data for model training.

The “processImagesMNIST” function returns a “dlarray”, and the “trainNetwork” method does not accept “dlarray” as an input parameter. In order to resolve the error, you need to convert the “dlarray” into a numerical array using the “extractdata” function and then pass the numerical array to the “trainNetwork” function.

You can convert the “dlarray” to a numerical array in the following manner:

XTrain = extractdata(XTrain);

Please refer to the following links to know more about the “trainNetwork” function and the list of the functions that support “dlarray”:

https://www.mathworks.com/help/deeplearning/ref/trainnetwork.html

https://www.mathworks.com/help/deeplearning/ug/list-of-functions-with-dlarray-support.html

Hope this helps and resolves your error!

Regards,

Aman Mehta