Transposing matrices of a Dataset

Hello! I'm new of the comunity.

I imported a dataset containing the mat files where each of them represents a 3D matrix (43x512x210).

i need to transpose each of these matrices into matrices 210x512x43. The name I associated with the dataset is "imds" and the code to import it is:

digitDatasetPath = fullfile('C:\Users\User\Desktop\DB_BCI\DB');
imds = imageDatastore(digitDatasetPath, ...
    'IncludeSubfolders',true,'FileExtension', '.mat','LabelSource','foldernames');

the files mat in the dataset are 120:

would anyone be able to help me? Thank you!

Answer 1

[Bjorn Gustavsson]

Take a look at the help and documentation of the matlab-function permute.

HTH

[Gianmarco Plutino]

I have seen the documentation of the matlab function permute, but i don't know how to apply this function for each element of the dataset.

thank you for the answer.

[Bjorn Gustavsson]

So straight from the help:

    PERMUTE and IPERMUTE are a generalization of transpose (.') 
    for N-D arrays.
 
    Example:
       a = rand(1,2,3,4);
       size(permute(a,[3 2 1 4])) % now it's 3-by-2-by-1-by-4.

That means that for an array, A, with size 43 x 512 x 210, you will have to do something like this:

       A = permute(A,[3 2 1]);

If you have your 120 data-sets in a cell-array, D, you'll have to repeat this cell-element by cell-element:

for iD = 1:120,
    D{iD} = permute(D{iD},[3 2 1]);
end

[Gianmarco Plutino]

Thanks man i'll try this solution!

[Gianmarco Plutino]

Hello my friend,

i tried this solution:

for i=1:120
    imds.Files{i,1}=permute(imds.Files{i,1},[3 2 1]);
end

and the command is executed without errors, but when i try to train the net it gives me the following error:

>> net = trainNetwork(imdsTrain,layers,options);
Error using trainNetwork (line 154)
The training images are of size 43x512x210 but the input layer expects images of size 210x512x43.

Caused by:
    Error using nnet.internal.cnn.util.NetworkDataValidator/assertCorrectDataSizeForInputLayer (line 262)
    The training images are of size 43x512x210 but the input layer expects images of size 210x512x43.
    

which makes me understand that the transformation of the matrices has not been successful.

[Bjorn Gustavsson]

Check with:

size(imds.Files{i,1})

before and after the permutation.