Trying to move images to different folder based on CNN classification, and failing

Hi folks, I'm trying to pass my images for classification to a CNN and then move each image (after being classified) to a folder based on its classification. Currently, I don't get any errors but the code just compiles without any outputs...Any help would be most appreciated!

for k = 1:fileNumber

    currentImage = fullfile(imageList(k).folder, imageList(k).name);
    img = imread(currentImage);
    labelName = classify(DaveNet, img);
    
    switch labelName
        case 'Anisotropy'
            movefile(currentImage, AnisotropyClassificationPath);
        case 'Isotropy'
            movefile(currentImage, IsotropyClassificationPath);
        case 'Filler'
            movefile(currentImage, FillerClassificationPath);
        otherwise
            movefile(currentImage, BadCropsClassificationPath);
    end 
    
end    

My CNN for reference:

imds = imageDatastore(TrainingDatasetPath, 'IncludeSubfolders',true,...
    'LabelSource','foldernames');
labelCount = countEachLabel(imds)
numTrainFiles = 500;
[imdsTrain,imdsValidation] = splitEachLabel(imds,numTrainFiles,'randomize');

layers = [
    imageInputLayer([227 227 3])
    
    convolution2dLayer(3, 8, 'Padding', 'same')
    batchNormalizationLayer
    reluLayer
    
    maxPooling2dLayer(2, 'Stride', 2)
    
    convolution2dLayer(3, 16, 'Padding', 'same')
    batchNormalizationLayer
    reluLayer
    
    maxPooling2dLayer(2, 'Stride', 2)
    
    convolution2dLayer(3, 32, 'Padding', 'same')
    batchNormalizationLayer
    reluLayer
    
    maxPooling2dLayer(2, 'Stride', 2)
    
    convolution2dLayer(3, 64, 'Padding', 'same')
    batchNormalizationLayer
    reluLayer
      
    maxPooling2dLayer(2, 'Stride', 2)
    
    convolution2dLayer(3, 128, 'Padding', 'same')
    batchNormalizationLayer
    reluLayer
    
    fullyConnectedLayer(4)
    softmaxLayer
    classificationLayer];

options = trainingOptions('sgdm','InitialLearnRate',0.001, ...
    'MaxEpochs', 5,'Shuffle','every-epoch', ...
    'ValidationData',imdsValidation,'ValidationFrequency',30, ...
    'Verbose',false, 'Plots','training-progress');

net = trainNetwork(imdsTrain, layers, options);
YPred = classify(net, imdsValidation);
YValidation = imdsValidation.Labels;
accuracy = sum(YPred == YValidation)/numel(YValidation)

Answer 1

[Jon]

I would suggest checking the return status and error messages from movefile by assigning some left hand arguments to the movefile command.

For example

    switch labelName
        case 'Anisotropy'
            [isGood,msg] = movefile(currentImage, AnisotropyClassificationPath);
        case 'Isotropy'
            [isGood,msg] = movefile(currentImage, IsotropyClassificationPath);
            % and so on
    end
    % check for errors
    if ~isGood
        disp(msg)
    end

Maybe the destination path is not formulated correctly or you have a permission issue with the destination folder

Output Arguments

collapse all

status — Move status

0 | 1

Move status, indicating if the attempt to move the file or folder is successful, returned as 0 or 1. If the attempt is successful, the value of statusis 1. Otherwise, the value is 0.

Data Types: logical

msg — Error message

character vector

Error message, returned as a character vector. If an error or warning occurs, msg contains the message text of the error or warning. Otherwise, msg is empty, ''.

msgID — Error message identifier

character vector

Error message identifier, returned as a character vector. If an error or warning occurs, msgID contains the message identifier of the error or warning. Otherwise, msgID is e

[Teshan Rezel]

Cheers Jon, I've tried doing that and it seems to have resolved my issue somewhat, but now I'm stuck with the number of elements always defaulting to 9 (it should be 9000 or so)...not sure why this is.

Before, the numel function returned 0...

[Jon]

Are you referring to the numel of YValidation?

I'm not that familiar with the CNN functions that you are using, but if you have less validation data than you think you should have maybe the total number of data sets is less than you think so when you use

numTrainFiles = 500;
[imdsTrain,imdsValidation] = splitEachLabel(imds,numTrainFiles,'randomize');

if the length of imds is say only 509 then you will only have 509-numTrainFiles, or 509-500 = 9

Just a thought.

Otherwise I would suggest single stepping through your code using the MATLAB debug tool and looking at the values of the various variables as you go, and you should be able to see where things are going wrong. If you aren't familiar with the debugger it is probably the single best thing you can learn about to make your life better in MATLAB, please see https://www.mathworks.com/help/matlab/matlab_prog/debugging-process-and-features.html