how can I extract predicted label and testlabel from already trained deep learning model. the code below gives error while running

load('MyVGG19Model.mat');
imdsTest = imageDatastore("C:\Users\Bashir\Desktop\Training dataset\Test set", ...
    'IncludeSubfolders',true, ...
    'LabelSource','foldernames');
imageAugmenter = imageDataAugmenter( ...
    'RandRotation',[-90,90], ...
    'RandScale',[1 1.1], ...
    'RandXTranslation',[-3 3], ...
    'RandYTranslation',[-3 3]);

imageSize = myvgg16.Layers(1).InputSize;
augmentedTestSet = augmentedImageDatastore(imageSize, imdsTest, 'DataAugmentation',imageAugmenter);


predictedLabels = predict(myvgg16, augmentedTestSet);
testLabels = imdsTest.Labels
% Tabulate the results using a confusion matrix.
confMat = confusionmat(testLabels, predictedLabels);

% Convert confusion matrix into percentage form
confMat = bsxfun(@rdivide,confMat,sum(confMat,2))

[Walter Roberson]

What error message do you observe?

[Bashir Abubakar]

thank u walter for the swift response. below is the error:

Error using confusionmat (line 71)

G and GHAT need to be the same type.

Error in samples (line 22)

confMat = confusionmat(testLabels, predictedLabels);

[Walter Roberson]

What shows up for class(testLabels) and class(predictedLabels)

I suspect that one is string and the other is categorical.

[Bashir Abubakar]

They are all categorical images. But how do I convert them all to be categorical

[Walter Roberson]

Could you confirm that

 class(testLabels)
 class(predictedLabels)

both show categorical ?

If so, then there is no need to convert them to categorical as they would already be categorical.

[Bashir Abubakar]

Thank you walter for support. TestLabels showed categorical While predictedLabels showed single. I want them all to be of class categorical. How do I go about that pls.

[Walter Roberson]

You could try

categorical(string(round(predictedLabels)))

but I would recommend looking more carefully at the values in predictedLabels. The values of the predictedLabels might possibly be class numbers, in which case you would want to use them to index the categories that were used in TestLabels .

Answer 1

[yanqi liu]

yes，sir，as Walter Roberson said，may be ensure data class，such as

load('MyVGG19Model.mat');
imdsTest = imageDatastore("C:\Users\Bashir\Desktop\Training dataset\Test set", ...
    'IncludeSubfolders',true, ...
    'LabelSource','foldernames');
imageAugmenter = imageDataAugmenter( ...
    'RandRotation',[-90,90], ...
    'RandScale',[1 1.1], ...
    'RandXTranslation',[-3 3], ...
    'RandYTranslation',[-3 3]);

imageSize = myvgg16.Layers(1).InputSize;
augmentedTestSet = augmentedImageDatastore(imageSize, imdsTest, 'DataAugmentation',imageAugmenter);


predictedLabels = predict(myvgg16, augmentedTestSet);
testLabels = imdsTest.Labels
% Tabulate the results using a confusion matrix.
confMat = confusionmat(double(testLabels), double(predictedLabels));

% Convert confusion matrix into percentage form
confMat = bsxfun(@rdivide,confMat,sum(confMat,2))