Facing matlab error "grouping variable must be vector, character array, string array"

Hi,

i facing some error in matlab while using both CNN and SVM for classification. However when i try run the program the error occur. The coding is as below

imds = imageDatastore('MerchData', 'IncludeSubfolders',true, 'LabelSource','foldernames'); 
[imdsTrain,imdsValidation] = splitEachLabel(imds,0.7);

testnet = resnet18
inputSize = testnet.Layers(1).InputSize

augimdsTrain = augmentedImageDatastore(inputSize(1:2),imdsTrain)
augimdsValidation = augmentedImageDatastore(inputSize(1:2),imdsValidation)

layer = 'pool5';
featuresTrain = activations(testnet,augimdsTrain,layer,'OutputAs','rows')
featuresTest = activations(testnet,augimdsValidation,layer,'OutputAs','rows');

whos featuresTrain

YTrain = imdsTrain.Labels;
YValidation = imdsValidation.Labels;

cvpt = cvpartition(featuresTrain,"KFold",5)
opt = struct("CVPartition",cvpt)
classifier = fitcecoc(featuresTrain,YTrain,"OptimizeHyperparameters","auto","HyperparameterOptimizationOptions",opt);

Below is the error that occur

Hope someone could help me on this. Thank you very much.

Answer 1

[Prateek Rai]

To my understanding, you are getting error in matlab while using both CNN and SVM for classification.

The error is mainly arising in the line:

cvpt = cvpartition(featuresTrain,"KFold",5)

In 'cvpartition' function, the first argument should correspond to number of observations in the sample data but you are passing featuresTrain which causes an error. Instead pass the number of observations which can be done by:

cvpt = cvpartition(augimdsTrain.NumObservations,"KFold",5)

So the whole code can be rewritten as:

imds = imageDatastore('MerchData', 'IncludeSubfolders',true, 'LabelSource','foldernames'); 
[imdsTrain,imdsValidation] = splitEachLabel(imds,0.7);

testnet = resnet18
inputSize = testnet.Layers(1).InputSize

augimdsTrain = augmentedImageDatastore(inputSize(1:2),imdsTrain)
augimdsValidation = augmentedImageDatastore(inputSize(1:2),imdsValidation)

layer = 'pool5';
featuresTrain = activations(testnet,augimdsTrain,layer,'OutputAs','rows')
featuresTest = activations(testnet,augimdsValidation,layer,'OutputAs','rows');

whos featuresTrain

YTrain = imdsTrain.Labels;
YValidation = imdsValidation.Labels;

%% -- Modified code -- %%
cvpt = cvpartition(augimdsTrain.NumObservations,"KFold",5)
%% -- %%

opt = struct("CVPartition",cvpt)
classifier = fitcecoc(featuresTrain,YTrain,"OptimizeHyperparameters","auto","HyperparameterOptimizationOptions",opt);

You can refer to cvpartition MathWorks documentation page to find more on cvpartition function .

[Teo]

Thank you @Prateek Rai for the explanation, been trying this for days.