Wanted to use the function multisvm under Image Processing, since it has been removed, please suggest an alternative for the same.

result = multisvm(Train_Feat);

Answer 1

[Mrutyunjaya Hiremath]

'multisvm' was never a built-in MATLAB function but seems to be a user-provided one that has been circulated in various forums and online platforms for multi-class SVM classification. MATLAB has the Statistics and Machine Learning Toolbox, which provides a way to perform multi-class classification with SVM.

Here's a rough outline of how you might use the built-in functions for a multi-class SVM:

Train the SVM:

When using the fitcecoc function, it internally trains binary SVM classifiers for each pair of classes and uses them for multi-class classification.

t = templateSVM('KernelFunction', 'polynomial', 'PolynomialOrder', 2);
Mdl = fitcecoc(Train_Feat, Train_Label, 'Learners', t);

Where 'Train_Feat' are your training features and 'Train_Label' are your training labels.

Predict using the trained SVM:

result = predict(Mdl, Test_Feat);

Where 'Test_Feat' are your test features.

It's quite straightforward using the built-in functions, and they're optimized and well-integrated into MATLAB's ecosystem.

Note: Ensure you have the Statistics and Machine Learning Toolbox installed and licensed in MATLAB.