.mat trained model linear svm to android

Hello, i have trained my data using Linear SVM machine learning from classification learner app in Matlab. Then i exported the model and saved it using .mat extension. Now, i want to import this model into android studio, so i can make the predictions real time from the smartphone. Is there any simple way to do that?

Answer 1

[Sruthi Yenugula]

Hi Michalis Gelastos,

You can check the following examples to achieve that

https://www.mathworks.com/help/supportpkg/android/ref/human-activity-recognition-simulink-model-for-smartphone-deployment.html

https://www.mathworks.com/help/supportpkg/android/ref/recognize-handwritten-digits-using-mnist-on-android.html#:~:text=On%20deployment%2C%20the%20Simulink%20model,along%20with%20the%20prediction%20probability

.

Answer 2

[Sruthi Yenugula]

Hi Michalis Gelastos,

We do not yet have an ECOC Simulink block. For ECOC models (and models for which we do not yet have Simulink blocks), you need to use a MATLAB Function block (or a System block, but MATLAB Function block is more efficient). Here’s an example that should apply to all models:

https://www.mathworks.com/help/stats/predict-class-labels-using-matlab-function-block.html

Our example in Human Activity Recognition demo also uses the MATLAB function block, so this makes me think that the issue is with the exported model from the app; the model exported from the Classification Learner App is not the model itself, but a struct that wraps the model. You can refer to this example to get the model:

https://www.mathworks.com/help/stats/code-generation-and-classification-learner-app.html

So,you extract the model first (and underlying preprocessing code if there is any) and then go to MATLAB function block

[Michalis Gelastos]

I'm trying to save the model from the struct using saveLearnerForCoder and i'm getting this error "Categorical array class labels are not supported for code generation."

[Michalis Gelastos]

I figured out that the results from the trained model were string, so i changed the to numbers and following your previous post, it worked! Thank you very match!