Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

How to test the auto train SVM

I have some doubt of testing my SVM classifier. I have trained some SVMs with train_auto (I am using OpenCV 2.4.9, because the 3.0.0 was not released when the application was done, and I do not want to change everything now). And now I want to test them and I am not really sure how to do that.

It works fine for C_SVC, because the prediction is < 0 or > 0, and if it is on one side it is one class, and if it is on the other side is the other class, am I right? NU_SCV is the same, but if I have or C_SVR, NU_SVR it is not the same way (does it needs both the positives and the negative samples, too?).

More, in case of ONE_CLASS what is happening, and how do I test it?

This was done for two classes (positives and negatives). I also know that predict(image, false) will return the class label, but I just want to get a little deeper in testing the SVM classifier.