I am working on a novelty detection project (on Android) using background subtraction. I would like to do a simple classification after finding an object (e.g. Person, Dog, Car, Motorbike). I read some literature on how to do this, for example, Bag Of Words, the Cascade classifier, but I'm still not sure how to approach the task. In fact, I want to preserve the novelty as an image, and then mark it with some class, such as Person, Dog, etc., Even if it corresponds to a specific person. The number of classes should not be too large - 2 or 3. Therefore, my questions:
1) What is the best / easiest way to do this? Cascade / Bag of words / other methods?
2) How to transfer this classifier to an Android application, so that after I save the new product as an image, I can simply pass it to the classifier, which then applies the label. I know that Cascade classifiers have associated xml files, but is this the only way?
Sorry if the questions seem trivial, but it's hard for me to find literature on the Internet.
source
share