How do i using openCV create or improve on an a already existing cascade classifier. I have a ROI consisting of of the object which i want to track (positive image) and the rest of the frame is what i don't want to detect (negative image). I just read somewhere that OpenCV provides an application which is capable of training these classifiers, but i am more interested in using an already existing one, like the haarcascades_Frontal_face_alt2, and improve it such it is capable of tracking only my face from different angles, poses and so on, is it possible to code such a solution?, and how would one do so.
I am thinking making the learning unsupervised or semi supervised. but I am not sure on how to handle this issue code wise.