Hi I'm working on a car occupant counter system that must count occupant of cars in a video stream (see attached image). I read some related papers and found that in all models they firstly define tow ROIs by detecting the windshield and rear window, then look up for an occupant in these ROIs. I have searched for different object detectors in OpenCV and found two implemented object detectors (Cascade Classifier and YOLO). I have already used YOLO for detecting cars, but I'm hesitated about which model to pick for detecting windows. Is there any advise?
