I am generating ground truth using image labeller app and my images are from CCTV camera to detect different vehicles on the road. I have 4 classes including car,van,pedestrians,bike. I am thinking to substract background from the mentioned classes and define it as a seperate class to improve my accuracy. is there anyway to do it automatically since I have 73 images.
logically like this : background = 1-(car|pedestrians|van|bike).
Then save it all as pixel label data.
I would appreciate the help.