Computing depth of identified object, not whole disparity map Python OpenCV

Question

I\'m building a turret that will need minimum 20 FPS at at least 640x480px and the vision system will be using two USB cameras as stereo vision.

I plan to use OpenCV\'s s