Why put a single camera on a servo when a half dozen with a compute engine provide a better result?