How do you handle edge cases in a computer vision pipeline processing raw camera feeds for autonomous landing on Zipline drones?
