Some time ago it was mentioned that SDCs can't use deep-learning because it is not sufficiently reliable, and that instead techniques like LiDAR should be used to recognize objects.
However, this made me wonder: how does the car recognize road-signs and traffic lights?