Every second, machines like drones, smartphones, and self-driving cars capture floods of visual data. But turning this data into meaning—like recognizing a red light or reading a sign—takes massive ...