Labeling is the fuel that makes machine learning run. A label tells an artificial neural network what the correct output is for a given input. For example, the pixels of empty roadway in a video are labeled as "free space". The pixels that correspond to vehicles, pedestrians, sidewalks, traffic cones, barriers, and so on are not labeled as free space. Given enough labeled examples, neural networks are good at learning which patterns of pixels correspond to road surface and which patterns correspond to obstacles. When shown a new video, a trained network can give the