The View from the Driver’s Seat

Stacey Svetlichnaya

Self-driving cars are a compelling futurist dream: all the freedom of a personal vehicle, none of the mental effort. A first step towards this is understanding the roadside environment and at least avoiding obstacles: pedestrians, cyclists, parked cars, the vehicle ahead in the lane, etc. In this report, I explore an existing semantic segmentation model which parses scenes from dashboard cameras into 20 categories of objects or materials: car, bicycle, building, trees, sky, sidewalk, etc. It turns out that finding humans is harder than it seems…

Join our mailing list to get the latest machine learning updates.