Fast Feature Field gives event cameras a general-purpose, real-time output for robots

Penn's F³ learns to predict future events, computes only where events occur, and outputs a multi-channel image that standard vision models can read. It runs at 120 Hz at HD resolution.