What is event-based vision?

An event camera contains pixels that independently report changes in log-brightness as timestamped events. Instead of capturing fixed frames on a clock, each pixel fires only when something changes.

That matters because conventional cameras waste bandwidth on static regions and blur under fast motion. Event cameras sample only what changes, which gives them microsecond temporal resolution and high dynamic range—properties that robotics researchers care about when robots move quickly or work in difficult lighting.

Each pixel maintains a reference brightness. When brightness changes past a threshold, the pixel fires an event and resets. The resulting stream is sparse, asynchronous, and tightly tied to motion in the scene.

Event-based vision connects to spiking networks, neuromorphic processors, optical flow, tracking, and SLAM under motion. In the Neuromorphic Robotics field map, it sits at the center of Sensing.