A 256x256 6.3pJ/pixel-event Query-driven Dynamic Vision Sensor with Energy-conserving Row-parallel Event Scanning
Rajkumar Kubendran, Akshay Paul, Gert Cauwenberghs
Abstract
This work reports a query-driven dynamic vision sensing (q DVS) approach to computational imaging that substantially improves on the achievable pixel density and energy efficiency of visual event coding. The chip architecture combines complementary advantages of frame scanning active pixel sensors (APS) and event-driven DVS (e DVS) CMOS imagers. The 256x256-pixel q DVS with 33% fill factor achieves 10% temporal contrast sensitivity and consumes 0.5mW from 1.2V at a peak event rate of 80Meps, yielding an energy efficiency Figure-of-Merit (FoM) of 6.3pJ/pixel-event.
Topics & Concepts
PixelComputer scienceFigure of meritEvent (particle physics)Frame rateImage sensorEnergy (signal processing)Computer visionFrame (networking)Artificial intelligencePhysicsTelecommunicationsQuantum mechanicsAdvanced Memory and Neural ComputingCCD and CMOS Imaging SensorsNeuroscience and Neural Engineering