Ultra-Fast Computer Vision using Highly Parallel Optical Memristor CNN Networks

Project Description

Standard vision recognition systems are inherently limited by data transfer bottlenecks and the slow, serial processing of information. Focal-plane image processing offers exceptional speed, yet is limited in resolution by circuit complexity. A new type of optical vision recognition system is proposed based on optical memristors (OM) and cellular nonlinear networks (CNN). This leverages the unique capacity of OM’s to both detect light and store information whilst taking advantage of CNN’s ability to simultaneously processes all cells at once, enabling ultra-fast (parallel) computation. The research should pave the way to a new generation of ultra-fast, high-resolution vision recognition systems.