Visual processing sparks a new class of processors

Abstract

Summary form only given. Augmented reality, gesture interfaces, and automotive driver assistance systems enable novel user experiences, safer rides, and new usage models. Bringing these systems to market requires a power-efficient architecture and many billions of operations per second of processing. Widely adopted processing architectures like CPUs and GPUs can't fulfill the requirements, sparking a new class of video and vision processors. In this talk we'll give a quick overview of applications, typical algorithms, and their implications on computer architecture. We will focus on the automotive market, where computer vision is the key technology to enabling autonoumous vehicles hitting the road.