Traditional Vision
on DNN IP
Hosted by Paul Brasnett
Summary
There has been a very significant shift in the computer vision industry over the past few years from traditional vision algorithms to deep neural network algorithms. Many companies with experience and investment in classical vision algorithms want to utilise DNNs without discarding their existing investments. When selecting IP for upcoming products there are likely to be competing needs to support both traditional vision and DNN-based vision.
In this talk, we’ll look at the basic operations supported by dedicated DNN IP (such as the PowerVR 2NX) and how traditional computer vision algorithms can be mapped to those operations. This approach can provide a cost-efficient way to develop silicon capable of supporting both traditional and DNN-based vision.
About the speaker
Paul Brasnett, Senior Research Manager Vision & AI, PowerVR, Imagination
Paul is a Senior Research Manager leading a team responsible for imaging, image processing and computer vision algorithms within the PowerVR division of Imagination Technologies. Prior to this he was a senior research engineer at Mitsubishi Electric involved in image and video processing and participated in MPEG standardization work. Paul has PhD and MEng degrees from the University of Bristol, UK.
