AI for Networked IoT Devices

Lau reviews the hardware options and essential software tools for creating AI-powered vision systems tailored for Internet of Things (IoT) applications. The emergence of high-performance Arm-based processors for cellphones and tablets has spurred an increasing convergence of embedded and desktop computing, with Raspberry Pis that run Ubuntu taking over for real-time embedded operating systems.

This convergence has made computer vision systems particularly easy to build, with a minimal amount of programming instruction, using interpreted scripting languages such as Python and Java instead of hardware-targeted assembly or C. Furthermore, Wi-Fi is now widespread, and there is increasing interest in IoT and accumulating information in the cloud. But streaming video to the cloud remains a challenge, especially in environments that may already be saturated with data. Therefore, Lau focuses on vision processing in the sensor and using AI operations to identify features of importance housed inside the camera and streaming only a minimal amount of data to the cloud.

About the presenter
Daniel L. Lau, Ph.D., received his B.Sc. degree with highest distinction in electrical engineering from Purdue University West Lafayette, Indiana, and his Ph.D. degree from the University of Delaware, Newark, in 1995 and 1999, respectively. Currently, he is a professor and the director of graduate studies for electrical and computer engineering at the University of Kentucky, Lexington. Among his many published works is an article in Proceedings of the IEEE and his book Modern Digital Halftoning (published by CRC Press). His work has also been featured in trade magazines such as Photonics Spectra, Vision Systems Design, Imaging Insight, Prosilica Camera News, and Inspect Magazine. His research interests include image and signal processing, 3D imaging, and machine vision. He is a Senior Member of IEEE.