Electrical and Computer Engineering
Dean of Engineering
Faculty Host: Lin Zhong
Associate Research Scholar
ECE Seminar Series: Video analytics at scale for mobile Internet-of-things platform (698/699)
Tuesday, March 28, 2017
to 5:00 PM
1064 Duncan Hall
6100 Main St
Houston, Texas, USA
Content-rich sensors, such as video cameras, are growing at an astonishing rate with emerging applications in traffic monitoring, surveillance for smart cities, and connected vehicles such as drones. Today, most of this collected video sensor data is stored close to the point of capture and is not easy to search even when timely remote access will provide valuable real-time insights. Emerging applications of Internet of things (IoT) need intelligent solutions to decide what data to analyze locally, what data to transmit to server without taxing network constraints to get a timely response, and how to use the responses for control and coordination among a network of IoT devices. My research focuses on the design of novel architectures and algorithms for IoT that can intelligently leverage edge computing and networking to minimize network traffic and response latencies.
In the first part of the talk, I discuss a system "Vigil," which uses content-aware compression to enable real-time tracking and surveillance across smart cities. Content-aware compression uses lightweight feature extraction on the video frames at/near camera nodes to transmit prioritized content that is analyzed further at the cloud based on available bandwidth. The proposed approach combined with intelligent traffic scheduling results in a 10x wireless capacity savings over systems that upload all videos to the cloud, while maintaining the same accuracy in detecting objects of interest. In the second part of the talk, I discuss a system "Optasia" to enable efficient queries for traffic monitoring over large-scale city-wide camera installations. This system modularizes the vision modules, such as classifying vehicles by color and type, so that we can apply SQL-type relational dataflow to process the video data efficiently by discarding irrelevant columns early, de-duplicating common modules, and parallelizing the processing. This system brings together advances from two areas—machine vision and big data analytics systems to automatically parallelize computation as video input size grows or number of cameras increase. I conclude the talk by discussing how I am applying similar techniques on video from drone platforms.
Biography of Aakanksha Chowdhery:
Dr. Aakanksha Chowdhery is an Associate Research Scholar at Princeton University. Her research focuses on the network architectures and data analytics for next-generation Internet-of-Things (IoT) applications. Her work has contributed to industry standards and consortia, such DSL standards and OpenFog Consortium. She completed her PhD in Electrical Engineering from Stanford University in 2013 and was a postdoctoral researcher at Microsoft Research in Mobility and Networking Group until 2015. In 2012, she became the first woman to win the Paul Baran Marconi Young Scholar Award, given for the scientific contributions in the field of communications and the Internet. She also received the Stanford School of Engineering Fellowship and the Stanford's Diversifying Academia Recruiting Excellence (DARE) fellowship. Prior to joining Stanford, she completed her Bachelor's degree at IIT Delhi where she received the President's Silver Medal Award.