Towards Understanding Online Visual Media

Dr. Gang Hua, Nokia Research Center


Abstract:

The proliferation of internet and visual sensors have created tremendous amount of online visual media. To assist the users to more conveniently access and share these visual media, it is an emerging need to intelligently analyze, understand, and annotate the content of them. It is not a trivial problem due to the unconstrained setting - essentially we are facing all the possible challenges a computer vision system can expect. Moreover, we are also confronted by scalability issues, both in terms of computation and memory storage.

In this talk, I will present how to approach to this problem by synergistically leveraging four elements, namely feature, learning, context and users. In particular, I will present our latest work on learning computationally efficient, robust, and yet compact local image descriptors. In addition, I will present one case study: contextual face recognition for online photo annotation, to demonstrate how different levels of contexts, e.g., spatial, temporal, and even social level context, can be employed to improve computer vision algorithms. The importance of having users in the loop will also be discussed. If time permitted, in the end of the talk, I will present one slide summary of several other works related to this topic.

 


Short Bio:

Dr. Gang Hua is a Senior Researcher in Nokia Research Center, Hollywood. Before that, he was a Scientist in Microsoft Live Labs Research from 2006 to 2009. He obtained his Ph.D. degree in Electrical and Computer Engineering from Northwestern University in 2006. His current research interests include computer vision, pattern recognition, and machine learning. He is the lead guest editor of IEEE Transaction on Pattern Analysis and Machine Intelligence Special Issue on "Real-world Face Recognition, a program co-chair of the IEEE International Workshop on "Mobile Vision" As of March, 2010, he holds 1 US patent and has 18 more patents pending. He is a member of both IEEE and ACM.

Host: Professor J. Griffioen

Speaker's personal page.