Tuesday, February 10, 2015 - 10:00am to 11:00am
KEC 1007

Speaker Information

Fuxin Li
Research Scientist
School of Interactive Computing
Georgia Institute of Technology

Abstract

As visual object recognition advances, an important challenge is to understand scenes in detail, such as segmenting out all the pixels associated with each object, understanding the interactions between objects, learning the affordance of objects for performing actions, or a full 3D representation of the scene. Such tasks are important for the application of computer vision to other domains such as video surveillance, robotics and autonomous driving. A lot of those tasks involves the understanding of object shape. In this talk, I will present a bottom-up methodology for image and video object recognition that starts from an unsupervised generation of segment proposals using figure-ground segmentation algorithms. Those proposals reflect educated guesses of object shapes, and machine learning models on top of them can make predictions with the object shape in mind. A composite statistical inference framework is then presented for inferring a detailed scene interpretation, based on a set of overlapping segment proposals associated with conflicting predictions. This framework is the basis of our winning systems of the PASCAL VOC Segmentation challenge between 2009-2012. It is shown to be statistically consistent, computationally efficient and delivering state-of-the-art performances in difficult benchmarks from semantic segmentation, video segmentation, pose estimation and 3D reconstruction problems.

Speaker Bio

Dr. Fuxin Li is a research scientist from the School of Interactive Computing, Georgia Institute of Technology, working with Dr. James M. Rehg. He obtained a Ph.D. degree in 2009 from the Institute of Automation, Chinese Academy of Sciences and had since held postdoctoral appointments in the University of Bonn and Georgia Institute of Technology. Dr. Li is interested in the intersection of machine learning and computer vision, especially segmentation in images/videos and visual object recognition. He and his colleagues have won the prestigious PASCAL Visual Object Recognition challenge in Segmentation from 2009-2012. He has won a Microsoft research award, 2 best reviewer awards and is currently leading an NSF project.