When: Friday, May 09, 3:00 p.m.

Where: 1305Newell-Simon Hall

Peter Tu and Xiaoming Liu,
GE Global Research

VASC Special Seminar

Abstract:
The first part of the talk will give an overview of the various intelligence video projects being developed at the Visualization and Computer Vision Lab of GE Global Research. The second part of the talk presents a discriminative framework for efficiently aligning images. Although conventional generative model based Active Appearance Models (AAM) have achieved some success, they suffer from the generalization problem, i.e., how to align any image with a generic model. We treat the iterative image alignment problem as a process of maximizing the score of a trained two-class classifier that is able to distinguish correct alignment (positive class) from incorrect alignment (negative class). During the modeling stage, given a set of images with ground truth landmarks, we train a conventional Point Distribution Model (PDM) and a boosting-based classifier, which we call Boosted Appearance Model (BAM). When tested on an image with the initial landmark locations, the proposed algorithm iteratively updates the shape parameters of the PDM via the gradient ascent method such that the classification score of the warped image is maximized. The proposed framework is applied to the face alignment problem. Using extensive experimentation, we show that, compared to the AAM-based approach, this framework greatly improves the robustness, accuracy and efficiency of face alignment by a large margin, especially for unseen data. We will also present two recent extensions to BAM. On the modeling side, we have employed a ranked learning scheme to ensure that a convex alignment cost surface is obtained. On the feature representation side, extracting features from the image observation space enables the fitting process to explicitly take advantage of edge information. Parts of this talk have been/will be published in several vision conferences, including BMVC 06,07, CVPR 07,08, and ICCV 2007.

Dr. Peter Tu, Ph.D. Oxford University. Dr. Tu joined GE Global Research in 1997. Prior to this, he was a member of the Sony Computer Vision Group in Tokyo, Japan. He has developed a number of algorithms for latent fingerprint matching that have been incorporated into the FBI AFIS system. Dr. Tu has also made contributions to GE's optical metrology systems, which are used to make high precision 3D shape measurements on manufactured parts. He has developed a number of techniques based on Helmholtz imaging which directly addresses issues associated with specularity and high curvature. Currently, Dr. Tu is focused on multi-view surveillance with the aim of achieving reliable behavior recognition in complex environments. He manages the $4 million intelligent video research effort. He has authored more than 25 publications, has more than 20 U.S. patents pending. Peter has served as PI for the NIJ "High Quality 3D Facial Images from Surveillance Video" the FBI "ReFace" program.

Dr. Xiaoming Liu is a research scientist at General Electric (GE) Global Research. He received the B.E. degree from Beijing Information Technology Institute, Beijing, China and the M.E. degree from Zhejiang University, Hangzhou, China, in 1997 and 2000 respectively, both in Computer Science, and the Ph.D. degree in Electrical and Computer Engineering from Carnegie Mellon University (CMU), in 2004. His research interests include computer vision, pattern recognition, and machine learning, with a recent focus on facial image processing in the context of surveillance videos. At GE, he is the PI for the current NIJ "Site-Adaptive Face Recognition at a Distance" program and the project leader of the BIRD "ID Kiosk" program. He has lead the execution of the NIJ "Active 3D Face Capture" program and was the main contributor of the NIJ "High Quality 3D Facial Images from Surveillance Video" program. He has authored more than 40 peer-reviewed scientific publications, and has over 10 U.S. patents pending.

Appointments: pm1e@andrew.cmu.edu

<< Back