Abstract:
Traditional face recognition research has focused on the PIE (Pose, Illumination and Expression) problem. In this talk I discuss a harder face recognition problem which we named UROPA challenge that reflects the real-world challenges faced by law enforcement almost everyday when trying to identify suspects from degraded face images. These faces images are typically acquired in the presence of real-world degradations of Unconstrained Resolution, Occlusion, Pose and Aging (UROPA). We present a unified framework for handling simultaneously off-angle pose, occlusion and low resolution. Many algorithms can address one or two of these challenges, but few can handle all of these simultaneously. This unified face model can be used to recover the whole face including the 3D face depth structure from an occluded 2D face which may also be at an off-angle pose and under low resolution. Additionally this unified face model can be trained from partial and dimensionality-deficient training data (e.g., faces which may not have 3D data, are occluded etc). We show various results under these simultaneous real-world challenging degradations.

We conclude this talk with one of the most challenging scenarios, which is identifying mostly-masked faces where only the eye-regions (periocular) are visible. We will describe a dimensionality-weighted K-SVD approach to reconstruct the whole face when only this periocular region is visible for observation. I present whole face reconstructions and matching results in real-world law-enforcement scenarios and show generalization of successful full-face reconstruction from just the periocular part of the faces in the NIST face recognition grand challenge (FRGC) face dataset.

Biography:
Marios Savvides is a Research Professor at the Electrical & Computer Engineering Department and in CyLab at Carnegie Mellon University. He is also the Founder and Director of the CyLab Biometrics Center and one of the tapped researchers to form the Office of the Director of National Intelligence (ODNI) 1st Center of Academic Excellence in Science and Technology. His research is mainly focused on developing algorithms for robust face and iris biometrics as well as pattern recognition, machine vision and computer image understanding for enhancing biometric systems performance. He has authored and co-authored over 170 journal and conference publications, including several book chapters in the area of Biometrics and served as the area editor of the Springer's Encyclopedia of Biometrics. He is the IEEE VP of Education of the IEEE Biometric Council. His achievements include leading the R&D in CMU's past participation at NIST's Face Recognition Grand Challenge 2005 (CMU ranked #1 in academia and industry at the hardest experiment #4 open challenge) and also in NIST's Iris Challenge Evaluation (CMU ranked #1 in academia and #2 against iris vendors) and developing the first long range iris system capable of capturing enrollment quality irises 40 feet away. He has filed over 12 patent applications in the area of biometrics security and is the recipient of CMU's 2009 Carnegie Institute of Technology (CIT) Outstanding Faculty Research Award.

Host:
Dr. Arun Ross and Dr. Anil K. Jain