Advanced Lecture Series in Pattern Recognition
题 目 (TITLE)：Probabilistic Elastic Part Model: A Pose-Invariant Representation for Face Recognition
讲 座 人 (SPEAKER)：Dr. Gang Hua (Computer Science in Stevens Institute of Technology)
主 持 人 (CHAIR)： Prof. Liang Wang
时 间 (TIME)：November 12(Thursday), 2015, 10:00AM
地 点 (VENUE)：No.2 Conference Room (3rd floor), Intelligence Building
One of the major visual complications confronting face recognition is pose variation. It is generally perceived that a part based representation for faces would be more robust to such pose variations. Instead of adopting a set of hand-crafted parts, we take a data driven approach to a probabilistically aligned part model, namely probabilistic elastic part (PEP) model. The model is achieved by fitting a spatial appearance Gaussian mixture model (GMM) on dense local features extracted from a set of pose variant face images. For a single face image or a track of face images, each mixture component of the learned spatial appearance GMM selects one local feature which induced the highest probability on it. These selected local features are concatenated to form the final pose invariant representation, namely the PEP representation. Based on the PEP representation, we further develop the Eigen-PEP model and representation, which is a video/subject level representation of only 100 dimensions. We apply the PEP representation for both face verification and identification in the wild, and unsupervised face detector adaptation. For face verification, the PEP model achieved state-of-the-art accuracy on the Labeled face in the Wild dataset, the YouTube Video face datasets, and the recent PaSC Video Face and Person Identification Competition. For video face identification, we achieve the top performance on the recently published Celebrity-1000 video face database. For unsupervised face detector adaptation, we observed significant detection performance improvement adapting two state-of-the-art face detectors on three different datasets.
Bio: Gang Hua is a Senior Research Manger in the Visual Computing Group at Microsoft Research Asia. He was an Associate Professor of Computer Science in Stevens Institute of Technology between 2011 and 2015. He held an Academic Advisor position at IBM T. J. Watson Research Center between 2011 and 2014. He was a visiting researcher at Microsoft Research Asia in 2013, and a Consulting Researcher at Microsoft Research in 2012. Before joining Stevens, he had worked as full-time Researchers at leading industrial research labs for IBM, Nokia, and Microsoft. He received the Ph.D. degree in Electrical and Computer Engineering from Northwestern University in 2006. His research in computer vision studies the interconnections and synergies among the visual data, the semantic and situated context, and the users in the expanded physical world, which can be categorized into three themes: human centered visual computing, big visual data analytics, and vision based cyber-physical systems. He is the author of more than 100 peer reviewed publications in prestigious international journals and conferences. His research was funded by NSF, NIH, ARO, ONR, Adobe Research, Google Research, Microsoft Research, and NEC Labs. He is the receipient of the 2015 IAPR Young Biometrics Investigaotor Award. To date, he holds 14 U.S. patents and has 10 more U.S. patents pending. He is a Senior Member of the IEEE and a life member of the ACM.