Department of Computer Engineering
S E M I N A R
Unconstrained-Pose 2D Face Recognition by Matching Using Graphical Models
Asst. Prof. Dr. Shervin Rahimzadeh Arashloo
Tarbiat Modares University, Tehran, Iran
Face recognition in unconstrained settings will be discussed. The presentation will cover a Markov random field (MRF)-based methodology applied to the face matching problem in 2D along with several innovative approaches taken in this direction. By virtue of a dense pixel-wise MRF image matching model, the proposed approach can tolerate moderate pose, scale, rotation and some other undesired image perturbations. The proposed method can operate using only a single image per person and circumvents the need for large training sets or non-frontal training images. From a technical point of view, different aspects of the algorithm such as - deformable image matching - Shape regularisation using a statistical prior - Multi-scale relaxation - A GPU implementation - A class-specific KDA classifier - And last but not least, a technical comparison of the proposed approach to other methods will be discussed will be covered. Shervin Rahimzadeh Arashloo Assistant professor, Tarbiat Modares University, Tehran, Iran S.Rahizmadeh@modares.ac.ir Visiting Research Fellow, University Of Surrey, Guildford, UK S.Rahimzadeh@surrey.ac.uk
Shervin Rahimzadeh Arashloo obtained his Ph.D. in computer vision from the Centre for Vision, Speech and Signal Processing (CVSSP) at university of Surrey, UK. His research interests include cognitive vision, pattern analysis and machine learning. He has published in major computer vision and machine learning journals including PAMI, CVIU, PRL, TMM, TIFS, PAA, MVA, VCIR, etc. He is currently an assistant professor with Tarbiat Modares university, Tehran, Iran and also holds a visiting research fellow position with CVSSP, university of Surrey, UK.
DATE: 15 January 2018, Monday @ 13:40