ECE Colloquium Series with Dr. Fatih Porikli

  • Starts: 4:00 pm on Wednesday, November 17, 2010

Learning on Manifolds

With Dr. Fatih Porikli, Mitsubishi Electric Research Laboratory, Cambridge, Mass.

Faculty Host: Janusz Konrad

Refreshments will be served outside Room 339 at 3:45 p.m.

About the Presentation: A large number of natural phenomena can be formulated as inference on differentiable manifolds. More specifically in computer vision, such underlying notions emerge in multi-factor analysis including feature selection, pose estimation, structure from motion, appearance tracking, and shape embedding. Unlike Euclidean spaces, differentiable manifolds do not exhibit global homeomorphism, thus, differential geometry is applicable only within the local tangent spaces. This prevents direct application of conventional inference and learning methods that require vector norms. Instead, distances are defined through curves of minimal length connecting two points. Recently, Dr. Porikli and his team introduced appearance based descriptors and motion transformations that exhibit Riemannian manifold structure on positive definite matrices and enable projections onto the tangent spaces. In this manner, they do not need to flatten the underlying manifold or discover its topology. For instance, by imposing weak classifiers on tangent spaces and establishing weighted sums via Karcher means, they bootstrap an ensemble of boosted classifiers with logistic loss functions for object classification. This talk will demonstrate promising results of manifold learning on human detection, regression tracking, unusual event analysis, and affine pose estimation.

About the Speaker: Dr. Fatih Porikli is currently a senior principal research scientist and technical manager at Mitsubishi Electric Research Laboratory (MERL). Before joining MERL in 2000, he developed satellite applications at Hughes Research Laboratories in 1999 and 3D imaging at AT&T Research Laboratories in 1997. His research interests include system design, sparse optimization, pattern recognition, online learning, computer vision, multimedia processing, medical data analysis, and data mining with many applications ranging from surveillance to intelligent transportation to automation to visualization. He is the associate editor for two journals, the general chair of the 2010 IEEE AVSS, co-organizer of more than 20 workshops, and a member of the organizing committee of flagship vision conferences. He has written more than 90 publications, authored over 50 patents, and mentored more than 30 PhD students. Dr. Porikli was the recipient of the R&D 100 Awards Scientist of the Year honor in 2006, the Best Paper Runner-Up Award at CVPR 2007, the Best Paper Award at OTCBVS of CVPR 2010, and half-a-dozen MELCO/MERL awards.

8 Saint Mary’s St., Room 211

Back to Calendar