About the Event
This talk addresses one of the basic questions in computer vision, that of finding a general-purpose object representation which could be used for diverse computations of image understanding. The specific representation we consider is a stochastic image grammar (SIG). Since many objects of interest in the real world are spatially localized, their image projections correspond to regions, which we take as the primitives of SIG. The SIG rules encode probabilistic object definitions in terms of properties of image regions, including their photometry and geometry (e.g., color, texture, shape), structure (e.g., number of parts and their layout), and spatial context (e.g., spatial layout of neighboring regions). In this talk, we present algorithms for learning and inference of SIG from images in a specific context of the following problems: (i) Given an arbitrary, unlabeled set of images, discover and learn a taxonomy of (unknown) object categories occurring in the set; (ii) Identify and segment any textures present in a given image; and (iii) Provide a globally consistent explanation of an image in terms of recognizing and segmenting all instances of objects and their parts present.
Dr. Sinisa Todorovic received his Ph.D. degree in electrical and computer engineering at the University of Florida in 2005. He was Postdoctoral Research Associate in the Beckman Institute at the University of Illinois Urbana-Champaign, between 2005-2008, where he collaborated with Prof. Narendra Ahuja. Currently, Dr. Todorovic is Assistant Professor in the School of EECS at the Oregon State University. His research focuses on computer vision and machine learning problems. He is Associate Editor of Image and Vision Computing, and Pattern Recognition Letters. He is also on Editorial Advisory Board of Computer Vision Central. He co-organized 1st International Workshop on Stochastic Image Grammars in 2009. For a paper published in IEEE Transactions on Vehicular Technology, he was awarded 2004 Jack Neubauer Best Paper Award by the IEEE Vehicular Technology Society.