EE Seminar: Scene Labeling with Supervised Contextual Models
  • FENS
  • EE Seminar: Scene Labeling with Supervised Contextual Models

You are here

Title: Scene Labeling with Supervised Contextual Models

Speaker: Dr. Tolga Tasdizen

Date/Time: 24th December-Wednesday-at 12:40

Place: FENS G029


Abstract:  Scene labeling is the problem of assigning an object label to each pixel. It unifies the image segmentation and object recognition problems. In this talk, we will describe our contextual hierarchical model (CHM), which learns contextual information in a hierarchical framework for scene labeling. At each level of the hierarchy, a classifier is trained based on down sampled input images and outputs of previous levels. CHM then incorporates the resulting multi-resolution contextual information into a classifier to segment the input image at the original resolution. This training strategy allows for optimization of a joint posterior probability at multiple resolutions through the hierarchy. We will also introduce a novel classifier that we call Logistic Normal Disjunctive Networks, which allows efficient training for CHM. Our approach is applicable to a variety of problems such as object segmentation and edge detection. We demonstrate that CHM outperforms state-of-the-art methods on the Stanford background and the Weizmann horse datasets. It also outperforms state-of-the-art edge detection methods on the NYU depth dataset and the Berkeley segmentation dataset (BSDS 500). Finally, we will demonstrate our results on segmentation of electron microscopy images of neuropil for connectomics.

Bio: Dr. Tolga Tasdizen is an Associate Professor of Electrical and Computer Engineering at the University of Utah. He is currently a visiting faculty member on sabbatical leave at Sabanci University. Dr. Tasdizen received his PhD from Brown University in 2001. He was a postdoctoral associate with the Scientific Computing and Imaging Institute at the University of Utah 2001-4. He is a Senior Member of the IEEE, IEEE Signal Processing Society and IEEE Computer Society. Dr. Tasdizen's research interests are in the general areas of image processing and pattern recognition with applications in bioimaging and medical image analysis. Dr. Tasdizen is the recipient of several awards including the NSF Early Career Award (2012), University of Utah College of Engineering Outstanding Teacher Award (2012) and the TUBITAK Fellowship for Visiting Scientists and Scientists on Sabbatical Leave (2014). Dr. Tasdizen is currently an Associate Editor for IEEE Signal Processing Letters. In the past, he as served as an Associate Editor for BMC Bioinformatics (2012-14). He has served in the IEEE Signal Processing Society's Bio Imaging and Signal Processing TC as an Associate Member (2009-11) and as a Regular Member (2012-14).
Contact: Mehmet Keskinoz