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.