Supervised Hierarchical Pitman-Yor Process for Natural Scene Segmentation

From conventional wisdom and empirical studies of annotated data, it has been shown that visual statistics such as object frequencies and segment sizes follow power law distributions. Using these two as prior distributions, the hierarchical Pitman-Yor process has been proposed for the scene segmentation task. In this paper, we add label information into the previously unsupervised model. Our approach exploits the labelled data by adding constraints on the parameter space during the variational learning phase. We evaluate our formulation on the LabelMe natural scene dataset, and show the effectiveness of our approach.

• Type: Journal
• Year: 2011
• Where: CVPR
• Authors: Alex Shyr, Trevor Darrell, Michael Jordan, Raquel Urtasun
• Keywords: Computer Vision, CVPR 2011, Hierarchical Pitman-yor Process, Power Law Distributions, Scene Segmentation Task