Dr. Yunhai Wang
Visual Computing Research Center
Shenzhen Institutes of Advanced Technology

Projective Analysis for 3D Shape Segmentation

Department of Computer Science
Thursday, November 14, 2013, 1:00 p.m., Room EN-2022

---

Abstract

We introduce projective analysis for semantic segmentation and labeling of 3D shapes. The analysis treats an input 3D shape as a collection of 2D projections, labels each projection by transferring knowledge from existing labeled images, and back-projects and fuses the labelings on the 3D shape. The image-space analysis involves matching projected binary images of 3D objects based on a novel bi-class Hausdorff distance. The distance is topology-aware by accounting for internal holes in the 2D figures and it is applied to piecewise-linearly warped object projections to compensate for part scaling and view discrepancies. Projective analysis simplifies the processing task by working in a lower-dimensional space, circumvents therequirement of having complete and well-modeled 3D shapes, and addresses the data challenge for 3D shape analysis by leveraging the massive available image data. A large and dense labeled set ensures that the labeling of a given projected image can be inferred from closely matched labeled images. We demonstrate semantic labeling of imperfect (e.g., incomplete or self-intersecting) 3D models which would be otherwise difficult to analyze without taking the projective analysis approach.

Biography:

Dr. Yunhai Wang is a is a postdoctoral researcher at department of Computer Science, MUN. He also holds a faculty position in Shenzhen Institutes of Advanced Technology, China. He received his Ph.D. degree from Supercomputing Center, Chinese Academy of Sciences in 2011. His research deals with the development of visualization and computer graphics techniques that help people see and understand the data. He is particularly interested in developing machine learning algorithms for data visualization and shape analysis.