Title: Skeletons and Canonical Views for Fast Model Understanding
	Speaker: Wencheng Wang
	Bio: Wencheng Wang is now a professor and a deputy director of the State Key Laboratory of Computer Science at ISCAS. His interests include scientific visualization, virtual reality, image editing and computational geometry, and with focus on fast rendering and object understanding.
	Abstract: It is much required to quickly investigate 3D models in many applications, especially when 3D models are increasing greatly in nowadays. This requires fast understanding of both the intrinsic structures of 3D models and the surface details of 3D models. With regard to this, we recently developed methods to improve skeleton extraction and view selection. For curve skeletons, a well-known compact representation of 3D models, we improve their centeredness to represent 3D models more effectively, and meanwhile speed up the extraction by about one order of magnitude, compared with the state-of-the-art techniques. View selection aims at finding good views that can watch meaningful contents as many as possible. However, existing methods are inefficient and may mistake low-quality views as good views. Our developed method proposes to constrain view sampling in the regions that are very possible to watch many contents, which generally take up 5% of the surface of the viewing sphere. Thus, we can speed up view selection by about one order of magnitude, and guarantee that the obtained views are really good views.