时间：12月10日星期二 下午15：00—16：30
地点：浙大紫金港校区CAD&CG国家重点实验室418室
题目：3D Modeling based on Geometry and Structure Priors
报告人：Dr. Tianjia Shao
主持人：吴鸿智 博士

Abstract:3D modeling is the key part in geometric graphics. As recent advances in computer graphics technologies dramatically increased the requirements of 3D models, traditional methods don't have enough power to efficiently generate models of high quality or with semantics, especially for complex objects or scenes. Recent research effort has shifted its focus to modeling with the help of human-provided prior knowledge. In this presentation, I will present three of my recent works on 3D modeling based on geometry and structure priors. The first work is about a sketch-based 3D model retrieval system, which uses a robust shape matching algorithm to measure the similarity between user’s sketches and 2D contours of 3D models. In my second work, a system to facilitate easy interpretation and exploration of concept sketches is presented. The algorithm generates an abstract functional model from concept sketches, with the geometry prior that product parts can be approximated as primitive shapes and the structure prior that a set of joint types are prescribed for contact relations between parts. In my third work, an interactive approach to semantic modeling of indoor scenes with a consumer-level RGBD camera is proposed, with the geometry prior that the depth data and 3D models of objects of same types are similar.

Bio:Tianjia Shao is a Ph.D. candidate in Institute for Advanced Study, Tsinghua University. Before this, he received the B.Eng. from the Department of Automation, Tsinghua University, in 2008. He was a visiting student in Univeristy College London, UK, supervised by Prof. Mitra, from 2012 to 2013. From 2008 to 2012, he was a research intern in IG Group in MSRA. His research interests include indoor scene analysis and modeling, 3D model analysis and retrieval, and sketch ananlysis.