Proactive 3D Sensing and Reconstruction

Invited Talk- Baoquan Chen

Endowed professor of computer science, executive director of the Center o​n Frontiers of Computing Studies (CFCS), Peking University

3D modeling of urban environments starts to play an increasingly important role in applications ranging from digital content production to augmented reality, and self-driving car. Despite the ever increasing intelligence of data processing, one fundamental challenge for geometry reconstruction of complex scenes is dealing with significant variance on data quality and the usual insufficiency. In this talk, I will introduce our recent development on using robot-based scanning system, which facilitates proactive acquisition, thus a closed-loop approach to acquiring, understanding, and reconstructing 3D real-world environments. The core of the approach is the in-situ analysis of data, semantical parsing of the scene, and immediate guidance of the scanning trajectory. For 3D scene parsing, we have developed PointCNN, a convolutional network native to 3D point clouds.


Bio:

Baoquan Chen is an endowed professor of computer science, also ​executive director of the Center o​n Frontiers of Computing Studies (CFCS), Peking University. He is also the Chief Scientist of Beijing Innovation Center for Future Visual Entertainments, Beijing Film Academy. His research interests generally lie in computer graphics, computer vision, and visualization. His publications include 30+ ACM SIGGRAPH/Trans. on Graphics (TOG) papers. He serves/served on the editorial board of ACM TOG/IEEE TVCG. He has also served on the technical committee of all major conferences in computer graphics and visualization, and served as conference chair of SIGGRAPH Asia 2014 and conference co-chair of IEEE Visualization 2005/2006. Chen received an MS in Electronic Engineering from Tsinghua University, Beijing (1994), and PhD (1999) in Computer Science from the State University of New York at Stony Brook.

Web: https://cfcs.pku.edu.cn/baoquan/