3D Scene Reconstruction using RGB-D Cameras and AI: From Static to Dynamic Scenes

Abstract:  

Detailed 3D scene reconstruction is crucial for 3D scene understanding, with applications in domains like robotics, autonomous driving, 3D simulation or life support. How to capture 3D information from various sensors, track camera motion, track objects or people’s (non-rigid) deformation and fuse all captured data into a single unified 3D shape representation are key challenges that have driven many research in computer vision for the last decades. These goals are also the core motivation of my research. In this talk, I will discuss about the development of both hardware and AI solutions for indoor 3D scene reconstruction. We will see that the technology is getting mature for static indoor scene reconstruction, even at relatively large scales. I will also talk about reconstruction of dynamic 3D shapes such as the human body, and AI-based solutions that are opening new exciting possibilities for 3D shape estimation using simple set-ups such as even a single color image.

Short bio:

Diego Thomas received his master's degree in Informatics and Applied Mathematics from the Ecole Nationale Superieure d'Informatique et de Mathematiques Appliquees de Grenoble (ENSIMAG), France in 2008. He received his Ph.D. from the Graduate University for Advanced Studies (SOKENDAI), Japan in 2012. He was a post-doc researcher at the National Institute of Informatics (Tokyo, Japan) from April 2012 to March 2015.  From April 2015 to March 2017, he was a JSPS post-doc researcher at the University of Kyushu (Fukuoka, Japan). From April 2017, he is assistant professor of Graduate School of Information Science and Electrical Engineering (ISEE), Kyushu University. His research interests are related to 3D Vision and Computer Graphics: RGB-D SLAM, 3D reconstruction of dynamic scenes, human body animation and AR/VR. He has published more than 50 research papers in top-tier conferences and journals and has reviewed more than 100 papers for various top-tier international conferences and journals. He was the local chair of 3DV 2020. His research work has been supported by Japan Society for the Promotion of Science (JSPS), JST, Kyushu University and international companies, including Microsoft, Huawei and Panasonic. From August 2020 he is a trainee of the SENTAN-Q program of Kyushu University.