GAMES Webinar 2020 – 155期(几何专题) | Yifei Shi (NUDT), Hsueh-Ti Derek Liu (University of Toronto)
【GAMES Webinar 2020-155期】(几何专题)
报告嘉宾1：Yifei Shi (NUDT)
报告题目： SymmetryNet: Learning to Predict Reflectional and Rotational Symmetries of 3D Shapes from Single-View RGB-D Images
We study the problem of symmetry detection of 3D shapes from single-view RGB-D images, where severely missing data renders geometric detection approach infeasible. We propose an end-to-end deep neural network which is able to predict both reflectional and rotational symmetries of 3D objects present in the input RGB-D image. Directly training a deep model for symmetry prediction, however, can quickly run into the issue of overfitting. We adopt a multi-task learning approach. Aside from symmetry axis prediction, our network is also trained to predict symmetry correspondences. The proposed method achieves state-of-the-art performance on a new benchmark of 3D symmetry detection based on single-view RGB-D images. In particular, it is robust in handling unseen object instances with large variation in shape, multi-symmetry composition, as well as novel object categories.
Yifei Shi (施逸飞) is an Assistant Professor at National University of Defense Technology (NUDT). Before that, he obtained his PhD degree from NUDT in 2019, advised by Kai Xu. During Mar. 2017- Aug. 2018, he was a visiting student at Princeton University, advised by Thomas Funkhouser. His research interests are in computer vision, computer graphics and robotics. His past research efforts focus on object analysis, scene reconstruction and modeling.
报告嘉宾2： Hsueh-Ti Derek Liu (University of Toronto)
报告题目：3D modeling for everyone
Creating digital 3D objects has been a central task across different disciplines. Recent advances in immersive technologies (e.g., augmented and virtual realities) and 3D printing further expand the need of designing complex 3D shapes. Despite a large demand for novel 3D content, the current state of 3D modeling tools is so far from being usable by the general public. Unlike traditional 3D modeling which edits surface points directly, I propose new perspectives on manipulating 3D geometry. In this talk, I will cover three aspects: (1) appearance-driven editing, (2) normal-driven editing, and (3) data-driven machine learning modeling. I will show how these perspectives can lead to easy-to-use techniques for creating detailed and stylized 3D objects. I argue that such modeling techniques will push the boundaries of how we model 3D shapes, leading to a world where everyone can easily create novel 3D content in their daily lives.
Hsueh-Ti Derek Liu is a third-year Ph.D. student at the University of Toronto, advised by Prof. Alec Jacobson on digital geometry processing. He completed his M.S. with Prof. Keenan Crane and Prof. Levent Burak Kara at Carnegie Mellon University. Derek’s work is published at top-tier graphics venues, such as SIGGRAPH, SIGGRAPH Asia, Eurographics, and Symposium on Geometry Processing. His research is supported by the Adobe Research Fellowship, the Robert E. Lansdale and Okino Computer Graphics Graduate Fellowship, the Mary H. Beatty Fellowship, and the Mitacs Globalink Research Award.
徐凯，国防科技大学教授。2011年于国防科大计算机学院获得博士学位。西蒙弗雷泽大学、普林斯顿大学访问学者。研究方向为数据驱动的几何处理与建模、三维视觉及其机器人应用等。发表ACM SIGGRAPH/Transactions on Graphics论文20余篇。共发表CCF A类论文30余篇。担任ACM Transactions on Graphics、Computer Graphics Forum、Computers and Graphics和The Visual Computer等期刊的编委。担任CAD/Graphics 2017、ISVC 2018等国际会议的论文共同主席，以及SIGGRAPH、Eurographics等国际会议的程序委员。现任中国图象图形学会三维视觉专委会副主任，中国工业与应用数学学会几何设计与计算专委会秘书长。曾获湖南省自然科学一等奖、军队科技进步二等奖、全军优秀博士论文奖、几何设计与计算青年学者奖、湖湘青年英才奖、陆增镛CAD&CG高科技奖二等奖。获国家优秀青年基金和湖南省杰出青年基金。
GAMES主页的“使用教程”中有 “如何观看GAMES Webinar直播？”及“如何加入GAMES微信群？”的信息；