GAMES Webinar 2020 – 143期(计算机视觉专题) | Kaichun Mo(Stanford University), Fanbo Xiang(University of California San Diego), Fei Xia(Stanford University)
【GAMES Webinar 2020-143期】(计算机视觉专题)
报告嘉宾1：Kaichun Mo(Stanford University)
报告时间：2020年6月18号星期四上午10:00 – 10:30（北京时间）
报告题目：Part-level and Structural Understanding for 3D Shape Perception, Synthesis and Editing
Part-level and structural understanding of 3D shapes is essential for many vision, graphics, and robotics tasks, such as part segmentation, shape synthesis and robot-object interaction. In this talk, I will briefly summarize my recent research efforts on building PartNet (CVPR 19), a large-scale dataset providing hierarchical and fine-grained shape part segmentation; StructureNet (Siggraph Asia 19), a hierarchical graph network for learning structure-aware shape generation; and StructEdit (CVPR 20), learning local shape edits (shape deltas) space that captures both discrete structural changes and continuous variations.
Kaichun Mo is currently a 4th-year CS Ph.D. student at Stanford supervised by Prof. Leonidas Guibas in the Geometry Computing Group and Stanford AI lab. Before that, he received his B.S.E. in CS from the ACM class, Zhiyuan College, Shanghai Jiao Tong University. He has interned at Adobe Research, Autodesk Research and Facebook AI Research. His research interests include 3D vision, graphics, and robotics. His past research efforts focus on developing 3D deep learning framework on point cloud data (PointNet), building up fine-grained and hierarchical shape part segmentation dataset (PartNet), and structural understanding of 3D shapes (StructureNet, StructEdit).
报告嘉宾2： Fanbo Xiang(University of California San Diego)
报告时间：2020年6月18号星期四上午10:30 – 11:00（北京时间）
报告题目：SAPIEN: A SimulAted Part-based Interactive ENvironment
Building home assistant robots has long been a goal for vision and robotics researchers. To achieve this task, a simulated environment with physically realistic simulation, sufficient articulated objects, and transferability to the real robot is indispensable. We take one step further in constructing an environment that supports household tasks for training robot learning algorithm. Our work, SAPIEN, is a realistic and physics-rich simulated environment that hosts a large-scale set of articulated objects. SAPIEN enables various robotic vision and interaction tasks that require detailed part-level understanding. We hope that SAPIEN will open research directions yet to be explored.
Fanbo Xiang is a master in computer science and PhD candidate at UC San Diego, advised by professor Hao Su. He got his BS in computer science and BS in mathematics at University of Illinois at Urbana-Champaign. His research interests lie in graphics, simulation, and robotics. He is also interested in software engineering, HCI and music information retrieval.
报告嘉宾3： Fei Xia(Stanford University)
报告时间：2020年6月18号星期四上午11:00 – 11:30（北京时间）
报告题目：Gibson Environment: Photorealistic Simulation for Embodied Visual Agents
Gibson Environment is a simulation environment that supports the rendering of high-fidelity images from 3D reconstructed buildings. This environment is fundamental to train visuo-motor navigation skills for robotic agents. These models learn to navigate within the simulated environment based on visual information. We further developed our simulator to enable not only navigation, but also interactions with the simulated environment. This opens new avenues in robotics, enabling agents to be trained for new tasks in simulation while maintaining a simple sim2real transfer, thanks to the visual realism. We’ll present our techniques to improve photo realism (computer vision and computer graphics) and examples of navigation and interactive agents trained in Gibson (robotics, machine learning, and reinforcement learning based on visual information)
Fei Xia is a fourth-year PhD student at Stanford University. He is advised by Silvio Savarese and Leo Guibas. His research interests lie in Computer Vision and Machine Learning. In particular, he is interested in simulation to real-world transfer and domain adaptation for vision and robotics tasks.
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；