On the afternoon of July 1st at 2:30 pm, Professor Yao Yao from the school invited Professor Li Hongdong from the National University of Australia to give an academic presentation titled "From 3D Visual Reconstruction to Generative AI Automatic Creation of 3D Materials".
Abstract:
We live in a three-dimensional world. The perception, understanding, processing, and creation (imagination) of three-dimensional scenes and objects is one of the important tasks of our visual information processing, and it is also a classic topic in computer vision research, with a large number of important research results published. The recent advancements in deep learning and generative AI technologies, such as ChatGPT and SDXL, have provided new opportunities for 3D visual understanding and 3D model generation. In this report, I will briefly introduce some recent research topics related to 3D reconstruction and 3D generation in our laboratory. Specifically, it includes (1) how to use neural inverse rendering technology to achieve three-dimensional geometric modeling of complex material objects through multi camera images and controllable light sources, including restoration of non diffuse materials and geometry, and GaussSplating CT modeling of semi transparent volume objects; (2) How to effectively embed guidance from multi view geometry theory in AI automatic creation tasks of 3D materials using Gen-AI technology, including AI generation of 3D objects and user controlled generation of large-scale 3D scenes. Due to time constraints, the report focuses on the introduction of basic concepts and algorithms, rather than mathematical theoretical derivation. The target audience is researchers and graduate students in related fields such as AI-ML-CV, as well as interested senior undergraduate students. The report duration is one hour (including Q&A time).
Reporter:
Li Hongdong is a tenured professor at the National University of Australia (ANU) and has served as the Vice Dean of Research at the School of Engineering Sciences at ANU. He is also the Chief Investigator at the former Australian National Center for Robot Vision Research. He is an internationally renowned expert in 3D computer vision and machine learning, and has undertaken or participated in projects such as the Australian artificial eye project, virtual city 3D reconstruction, robot navigation and autonomous driving, mixed reality and augmented reality. He has won numerous international top research paper awards, including the IEEE-CVPR Best Paper Award and the ICCV Mare Award for Honorary Nomination. During his academic vacation from 2022 to 2023, he led the research and development of new technologies and products for Tencent XR as an outstanding scientist at Tencent and the head of Tencent XR's global research laboratory. He has served as the field chair of CVPR, ICCV, and ECCV for many years, as well as the senior editor of ICRA2022, the senior field chair (SAC) of CVPR-2023 and NeuroPS-2024, and the editorial board member of the top artificial intelligence journal IEEE-TPAMI.
