Yunchao Wei

Professor and Vice Dean, School of Computer Science, Beijing Jiaotong University

Recipient of the National High-level Talent Programme. He has conducted research at National University of Singapore, University of Illinois at Urbana-Champaign, and University of Technology Sydney. Selected as MIT TR35 China, Baidu Global High Potential Chinese Young Scholar, The Australian Top 40 Rising Star; awarded the Leading Science and Technology Award of the World Internet Conference (2023), the First Prize of Natural Science Award of Higher Education Institutions of the Ministry of Education (2022), the First Prize of Science and Technology Award of the Chinese Society of Graphics and Graphics (2019), the Australian Research Council Young Research Award (2019), IBM C3SR Best Research Award (2019), ImageNet Target Detection Champion of the Computer Vision World Cup (2014), and several CVPR competitions; published more than 100 papers in top journals/conferences, such as TPAMI, CVPR, etc., with more than 22000 Google citations. His current main research interests include visual perception for imperfect data, multimodal data analysis, and generative artificial intelligence.

Topic

From 3D to 4D, an Exploration of Rapid and Temporally Consistent 4D Content Generation

In recent years, 3D content generation has evolved rapidly with the development of 2D generative macromodels, however, dynamic 3D (4D) content generation has been less explored due to the lack of large 4D datasets and robust video pre-training models. This presentation aims to investigate fast and spatio-temporally consistent 4D content generation from a 3D generation perspective, and will introduce the latest work on 4D generation, 4DGen and Diffusion4D, by exploring how to design spatio-temporally high fidelity supervised signals and how to use 4D datasets to fine-tune video generation models. fine-tuned to enable fast, high-quality generation from text, image, or video control signals to 4D content. Outline: 1. introduce the background of AIGC development and explain the important work on 2D and 3D generation 2. Introduce 4D generation work 4DGen 3.Introduce 4D generation work Diffusion4D 4.Summarise the progress of 4D content generation and put forward the outlook