Kun Zhou: Next-Generation 3D Digital Technology - Building a Colorful Digital World

On the morning of December 1, 2022, Kun Zhou, a professor at the School of Computer Science at Zhejiang University, ACM Fellow, and IEEE Fellow, delivered an exciting presentation titled "Next-Generation 3D Digital Technology - Building a Colorful Digital World" at the "Peking University Intelligent Science Frontier Forum." The presentation, conducted online, attracted the participation of many professors and students from School of Intelligent Science and Technology, Institute of Artificial Intelligence, Wangxuan Institute of Computer Technology, and School of Computer Science.

Firstly, Professor Kun briefly reviewed the development of the internet and content formats, summarizing the transition from a text and image-based content dissemination method during the portal website era to the emergence of video content in the mobile internet era. He mentioned that 3D digital content will become a core element in the next development stage of the internet, contributing to the construction of a rich content ecosystem along with other content formats. An important aspect of 3D digital content is the digitization and reconstruction of three-dimensional objects.

Next, Professor Kun introduced the research line in the digital representation of 3D objects and the reconstruction of 3D digital objects from the real world. The digital representation of 3D objects includes the representation of geometric shapes and surface textures. The surface textures, as a high-dimensional function that changes with space, lighting, and viewing angles, depict the "colorful flow of light." Subsequently, Professor Kun introduced the current state of 3D digitalization technology and systems, summarizing the characteristics and existing problems of 3D scanners and texture scanners. He explained the challenges faced in the joint acquisition of geometric and texture information for complex objects and mentioned that data acquisition and reconstruction should be jointly optimized.

Then, Professor Kun introduced the research progress of his research group in the joint acquisition and reconstruction of geometric and texture information, including fixed texture scanners, fixed geometric and texture scanners, and handheld non-planar geometric and texture scanners. During the presentation, Professor Kun carefully explained the basic principles of their designed joint optimization software and hardware system, as well as the design ideas of the hardware devices and software algorithms. This research work maps physical sampling and computational reconstruction into differentiable deep neural networks, achieving end-to-end software and hardware joint optimization, ultimately obtaining high-precision reconstruction results with reduced sampling rates.

In the conclusion of the presentation, Baoquan Chen, Zhouchen Lin, and other teachers and students engaged in discussions on issues related to the light source mode settings in acquisition devices, simulation calculations for LED light sources, preparation of training data, and material acquisition and reconstruction for multi-object scenes. Professor Kun provided further explanations and shared more insights and perspectives, allowing the audience to better comprehend the essence of the methods and recognize the challenges and opportunities in this field. Baoquan Chen summarized and expressed gratitude for Professor Kun's excellent presentation, marking the successful conclusion of this exchange and presentation event.