What is it about?
IDE-3D is the first 3D generative model which is capable of rendering a 3D virtual portrait in real-time and further editing its local attributes interactively. We develop a UI platform for users to synthesize their own 3D virtual figures and edit them as they want. The editing is very simple. You only need to edit a semantic mask on a drawing board and then the edited 3D portrait would be rendered in real-time.
Featured Image
Photo by Jessica Felicio on Unsplash
Why is it important?
IDE-3D is a powerful 3D generative model with three characteristics: photorealism, facial disentanglement, and efficient editing. We adopt a tri-plane hybrid representation for efficient photorealistic rendering and present a method for decoupling facial attributes. Besides, our proposed canonical editor enables faithful real-time editing. In total, IDE-3D simplifies the process of making a customized 3D virtual portrait.
Perspectives
It is a great pleasure to write this paper because it lets more people realize the fantasy and practicability of 3D generation, and it would be exciting to see if IDE-3D could be deployed on mobile phones or pads in the future. Furthermore, we hope this paper would inspire more research combining other recent AI techniques, such as text to image, stylization, and so on.
Jingxiang Sun
Tsinghua University
Read the Original
This page is a summary of: IDE-3D, ACM Transactions on Graphics, November 2022, ACM (Association for Computing Machinery),
DOI: 10.1145/3550454.3555506.
You can read the full text:
Contributors
The following have contributed to this page
