What is it about?
The resolution for virtual reality display is vital. We present a technique for reconstructing a high-resolution 3D scene that can be viewed at new viewpoints even with only low-resolution input images. Our work only need images of the target scene and is free of additional training data.
Featured Image
Photo by stephan sorkin on Unsplash
Why is it important?
Our finding provide the first solution to super-resolve not a single image, but the whole 3D scene. Also, we don't need much data to train, only a few images of the scene.
Perspectives
I hope this article can further boost people's interest for 3D computer vision and virtual reality technologies. Although it still has some distance for common users, it is fast-evolving and life-changing. People will be amazed at how their life would be with these devices - a total virtual world beyond their imagination.
Chen Wang
Tsinghua University
Read the Original
This page is a summary of: NeRF-SR: High Quality Neural Radiance Fields using Supersampling, October 2022, ACM (Association for Computing Machinery),
DOI: 10.1145/3503161.3547808.
You can read the full text:
Resources
Contributors
The following have contributed to this page
