Reproduce “Nebula: Enable City-Scale 3D Gaussian Splatting in Virtual Reality via Collaborative Rendering and Accelerated Stereo Rasterization” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
He Zhu, Zheng Liu, Xingyang Li, Anbang Wu, Jieru Zhao, Fangxin Liu, Yiming Gan, Jingwen Leng and 1 more
Nebula introduces a collaborative rendering framework and accelerated stereo rasterization to enable city-scale 3D Gaussian Splatting in Virtual Reality. The system achieves a 12.1x motion-to-photon speedup and reduces bandwidth requirements by 1925% compared to lossy video streaming, while maintaining visual quality at VR-required frame rates.
Key insights
- For city-scale 3DGS, the Level-of-Detail (LoD) search, rather than rasterization, constitutes the primary computational and memory bottleneck, making it ideal for cloud offloading.
- The set of active Gaussians ("cut") for rendering exhibits high temporal coherence across consecutive frames, enabling efficient localized updates and reduced data transfer.
- Binocular VR rendering possesses inherent geometric similarities between left and right eye views, allowing for significant computational sharing without compromising visual fidelity.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.