Voyager:实时将城市规模的3D高斯函数绘制到您的手机上
Voyager: Real-Time Splatting City-Scale 3D Gaussians on Your Phone

原始链接: https://arxiv.org/abs/2506.02774

论文“Voyager:在手机上实时绘制城市规模的3D高斯 splatting”介绍了一种在资源有限的移动设备上渲染大型3D高斯 splatting (3DGS) 场景的方法。核心问题在于3DGS的计算需求,特别是对于城市规模的场景,这使得直接在移动设备上渲染变得不切实际,而简单的云端流式传输又过于带宽密集。Voyager的解决方案基于这样的观察:在正常的用户移动过程中,新可见的高斯数量保持相对稳定。它利用这一点,只从云端向移动客户端传输必要的高斯。这是通过在云端进行异步细节层次搜索来高效识别这些必要的高斯实现的。在客户端,渲染通过基于查找表的栅格化来加速。这些优化与其他运行时改进相结合,使Voyager能够在智能手机上提供低延迟的城市规模3DGS渲染,与现有方法相比,数据传输量减少了100多倍,速度提高了8.9倍,同时保持了类似的视觉质量。

The Hacker News thread discusses "Voyager," a real-time 3D Gaussian splatting technique for city-scale rendering on mobile phones. A user who worked on a similar project at Faro Inc. notes that the key innovation is efficient on-device rasterization using the mobile GPU's geometry shader. The conversation touches on related technologies, like Leica's Powerlock, and the recent resurgence of splatting techniques. Users note that while splatting isn't new, its application to NeRF (Neural Radiance Fields) and leveraging differentiable point samples for gradient descent, combined with deep learning advancements, has led to significant progress. Specifically, the use of Gaussian splats overcomes the earlier limitations of point splats (holes, inefficiencies) in light field capture setups. Others expressed interest in a demo and the availability of code. The Voyager demo site is currently suspended due to exceeding the limits of its free hosting plan, while the code release is expected soon.
相关文章

原文

View a PDF of the paper titled Voyager: Real-Time Splatting City-Scale 3D Gaussians on Your Phone, by Zheng Liu and 8 other authors

View PDF HTML (experimental)
Abstract:3D Gaussian Splatting (3DGS) is an emerging technique for photorealistic 3D scene rendering. However, rendering city-scale 3DGS scenes on mobile devices, e.g., your smartphones, remains a significant challenge due to the limited resources on mobile devices. A natural solution is to offload computation to the cloud; however, naively streaming rendered frames from the cloud to the client introduces high latency and requires bandwidth far beyond the capacity of current wireless networks.
In this paper, we propose an effective solution to enable city-scale 3DGS rendering on mobile devices. Our key insight is that, under normal user motion, the number of newly visible Gaussians per second remains roughly constant. Leveraging this, we stream only the necessary Gaussians to the client. Specifically, on the cloud side, we propose asynchronous level-of-detail search to identify the necessary Gaussians for the client. On the client side, we accelerate rendering via a lookup table-based rasterization. Combined with holistic runtime optimizations, our system can deliver low-latency, city-scale 3DGS rendering on mobile devices. Compared to existing solutions, Voyager achieves over 100$\times$ reduction on data transfer and up to 8.9$\times$ speedup while retaining comparable rendering quality.
From: Yu Feng [view email]
[v1] Tue, 3 Jun 2025 11:50:51 UTC (36,543 KB)
[v2] Wed, 4 Jun 2025 01:40:43 UTC (36,544 KB)
联系我们 contact @ memedata.com