Search papers, labs, and topics across Lattice.
This paper introduces Atlas, a novel framework for on-device city-scale 3D Gaussian splatting (3DGS) that addresses the memory limitations of mobile VR devices by dynamically loading only the necessary Gaussian data for rendering. By leveraging a hierarchical memory offloading mechanism and optimizing for temporal awareness in level of detail (LoD) search, Atlas achieves significant performance improvements, achieving an 18.5x speedup over traditional GPU baselines and a 3.9x speedup compared to existing 3DGS accelerators. The framework not only reduces computational redundancy but also integrates seamlessly with current 3DGS hardware with minimal overhead, resulting in 92.4% energy savings.
Achieving 18.5x faster city-scale 3D rendering on mobile VR devices while slashing energy consumption by 92.4% could redefine immersive experiences in urban environments.
3D Gaussian splatting (3DGS) has drawn significant attention in the architectural community recently. However, enabling city scale 3DGS on mobile VR devices remains challenging, as the memory requirement of large scale scenes far exceeds the memory capacity of today's mobile GPUs. This paper presents Atlas, an on device city scale 3DGS rendering framework that enables scalable rendering without runtime Internet access. The key insight is that although the full 3DGS model is massive, each frame only requires a small subset of Gaussians under the current pose and level of detail requirement. Based on this insight, Atlas introduces a hierarchical memory offloading mechanism that dynamically loads only necessary Gaussian data into device memory. To further improve performance, Atlas proposes temporal aware LoD search and stereo rasterization to avoid redundant computation in VR. We further show that our technique can be integrated with existing 3DGS accelerators with negligible hardware overhead. Overall, Atlas achieves 18.5x speedup over the GPU baseline and 3.9x speedup over the state of the art 3DGS accelerators, with 92.4% energy savings.
