arXiv 2511.04951

CLM: Removing the GPU Memory Barrier for 3D Gaussian Splatting

By Hexu Zhao, Xiwen Min, et al.

Published 2025-11-07

Citation lineage

Review the prior work and downstream research connected to this paper.

3D Gaussian Splatting (3DGS) is an increasingly popular novel view synthesis approach due to its fast rendering time, and high-quality output. However, scaling 3DGS to large (or intricate) scenes is challenging due to its large memory requirement, which exceed most GPU's memory capacity. In this paper, we describe CLM, a system that allows 3DGS to render large scenes using a single consumer-grade GPU, e.g., RTX4090.…

View the original paper on arXiv