arXiv 2511.04951

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

By Hexu Zhao, Xiwen Min, et al.

Published 2025-11-07

Mindmap

Browse the paper's core ideas, clusters, and relationships in a structured outline.

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