Scaling Point-based Differentiable Rendering for Large-scale Reconstruction

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhao, Hexu, Liu, Xiaoteng, Min, Xiwen, Huang, Jianhao, Deng, Youming, Li, Yanfei, Li, Ang, Li, Jinyang, Panda, Aurojit
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908730072760320
author Zhao, Hexu
Liu, Xiaoteng
Min, Xiwen
Huang, Jianhao
Deng, Youming
Li, Yanfei
Li, Ang
Li, Jinyang
Panda, Aurojit
author_facet Zhao, Hexu
Liu, Xiaoteng
Min, Xiwen
Huang, Jianhao
Deng, Youming
Li, Yanfei
Li, Ang
Li, Jinyang
Panda, Aurojit
contents Point-based Differentiable Rendering (PBDR) enables high-fidelity 3D scene reconstruction, but scaling PBDR to high-resolution and large scenes requires efficient distributed training systems. Existing systems are tightly coupled to a specific PBDR method. And they suffer from severe communication overhead due to poor data locality. In this paper, we present Gaian, a general distributed training system for PBDR. Gaian provides a unified API expressive enough to support existing PBDR methods, while exposing rich data-access information, which Gaian leverages to optimize locality and reduce communication. We evaluated Gaian by implementing 4 PBDR algorithms. Our implementations achieve high performance and resource efficiency: across six datasets and up to 128 GPUs, it reduces communication by up to 91% and improves training throughput by 1.50x-3.71x.
format Preprint
id arxiv_https___arxiv_org_abs_2512_20017
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Scaling Point-based Differentiable Rendering for Large-scale Reconstruction
Zhao, Hexu
Liu, Xiaoteng
Min, Xiwen
Huang, Jianhao
Deng, Youming
Li, Yanfei
Li, Ang
Li, Jinyang
Panda, Aurojit
Distributed, Parallel, and Cluster Computing
Graphics
C.0; I.3.2; I.4.5
Point-based Differentiable Rendering (PBDR) enables high-fidelity 3D scene reconstruction, but scaling PBDR to high-resolution and large scenes requires efficient distributed training systems. Existing systems are tightly coupled to a specific PBDR method. And they suffer from severe communication overhead due to poor data locality. In this paper, we present Gaian, a general distributed training system for PBDR. Gaian provides a unified API expressive enough to support existing PBDR methods, while exposing rich data-access information, which Gaian leverages to optimize locality and reduce communication. We evaluated Gaian by implementing 4 PBDR algorithms. Our implementations achieve high performance and resource efficiency: across six datasets and up to 128 GPUs, it reduces communication by up to 91% and improves training throughput by 1.50x-3.71x.
title Scaling Point-based Differentiable Rendering for Large-scale Reconstruction
topic Distributed, Parallel, and Cluster Computing
Graphics
C.0; I.3.2; I.4.5
url https://arxiv.org/abs/2512.20017