Spectrally Pruned Gaussian Fields with Neural Compensation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yang, Runyi, Zhu, Zhenxin, Jiang, Zhou, Ye, Baijun, Chen, Xiaoxue, Zhang, Yifei, Chen, Yuantao, Zhao, Jian, Zhao, Hao
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909186510553088
author Yang, Runyi
Zhu, Zhenxin
Jiang, Zhou
Ye, Baijun
Chen, Xiaoxue
Zhang, Yifei
Chen, Yuantao
Zhao, Jian
Zhao, Hao
author_facet Yang, Runyi
Zhu, Zhenxin
Jiang, Zhou
Ye, Baijun
Chen, Xiaoxue
Zhang, Yifei
Chen, Yuantao
Zhao, Jian
Zhao, Hao
contents Recently, 3D Gaussian Splatting, as a novel 3D representation, has garnered attention for its fast rendering speed and high rendering quality. However, this comes with high memory consumption, e.g., a well-trained Gaussian field may utilize three million Gaussian primitives and over 700 MB of memory. We credit this high memory footprint to the lack of consideration for the relationship between primitives. In this paper, we propose a memory-efficient Gaussian field named SUNDAE with spectral pruning and neural compensation. On one hand, we construct a graph on the set of Gaussian primitives to model their relationship and design a spectral down-sampling module to prune out primitives while preserving desired signals. On the other hand, to compensate for the quality loss of pruning Gaussians, we exploit a lightweight neural network head to mix splatted features, which effectively compensates for quality losses while capturing the relationship between primitives in its weights. We demonstrate the performance of SUNDAE with extensive results. For example, SUNDAE can achieve 26.80 PSNR at 145 FPS using 104 MB memory while the vanilla Gaussian splatting algorithm achieves 25.60 PSNR at 160 FPS using 523 MB memory, on the Mip-NeRF360 dataset. Codes are publicly available at https://runyiyang.github.io/projects/SUNDAE/.
format Preprint
id arxiv_https___arxiv_org_abs_2405_00676
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Spectrally Pruned Gaussian Fields with Neural Compensation
Yang, Runyi
Zhu, Zhenxin
Jiang, Zhou
Ye, Baijun
Chen, Xiaoxue
Zhang, Yifei
Chen, Yuantao
Zhao, Jian
Zhao, Hao
Computer Vision and Pattern Recognition
Recently, 3D Gaussian Splatting, as a novel 3D representation, has garnered attention for its fast rendering speed and high rendering quality. However, this comes with high memory consumption, e.g., a well-trained Gaussian field may utilize three million Gaussian primitives and over 700 MB of memory. We credit this high memory footprint to the lack of consideration for the relationship between primitives. In this paper, we propose a memory-efficient Gaussian field named SUNDAE with spectral pruning and neural compensation. On one hand, we construct a graph on the set of Gaussian primitives to model their relationship and design a spectral down-sampling module to prune out primitives while preserving desired signals. On the other hand, to compensate for the quality loss of pruning Gaussians, we exploit a lightweight neural network head to mix splatted features, which effectively compensates for quality losses while capturing the relationship between primitives in its weights. We demonstrate the performance of SUNDAE with extensive results. For example, SUNDAE can achieve 26.80 PSNR at 145 FPS using 104 MB memory while the vanilla Gaussian splatting algorithm achieves 25.60 PSNR at 160 FPS using 523 MB memory, on the Mip-NeRF360 dataset. Codes are publicly available at https://runyiyang.github.io/projects/SUNDAE/.
title Spectrally Pruned Gaussian Fields with Neural Compensation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2405.00676