Depth-Regularized Optimization for 3D Gaussian Splatting in Few-Shot Images

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Chung, Jaeyoung, Oh, Jeongtaek, Lee, Kyoung Mu
Natura: Preprint
Pubblicazione: 2023
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913185142931456
author Chung, Jaeyoung
Oh, Jeongtaek
Lee, Kyoung Mu
author_facet Chung, Jaeyoung
Oh, Jeongtaek
Lee, Kyoung Mu
contents In this paper, we present a method to optimize Gaussian splatting with a limited number of images while avoiding overfitting. Representing a 3D scene by combining numerous Gaussian splats has yielded outstanding visual quality. However, it tends to overfit the training views when only a small number of images are available. To address this issue, we introduce a dense depth map as a geometry guide to mitigate overfitting. We obtained the depth map using a pre-trained monocular depth estimation model and aligning the scale and offset using sparse COLMAP feature points. The adjusted depth aids in the color-based optimization of 3D Gaussian splatting, mitigating floating artifacts, and ensuring adherence to geometric constraints. We verify the proposed method on the NeRF-LLFF dataset with varying numbers of few images. Our approach demonstrates robust geometry compared to the original method that relies solely on images. Project page: robot0321.github.io/DepthRegGS
format Preprint
id arxiv_https___arxiv_org_abs_2311_13398
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Depth-Regularized Optimization for 3D Gaussian Splatting in Few-Shot Images
Chung, Jaeyoung
Oh, Jeongtaek
Lee, Kyoung Mu
Computer Vision and Pattern Recognition
Graphics
In this paper, we present a method to optimize Gaussian splatting with a limited number of images while avoiding overfitting. Representing a 3D scene by combining numerous Gaussian splats has yielded outstanding visual quality. However, it tends to overfit the training views when only a small number of images are available. To address this issue, we introduce a dense depth map as a geometry guide to mitigate overfitting. We obtained the depth map using a pre-trained monocular depth estimation model and aligning the scale and offset using sparse COLMAP feature points. The adjusted depth aids in the color-based optimization of 3D Gaussian splatting, mitigating floating artifacts, and ensuring adherence to geometric constraints. We verify the proposed method on the NeRF-LLFF dataset with varying numbers of few images. Our approach demonstrates robust geometry compared to the original method that relies solely on images. Project page: robot0321.github.io/DepthRegGS
title Depth-Regularized Optimization for 3D Gaussian Splatting in Few-Shot Images
topic Computer Vision and Pattern Recognition
Graphics
url https://arxiv.org/abs/2311.13398