The paper proposes a novel approach for surface reconstruction from Gaussian splatting models. Instead of relying on the locations of the Gaussian elements as a prior for surface reconstruction, the method leverages the superior novel-view synthesis capabilities of 3DGS.
The pipeline consists of:
The proposed framework also allows for the reconstruction of a specific object in the scene by segmenting the object using a combination of Segment-Anything (SAM) masks and depth map information.
The method reduces surface reconstruction time dramatically, taking only a small overhead on top of the 3DGS capturing of the scene. It was tested on the Tanks and Temples benchmark, where it surpassed the current state-of-the-art method for surface reconstruction from Gaussian splatting models. Additionally, extensive testing on in-the-wild scenes captured with a smartphone showed the method's superior reconstruction abilities compared to previous approaches.
To Another Language
from source content
arxiv.org
Önemli Bilgiler Şuradan Elde Edildi
by Yaniv Wolf,A... : arxiv.org 04-03-2024
https://arxiv.org/pdf/2404.01810.pdfDaha Derin Sorular