Abstract:Geometric priors are often used to enhance 3D reconstruction. With many smartphones featuring low-resolution depth sensors and the prevalence of off-the-shelf monocular geometry estimators, incorporating geometric priors as regularization signals has become common in 3D vision tasks. However, the accuracy of depth estimates from mobile devices is typically poor for highly detailed geometry, and monocular estimators often suffer from poor multi-view consistency and precision. In this work, we propose an approach for joint surface depth and normal refinement of Gaussian Splatting methods for accurate 3D reconstruction of indoor scenes. We develop supervision strategies that adaptively filters low-quality depth and normal estimates by comparing the consistency of the priors during optimization. We mitigate regularization in regions where prior estimates have high uncertainty or ambiguities. Our filtering strategy and optimization design demonstrate significant improvements in both mesh estimation and novel-view synthesis for both 3D and 2D Gaussian Splatting-based methods on challenging indoor room datasets. Furthermore, we explore the use of alternative meshing strategies for finer geometry extraction. We develop a scale-aware meshing strategy inspired by TSDF and octree-based isosurface extraction, which recovers finer details from Gaussian models compared to other commonly used open-source meshing tools. Our code is released in <a class="link-external link-https" href="https://xuqianren.github.io/ags_mesh_website/" rel="external noopener nofollow">this https URL</a>.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to solve the problem of insufficient geometric details and precision when using smartphones for indoor room 3D reconstruction. Specifically, the authors propose a method named AGS - Mesh, which adaptively integrates geometric prior information (geometric priors) into Gaussian Splatting to improve the accuracy and details of 3D reconstruction of indoor scenes. #### Main problems include: 1. **Limitations of low - resolution depth maps**: - The low - resolution depth sensors equipped on smartphones (such as the LiDAR sensor on iPhone) perform poorly when dealing with highly detailed geometric structures and edges. - These low - resolution depth maps produce inaccurate results when used for 3D mesh reconstruction in traditional methods (such as volumetric fusion). 2. **Poor multi - view consistency of monocular geometry estimators**: - Monocular geometry estimators perform poorly in terms of multi - view consistency, resulting in unstable quality of geometric prior information. 3. **Limitations of existing Gaussian Splatting methods**: - Existing Gaussian Splatting methods (such as 3DGS, 2DGS) do not extract fine - grained geometric details when dealing with large - scale indoor scenes, especially on surfaces with less texture and sparse outward - shooting data. ### Specific solutions proposed in the paper 1. **Adaptive Depth Regularization Strategy**: - Use Depth Normal Consistency (DNC) to filter noisy depth maps and ensure more accurate depth estimates on smooth surfaces. - Filter inaccurate depth estimates by comparing the normal vector consistency generated from noisy depth maps and pre - trained monocular networks. 2. **Adaptive Normal Regularization Strategy**: - Introduce Adaptive Normal Regularization (ANR) to dynamically adjust normal vector supervision during the optimization process and reduce the influence of uncertain areas. - Use fully pre - trained normal vectors for regularization at the beginning of training, and then rely on multi - view - consistent and more reliable filtered normal vectors. 3. **Improved Mesh Extraction Method**: - Propose an IsoOctree, an isosurface extraction method based on TSDF and octree, which can better preserve small geometric details. - Improve mesh quality by adjusting the truncation distance and combining with rendered normal maps. ### Summary The main goal of the paper is to improve the performance of the Gaussian Splatting method in indoor room 3D reconstruction by adaptively integrating geometric prior information from mobile devices and pre - trained monocular models, thereby achieving higher - quality geometric reconstruction and novel view synthesis.

AGS-Mesh: Adaptive Gaussian Splatting and Meshing with Geometric Priors for Indoor Room Reconstruction Using Smartphones

2DGS-Room: Seed-Guided 2D Gaussian Splatting with Geometric Constrains for High-Fidelity Indoor Scene Reconstruction

GaussianRoom: Improving 3D Gaussian Splatting with SDF Guidance and Monocular Cues for Indoor Scene Reconstruction

DN-Splatter: Depth and Normal Priors for Gaussian Splatting and Meshing

Mobile3DRecon: Real-time Monocular 3D Reconstruction on a Mobile Phone

Visual SLAM with 3D Gaussian Primitives and Depth Priors Enabling Novel View Synthesis

Integrating Meshes and 3D Gaussians for Indoor Scene Reconstruction with SAM Mask Guidance

Mobile3DScanner: an Online 3D Scanner for High-quality Object Reconstruction with a Mobile Device

GS2Mesh: Surface Reconstruction from Gaussian Splatting via Novel Stereo Views

SA-GS: Semantic-Aware Gaussian Splatting for Large Scene Reconstruction with Geometry Constrain

The Polygonal 3D Layout Reconstruction of an Indoor Environment via Voxel-Based Room Segmentation and Space Partition

Enhancement of 3D Gaussian Splatting using Raw Mesh for Photorealistic Recreation of Architectures

ES-Gaussian: Gaussian Splatting Mapping via Error Space-Based Gaussian Completion

MeshFormer: High-Quality Mesh Generation with 3D-Guided Reconstruction Model

StreetSurfGS: Scalable Urban Street Surface Reconstruction with Planar-based Gaussian Splatting

Space-time 2D Gaussian Splatting for Accurate Surface Reconstruction under Complex Dynamic Scenes

CityGaussianV2: Efficient and Geometrically Accurate Reconstruction for Large-Scale Scenes

LI-GS: Gaussian Splatting with LiDAR Incorporated for Accurate Large-Scale Reconstruction

MGSO: Monocular Real-time Photometric SLAM with Efficient 3D Gaussian Splatting

Indoor Scene Reconstruction with Fine-Grained Details Using Hybrid Representation and Normal Prior Enhancement

DyGASR: Dynamic Generalized Exponential Splatting with Surface Alignment for Accelerated 3D Mesh Reconstruction