2DGS-Room: Seed-Guided 2D Gaussian Splatting with Geometric Constrains for High-Fidelity Indoor Scene Reconstruction
Wanting Zhang, Haodong Xiang, Zhichao Liao, Xiansong Lai, Xinghui Li, Long Zeng
2024-12-09
Summary
This paper talks about 2DGS-Room, a new method for accurately reconstructing indoor scenes in 3D using advanced techniques that improve how we represent and visualize spaces from images.
What's the problem?
Reconstructing indoor scenes is difficult because they often have complex shapes and areas without clear textures. Traditional methods struggle to create accurate 3D models from images, especially in these challenging conditions, leading to incomplete or low-quality results.
What's the solution?
The authors introduce 2DGS-Room, which uses a technique called 2D Gaussian Splatting to create detailed indoor scenes. They use a 'seed-guided' approach that helps control how the 2D Gaussians (which represent parts of the scene) are distributed. This method optimizes the placement and density of these seed points to enhance geometric accuracy. Additionally, they incorporate depth information and surface normals to improve clarity in areas with details and textures. They also apply multi-view consistency constraints to ensure that images from different angles look coherent and realistic.
Why it matters?
This research is important because it advances the field of indoor scene reconstruction, making it easier to create high-quality 3D models from images. This has applications in virtual reality, architecture, and gaming, where accurate representations of indoor spaces are crucial for creating immersive experiences.
Abstract
The reconstruction of indoor scenes remains challenging due to the inherent complexity of spatial structures and the prevalence of textureless regions. Recent advancements in 3D Gaussian Splatting have improved novel view synthesis with accelerated processing but have yet to deliver comparable performance in surface reconstruction. In this paper, we introduce 2DGS-Room, a novel method leveraging 2D Gaussian Splatting for high-fidelity indoor scene reconstruction. Specifically, we employ a seed-guided mechanism to control the distribution of 2D Gaussians, with the density of seed points dynamically optimized through adaptive growth and pruning mechanisms. To further improve geometric accuracy, we incorporate monocular depth and normal priors to provide constraints for details and textureless regions respectively. Additionally, multi-view consistency constraints are employed to mitigate artifacts and further enhance reconstruction quality. Extensive experiments on ScanNet and ScanNet++ datasets demonstrate that our method achieves state-of-the-art performance in indoor scene reconstruction.