Reproduce “GaussianDreamerPro: Text to Manipulable 3D Gaussians with Highly Enhanced Quality” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Taoran Yi, Jiemin Fang, Zanwei Zhou, Junjie Wang, Guanjun Wu, Lingxi Xie, Xiaopeng Zhang, Wenyu Liu and 2 more
GaussianDreamerPro refines text-to-3D Gaussian Splatting by employing a geometry-guided approach, binding 3D Gaussians to a mesh surface for enhanced detail and clarity. This method produces manipulable 3D assets with improved surface quality, achieving a 56% user preference in a comparative study.
Key insights
- Binding 3D Gaussians to a mesh surface constrains their growth, preventing scattering and enabling more stable optimization for clearer surfaces and enriched details.
- The geometry guidance mechanism allows for the seamless manipulation of generated 3D assets, as the Gaussians follow transformations applied to the underlying mesh.
- A two-stage progressive refinement process, starting with coarse geometry and enhancing details with geometry-bound Gaussians, improves overall quality.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.