Reproduce “ARAP-GS: Drag-driven As-Rigid-As-Possible 3D Gaussian Splatting Editing with Diffusion Prior” 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
Xiao Han, Runze Tian, Yifei Tong, Fenggen Yu, Dingyao Liu, Yan Zhang
ARAP-GS introduces a method for intuitively editing 3D Gaussian Splatting scenes through drag operations while preserving geometric structure and visual quality. This approach combines As-Rigid-As-Possible (ARAP) deformation with diffusion model-based visual enhancement, enabling coherent changes across various scene types.
Key insights
- ARAP principles can be effectively adapted to 3D Gaussian Splatting by deforming a representative subset of Gaussians and interpolating changes to the entire scene.
- Combining geometric deformation (ARAP) with a diffusion prior (e.g., StableSR) significantly improves the visual quality and helps remove artifacts from edited 3DGS scenes.
- A mask-guided refinement and iterative optimization strategy against diffusion-enhanced images enables maintaining multi-view consistency during fine-tuning.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.