Reproduce “GATOR: Generative and Agentic 3D Object Reconstruction From Casual Images” 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
Qirui Wu, Stan Birchfield, Hesam Rabeti, Angel X. Chang, Bowen Wen
GATOR reconstructs complete, scene-placed 3D objects from sparse images, then uses an image-grounded agent to make local repairs. It mixes aligned RGB, masks, and depth-derived pointmaps before generating structure, geometry, and appearance with text guidance. On ScanNet++, across 229 targets in 49 scenes, it achieves mean ADD-SB 0.041; performance depends on estimated masks, cameras, and depth.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.