Reproduce “Any2Building: Unifying Point Clouds and Images for Building Reconstruction via Shared Spatial Queries” 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
Hongye Hou, Zhuoxiao Li, Liu Xiangyu, Tengxi Wang, Jing Ou, Dongli Wu, Shuai Zhang, Tongyan Hua and 1 more
Any2Building reconstructs compact LoD2 building models from point clouds, overhead images, or both. It encodes observations into shared queries on a 3D grid, which condition a structure generator. With 41 spatially cropped and subsampled points, adding a paired image raises final low-poly F@2% from 19.43% to 46.51% on 2,000 test buildings; this gain applies to that condition, not all inputs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.