Reproduce “Holi-Spatial 360: Generative Reconstruction from Sparse Panoramas for Spatial Rendering and Intelligence” 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
Yuanyuan Gao, Zhirui Liu, Yuning Gong, Yifei Liu, Yifan Zhan, Kuixuan Jiao, Xiaolong Zhou, Yifan Yang and 4 more
Holi-Spatial 360 uses sparse real-estate panoramas to build 3D scenes and spatial supervision. It anchors reconstruction with cross-view correspondences, then uses image and video models to restore intermediate views as appearance supervision. On RealSee scenes, F-score at a 2 cm threshold rises from 0.1991 to 0.3529; fine-tuning Qwen3.5-4B raises Holi360 benchmark classification macro accuracy from 46.77% to 85.25%.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.