Reproduce “OmniRay: Camera-Agnostic Monocular 3D Reconstruction with Composite Ray Fields” 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
Namitha Guruprasad, Abhay Yadav, Cheng Peng, Rama Chellappa, Anand Bhattad
OmniRay reconstructs camera-frame point clouds from single images without camera metadata by predicting each pixel’s viewing direction and distance along it, up to scale. Its ray field starts with a radial pattern, then adds spherical-harmonic and local corrections. On a held-out catadioptric configuration, it reports 1.120° mean ray error and 0.0131 relative Chamfer distance, but does not lead on every dataset.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.