Reproduce “Evaluating Alternatives to SFM Point Cloud Initialization for Gaussian Splatting” 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
Yalda Foroutan, Daniel Rebain, Kwang Moo Yi, Andrea Tagliasacchi
The method initializes Gaussian Splatting with points sampled from a briefly trained NeRF, then can add NeRF-estimated depth guidance during training. On OMMO, NerfAcc initialization with depth guidance reaches 29.85 average test PSNR, versus 29.22 without guidance and 29.29 with COLMAP initialization. This supports depth guidance in this setup, not universal replacement of SfM or accurate camera poses.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.