Reproduce “OV-Foam: How 3D Representations Shape Semantic Feature Lifting” 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
Yousef Rajeh, Alejandra Esparza Pelcastre, Ivan Viola, Dominik Engel
OV-Foam partitions space with bounded power cells and lifts image features using rendering-weighted one-shot averaging, reducing contributor overlap. In ten ScanNet scenes initialized from ground-truth geometry, it averaged 2.32 contributors per ray, versus 33.63 for 3DGS. With ground-truth class embeddings, mIoU was 90.59% versus 85.40%; evidence is limited to this controlled setup.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.