Reproduce “Martine: Benchmarking Multi-View 3D Surface Reconstruction Across Viewpoint Coverage, Resolution, and Lighting” 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
Robin Bruneau, Baptiste Brument, Frederic Giraud, Bastian Sigrist, Sascha Jecklin, Philipp Fürnstahl, Bjoern Menze, Lilian Calvet
Martine is a benchmark for measuring 3D reconstruction accuracy from photographs. It captures objects from 84 viewpoints under controlled single-light and multi-light conditions, and checks alignment between images and reference scans. Of 28 objects, 26 have scans for evaluation. At 0.5 millimeter and quarter resolution, NeuS2 scores 75.5% versus RNb-NeuS2 at 66.7%; Neuralangelo ranks highest in this setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.