Reproduce “MOMASPLAT: Dynamic Novel-View Demonstration Augmentation for Mobile Manipulation” 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
Sateesh Kumar, Arnav Balaji, Arpit Bahety, Skand Peri, Georgios Pavlakos, Roberto Martín-Martín
MOMASPLAT generates synthetic mobile-manipulation demonstrations for base poses absent from the source data. It reconstructs the mostly static scene and moving object separately, renders them from sampled viewpoints, and transforms actions into the new base frame. Across five real-world tasks, it reports 37% average success at unseen poses versus 17% for One-Thousand Trajectories; performance is not guaranteed for every pose.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.