Reproduce “LEGAU: Learning Semantic Gaussian Priors for Scalable Category-level Pose Estimation” 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

Hongli Xu, Zhaowei Lu, Junwen Huang, Jiaqi Hu, Peter KT Yu, Benjamin Busam, Federico Tombari, Slobodan ilic

LEGAU jointly learns category-conditioned canonical shape representations, NOCS correspondences, and pose from one RGB-D view, using a Semantic Gaussian Field to let global structure guide local alignment. On synthetic SOPE, it scores 44.2 versus GenPose++’s 40.2 on VUS@10°5cm. On real ROPE, however, it trails GenPose++ at that pose metric, scoring 24.4 versus 29.4.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.