Reproduce “Learning Signed Hyper Surfaces for Oriented Point Cloud Normal 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
Qing Li, Huifang Feng, Kanle Shi, Yue Gao, Yi Fang, Yu-Shen Liu, Zhizhong Han
SHS-Net estimates and orients point-cloud surface normals in one network. It combines features from a local patch with global shape context, then predicts a normal and its sign using separate normal and sign losses. Across six PCPNet conditions, its average oriented RMSE is 19.79°, versus 31.07° for HSurf-Net+ODP; NGLO reports a lower 18.49° average.
Key insights
- Implicitly modeling signed hyper surfaces allows for simultaneous learning of normal direction and consistent orientation within a unified framework.
- Integrating both local and global geometric features is crucial for resolving the inherent ambiguity in normal orientation.
- Attention mechanisms and specific loss functions (e.g., sin loss for direction, cross-entropy for sign) effectively guide the network to produce accurate and globally consistent normals.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.