BriefGPT.xyz
Apr, 2019
深度迭代表面法线估计
Differentiable Iterative Surface Normal Estimation
HTML
PDF
Jan Eric Lenssen, Christian Osendorfer, Jonathan Masci
TL;DR
该研究提出了一种基于图神经网络和自适应各向异性核的表面法线估计算法，不需要任何手动特征或预处理，优于现有深度学习方法，在保留尖锐特征和空间等变性的同时，速度和参数效率均为同类算法的两个数量级。
Abstract
This paper presents an end-to-end differentiable algorithm for anisotropic
surface normal estimation
on unstructured point-clouds. We utilize
graph neural networks
to iteratively infer point weights for a plane f
→