BriefGPT.xyz
Jul, 2018
3DFeat-Net:基于弱监督的点云局部3D特征匹配
3DFeat-Net: Weakly Supervised Local 3D Features for Point Cloud Registration
HTML
PDF
Zi Jian Yew, Gim Hee Lee
TL;DR
本文提出 3DFeat-Net 算法，利用弱监督学习 3D 特征检测器和描述符，通过对齐和注意机制学习 GPS/INS 标记的 3D 点云的特征对应关系，无需显式指定匹配点群，实验表明其在室外重力对齐数据集上取得了最优表现。
Abstract
In this paper, we propose the
3dfeat-net
which learns both 3D feature detector and descriptor for
point cloud matching
using
weak supervision
→