本文介绍了一种基于PointNet的3D重建方法，重点在于N-tuple loss的应用，这是一种许多到许多的损失函数。文章提出了点对点特征(PPF)，能够快速表示三维对象，为3D重建提供高效解决方案。
It's based on PointNet but the important thing is the N-tuple loss, which is a many-to-many loss. The basic idea is still the reproject error sum. So the authors call them point pair features (PPF). This gives a fast representation of objects for 3D reconstruction. Minors thing such as CGF, 3D Match are standard in the community.
扫一扫
评论 3
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
查看更多评论
添加红包
