PointNet++解析

返回总结列表  https://blog.csdn.net/albert199199/article/details/103531486

PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space

Key points:

• local+global

• Skip connection

• Sampling layer: iterative farthest point sampling (FPS)

• Grouping layer: input~~N*(d+C)+N'*d, output~~N'*K*(d+C), where K neighborhood.

• PointNet layer: input~~N'*K*(d+C), output~~N'*(d+C')

• Non-Uniform Sampling Density: Multi-Scale Grouping (MSG)\Multi-resolution Grouping (MRG)

• code