DRGCNN: Dynamic region graph convolutional neural network for point clouds

作者：

Highlights：

• DRGConv allows each point can gather different regional features.

• DRGConv module can be integrated to other backbone networks.

• Information fusion from both local and global contexts on point-clouds.

• DRGConv module has translational invariance.

摘要

•DRGConv allows each point can gather different regional features.•DRGConv module can be integrated to other backbone networks.•Information fusion from both local and global contexts on point-clouds.•DRGConv module has translational invariance.

论文关键词：Graph convolution,Dynamic region selection,Feature extraction,Classification and segmentation,Point cloud

论文评审过程：Received 28 October 2021, Revised 1 April 2022, Accepted 27 May 2022, Available online 30 May 2022, Version of Record 3 June 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.117663