Abstract

Deep learning with 3D data such as reconstructed point clouds and CAD modelshas received great research interests recently. However, the capability ofusing point clouds with convolutional neural network has been so far not fullyexplored. In this technical report, we present a convolutional neural networkfor semantic segmentation and object recognition with 3D point clouds. At thecore of our network is point-wise convolution, a convolution operator that canbe applied at each point of a point cloud. Our fully convolutional networkdesign, while being simple to implement, can yield competitive accuracy in bothsemantic segmentation and object recognition task.

 

Quick Read (beta)