One popular model that is used currently to solve these tasks is the PointNet, which trained a permutation invariant model on a set of points. However, a set of points could also be seen as a graph.

In this paper, we propose 3D graph convolution networks (3D-GCN), which uniquely learns 3D kernels with graph max-pooling mechanisms for extracting geometric features from point cloud data across ...

Graph Neural Networks have been recently applied to 3D object detection in point clouds. The works, however, have the problem of insufficient detection accuracy for small objects and objects in ...