Mining Point Cloud Local Structures by Kernel Correlation and Graph Pooling

Abstract

Unlike on images, semantic learning on 3D point clouds using a deep network is challenging due to the naturally unordered data structure. Among existing works, PointNet has achieved promising results by directly learning on point sets. However, it does not take full advantage of a point's local neighborhood that contains fine-grained structural information which turns out to be helpful towards better semantic learning. In this regard, we present two new operations to improve PointNet with a more efficient exploitation of local structures. The first one focuses on local 3D geometric structures. In analogy to a convolution kernel for images, we define a point-set kernel as a set of learnable 3D points that jointly respond to a set of neighboring data points according to their geometric affinities measured by kernel correlation, adapted from a similar technique for point cloud registration. The second one exploits local high-dimensional feature structures by recursive feature aggregation on a nearest-neighbor-graph computed from 3D positions. Experiments show that our network can efficiently capture local information and robustly achieve better performances on major datasets. Our code is available at http://www.merl.com/research/license#KCNet

Accepted in CVPR'18. *indicates equal contribution

References in corpus (7)

• Semi-Supervised Classification with Graph Convolutional Networks
• Caffe: Convolutional Architecture for Fast Feature Embedding
• Deep Convolutional Networks on Graph-Structured Data
• Generative and Discriminative Voxel Modeling with Convolutional Neural Networks
• 3D ShapeNets: A Deep Representation for Volumetric Shapes
• Graph Based Convolutional Neural Network
• FoldingNet: Point Cloud Auto-encoder via Deep Grid Deformation

Cited by in corpus (9)

• Dynamic Graph CNN for Learning on Point Clouds
• Patch-based Progressive 3D Point Set Upsampling
• PVNet: A Joint Convolutional Network of Point Cloud and Multi-View for 3D Shape Recognition
• PointGrow: Autoregressively Learned Point Cloud Generation with Self-Attention
• PointNetLK: Robust & Efficient Point Cloud Registration using PointNet
• Inferring Point Clouds from Single Monocular Images by Depth Intermediation
• PointSmile: Point Self-supervised Learning via Curriculum Mutual Information
• MeshNet: Mesh Neural Network for 3D Shape Representation
• A Deeper Look at 3D Shape Classifiers