Linked Dynamic Graph CNN: Learning on Point Cloud via Linking Hierarchical Features

TL;DR

This paper introduces LDGCNN, a novel graph CNN architecture that directly classifies and segments point clouds by linking hierarchical features, achieving state-of-the-art results on standard datasets.

Contribution

The paper proposes LDGCNN, removing the transformation network and linking hierarchical features from dynamic graphs to improve point cloud classification and segmentation.

Findings

Achieves state-of-the-art performance on ModelNet40.

Outperforms previous methods on ShapeNet.

Provides theoretical analysis and visualization of the network.

Abstract

Learning on point cloud is eagerly in demand because the point cloud is a common type of geometric data and can aid robots to understand environments robustly. However, the point cloud is sparse, unstructured, and unordered, which cannot be recognized accurately by a traditional convolutional neural network (CNN) nor a recurrent neural network (RNN). Fortunately, a graph convolutional neural network (Graph CNN) can process sparse and unordered data. Hence, we propose a linked dynamic graph CNN (LDGCNN) to classify and segment point cloud directly in this paper. We remove the transformation network, link hierarchical features from dynamic graphs, freeze feature extractor, and retrain the classifier to increase the performance of LDGCNN. We explain our network using theoretical analysis and visualization. Through experiments, we show that the proposed LDGCNN achieves state-of-art performance on two standard datasets: ModelNet40 and ShapeNet.