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

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.

References in corpus (8)

• Graph Neural Networks: A Review of Methods and Applications
• Dynamic Graph CNN for Learning on Point Clouds
• PointCNN: Convolution On $\mathcal{X}$-Transformed Points
• U-Net: Convolutional Networks for Biomedical Image Segmentation
• SpiderCNN: Deep Learning on Point Sets with Parameterized Convolutional Filters
• Sensor Fusion for Predictive Control of Human-Prosthesis-Environment Dynamics in Assistive Walking: A Survey
• Modeling Local Geometric Structure of 3D Point Clouds using Geo-CNN
• Directional PointNet: 3D Environmental Classification for Wearable Robotics

Cited by in corpus (23)

• Comprehensive Review of Deep Learning-Based 3D Point Cloud Completion Processing and Analysis
• Deep Learning-based 3D Point Cloud Classification: A Systematic Survey and Outlook
• Deep Learning for 3D Point Clouds: A Survey
• Attribute Artifacts Removal for Geometry-based Point Cloud Compression
• DeltaConv: Anisotropic Operators for Geometric Deep Learning on Point Clouds
• Sequential Decision Fusion for Environmental Classification in Assistive Walking
• Triangle-Net: Towards Robustness in Point Cloud Learning
• GRNet: Gridding Residual Network for Dense Point Cloud Completion
• Pointwise Rotation-Invariant Network with Adaptive Sampling and 3D Spherical Voxel Convolution
• DRINet++: Efficient Voxel-as-point Point Cloud Segmentation
• DiffGCN: Graph Convolutional Networks via Differential Operators and Algebraic Multigrid Pooling
• Residual Attention Net for Superior Cross-Domain Time Sequence Modeling
• Graph Signal Processing for Geometric Data and Beyond: Theory and Applications
• Grid-GCN for Fast and Scalable Point Cloud Learning
• Revisiting 2D Convolutional Neural Networks for Graph-based Applications
• DRINet: A Dual-Representation Iterative Learning Network for Point Cloud Segmentation
• PRIN/SPRIN: On Extracting Point-wise Rotation Invariant Features
• PatchFormer: An Efficient Point Transformer with Patch Attention
• Mesorasi: Architecture Support for Point Cloud Analytics via Delayed-Aggregation
• Multi-scale Receptive Fields Graph Attention Network for Point Cloud Classification
• A Convolutional Architecture for 3D Model Embedding
• Generating Unrestricted 3D Adversarial Point Clouds
• MARNet: Multi-Abstraction Refinement Network for 3D Point Cloud Analysis