Deep Learning Neural Networks for 3D Point Clouds Shape Classification: A Survey

Point clouds data acquisition is increasingly important over these years because of its wide applications such as autonomous driving, robotics, virtual reality, and medical treatment. Deep learning neural networks are commonly used to process 3D point clouds for tasks such as shape classification nowadays. It can be generally classified into four main categories, i.e convolution-based method, point-wise MLP method, graph-based method, and hierarchical Data Structure-based methods. This paper demonstrates a comprehensive review of these latest state-of-the-art 3D point clouds classification methods. It also presents a comparative study on the advantages and limitations of these point clouds classification methods.