Survey and Performance Analysis on Point Cloud Classification Models

Point clouds are a popular representation for 3D data, widely used in applications such as autonomous driving, robotics, and computer graphics. Classifying objects within point clouds poses unique challenges due to the unordered nature of the data and the lack of spatial structure. In this paper, we propose a survey and performance analysis of SOTA models on point cloud data classification. PointNet leverages a novel neural network architecture that directly consumes raw point clouds, eliminating the need for pre-processing steps like voxelization or rendering. Our work is to better analyse the mnodel's performance in with different dataset such as ModelNet and compare them. PointNet was the first ground-breaking approach to handle the classification problem which is still giving competetive performance qulality.