Comparison of Aggregation Functions for 3D Point Clouds Classification

The three-dimensional data is the core tool behind environment aware algorithms used in e.g. SLAM or autonomous driving. As a data format, point clouds are becoming increasingly popular, due to their high-resolution and mapping fidelity. However, representing data as points, rather than voxels, comes with very high processing complexity, as machine learning models need to deal with permutation-invariance within samples. The PointNet architecture provides an easy and efficient way to deal with the point cloud data, by performing feature extraction for each point separately and then computing feature-wise max function. In this work, we present a comparison of different permutation-invariant functions used for this aggregation evaluated on the ShapeNet dataset for the classification task.