Risk-optimized Outlier Removal for Robust 3D Point Cloud Classification

With the growth of 3D sensing technology, deep learning system for 3D point clouds has become increasingly important, especially in applications like autonomous vehicles where safety is a primary concern.