An Improved Explainable Point Cloud Classifier (XPCC)

Classification of objects from 3-D point clouds has become an increasingly relevant task across many computer-vision applications. However, few studies have investigated explainable methods. In this article, a new prototype-based and classification method called eXplainable cloud classifier (XPCC) is proposed. The XPCC offers several advantages over previous nonexplainable First, the uses local densities global multivariate generative distributions. Therefore, provides comprehensive interpretable object-based classification. Furthermore, proposed built on recursive calculations, thus, computationally very efficient. Second, model learns continuously without need for complete retraining domain transferable. Third, expands underlying learning deep neural networks (xDNN), specific to 3-D. As such, following three layers are added original xDNN architecture: 1) feature extraction, 2) compound prototype weighting, 3) SoftMax function. Experiments were performed with ModelNet40 benchmark, which demonstrated that only one increase accuracy relative base algorithm when applied same problem. addition, article proposes novel visual representation model- explanations. superimposed create prototypical class their data density within space, cloud. They allow user visualize aspects identify object regions contribute in human-understandable way.