A Brief Overview on 3D Perception using Deep Neural Networks for Automotive

Technologies in automotive industry have been improved very much in the last few years. Since technology enhances our daily activity, many automotive companies are looking for implementing different features for cars to make our life easier. One of the most advanced capabilities is represented by autonomous driving. This task typically implies intensive 3D perception of the environment. Perception involves the collection of the data coming from sensors which are considered to be relevant for learning the environment. The aim of the current paper is to review the point cloud-based 3D object perception using deep neural networks works with application in automotive. The study refers to the whole problematic associated with this challenging task, including the necessary hardware for acquiring the 3D data, the publicly available point cloud datasets and the deep neural network models which are able to process point cloud data.