1. The goal of this project is to employ deep learning technologies to build advanced techniques for the semantic segmentation and noise removal of point clouds. Convolutional neural networks (CNNs), which are intended to process unordered and irregular point cloud data efficiently and capture both local and global information, are used in the suggested approach. We provide an efficiency architecture that emphasizes the incorporation of contextual information to improve the accuracy of segmentation in intricate scenarios. An extensive framework for validation is constructed in order to guarantee the universality and dependability of the generated model. This paradigm includes strong measures, such as the intersection over union (IoU), recall, accuracy, and precision, or assessing segmentation performance for advanced driver assistance systems.

2. Physiable implementation and validation in Window+Linux platform and embedded system.

3. Hardware-Software co-integration framework for a Lidar system