Semantic Segmentation

This project is about the implementation of Semantic segmentation of LIDAR point cloud data. Since a lot of self-driving cars rely mostly rely on the LIDAR point clouds, it is essential to differentiate the pseudo-dynamic obstacles such as cars, to better simultaneous localization and mapping. In this work, I initially present the data that I am using for segmentation, followed by the Iterative closest point implementation for mapping the LIDAR , and then followed by the segmentation using neural networks.

Report for this project can be downloaded here.

Instructor: Nitin Sanket