The objective of this project is to enhance the detection precision of point cloud panorama segmentation algorithms by introducing a novel network architecture that leverages polar coordinate bird's-eye-view projection. Designed to excel even with limited datasets, this network substantially boosts model performance, fulfilling the criteria for both panorama segmentation and real-time performance in autonomous driving applications. It achieves an impressive mean Intersection over Union (mIOU) of 59.40% and a Frame Per Second (FPS) rate of 14.
