Enhancing point cloud resolution for autonomous driving with deep learning AI models

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In the fast-evolving world of Lidar technology, our study tackles the growing need for top-quality Lidar data. This demand spans various uses, like self-driving cars, environmental tracking, and robot awareness. Elevating point cloud resolution, while vital for robust environmental perception, often entails increased costs due to the integration of additional lasers. This predicament is especially pronounced in self-driving vehicles, where cost-effectiveness is paramount. Our research endeavors to democratize high-resolution Lidar data by leveraging custom trained deep learning AI models (based on PU-Net and PU-GCN) to tackle the pivotal issue of point cloud upsampling. Our mission is to make high-resolution Lidar data a cost-effective solution for applications demanding acute environmental perception. Embracing a diverse set of state-of-the-art techniques, including Chamfer Distance and Earth Mover’s Distance, tailored to the probabilistic nature of point clouds, we meticulously evaluate point cloud dissimilarity. Beyond the technical intricacies, our work resonates with the broader goal of enhancing the resolution of Lidar data, thereby contributing to the precision and safety of autonomous systems.
Original languageEnglish
Title of host publication2024 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Workshops)
PublisherIEEE
ISBN (Electronic)9798350304367
ISBN (Print)9798350304374
DOIs
Publication statusPublished - 23 Apr 2024

Publication series

NameIEEE Conference Proceedings
PublisherIEEE
ISSN (Print)2836-5348
ISSN (Electronic)2766-8576

Fingerprint

Dive into the research topics of 'Enhancing point cloud resolution for autonomous driving with deep learning AI models'. Together they form a unique fingerprint.

Cite this