Long-Tail Learning for Rare Event Detection in Autonomous Vehicles

EasyChair Preprint 14039

Abstract

Autonomous vehicles must navigate a wide range of driving scenarios, including rare events such as adverse weather conditions and unusual road obstacles. Traditional deep learning models often struggle with these rare events due to the limited data available for training. This research explores advanced methods for long-tail learning to enhance the capability of deep learning models in identifying and responding to rare events on the road. By leveraging techniques such as data augmentation, transfer learning, and few-shot learning, this study aims to improve the performance and reliability of autonomous vehicles in handling uncommon yet critical situations. The research evaluates the effectiveness of these methods through simulation and real-world testing, highlighting the potential for long-tail learning to contribute to safer and more dependable autonomous driving systems.

Keyphrases: Transfer Learning, autonomous driving, long-tail learning

@booklet{EasyChair:14039,
  author    = {Dylan Stilinki and Kaledio Potter and Seyi Oladimeji},
  title     = {Long-Tail Learning for Rare Event Detection in Autonomous Vehicles},
  howpublished = {EasyChair Preprint 14039},
  year      = {EasyChair, 2024}}