Real-time Hazard Detection in Autonomous Vehicles: Deep Learning Approaches for Enhancing Road Safety

Abstract

This examination explores ongoing risk location in independent vehicles through the use of profound learning philosophies, expecting to upgrade street security. Utilizing Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), the review centers around distinguishing and grouping dangers like people on foot, vehicles, obstructions, and street abnormalities in assorted driving situations. By examining information from locally available sensors including cameras, LiDAR, and radar, the profound learning models learn perplexing examples and portrayals, empowering them to settle on informed choices in powerful conditions. The trial results exhibit the viability of the proposed approaches, with CNN-based and RNN-based models accomplishing high exactness, accuracy, review, and F1-score upsides of 0.95, 0.93, 0.96, and 0.94 individually. These outcomes outperform those of conventional AI calculations and pattern techniques, displaying the predominance of profound learning in danger identification undertakings. Also, examinations with related work highlight the headways accomplished, featuring the capability of profoundly figuring out how to upgrade the well-being abilities of independent vehicles. In general, this examination adds to the continuous endeavors in creating solid and effective danger recognition frameworks, making ready for more secure and more dependable independent driving innovation.