Year: 2024
Type: article
Abstract: This study evaluates the performance of the YOLOv5 model in the detection of traffic signs under a diverse range of environmental conditions, assessing its performance through a comprehensive set of experiments.This study assesses the model's precision in identifying signage categories across a variety of lighting conditions and perspectives by employing a robust dataset that includes 1,596 images of a wide range of traffic signs.The model's ability to maintain high detection accuracy in optimal conditions is the primary focus of the analysis, which also emphasizes the challenges encountered in adverse lighting conditions such as direct sunlight and low-light settings in parking lots.The results indicate that YOLOv5 is highly reliable in unobstructed and clear conditions, but its reliability decreases in complex environments.This paper examines potential enhancements and future research directions, such as exploring of alternative model architectures and the implementation of advanced data augmentation techniques, to improve the adaptability and robustness of traffic sign detection systems.
Authors: Nur Izzati Anaz Anizan, Fazlina Ahmat Ruslan, Noorfadzli Abd Razak, Mohd Azri Abdul Aziz, Juliana Johari
Institutions: Universiti Teknologi MARA System, Universiti Teknologi MARA
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
Funder: Universiti Teknologi MARA
OpenAlex ID: W4404464790