Road accidents pose a significant global threat, exacerbated by diverse road conditions and inadequate traffic management systems in Bangladesh. This research addresses these issues by improving traffic sign detection and identification technology, which is critical for ITS and ADAS. This work emphasizes the critical role of accurate road sign identification in improving roadway security and traffic management, and it makes use of advanced models, namely the YOLO models. A major contribution of this work is the development of the Bangladeshi Road Sign Detection Dataset (BRSDD), comprising 9,133 annotated images spanning 29 distinct classes of Bangladeshi road signs. This dataset fills a critical gap, providing a comprehensive resource to train and evaluate detection models tailored specifically to Bangladesh. Evaluation of the YOLOv8 model on the BRSDD demonstrates significant performance with a precision of 0.986, with a recall of 0.999, surpassing earlier versions, YOLOv5 and YOLOv7, in mean average precision. Integrating YOLOv8 with ADAS enhances real-time road sign recognition, supporting improved road safety and traffic management practices. This research lays the groundwork for integrating advanced computer vision into ITS, customized to local road conditions, to effectively advance traffic safety and management strategies. In conclusion, this particular research work advances road sign detection technology and offers practical solutions to enhance road safety, traffic management, and overall transportation efficiency in Bangladesh.
A YOLO-Based Framework for Road Sign Detection and Recognition in the Context of Bangladesh
2024-09-25
1754476 byte
Conference paper
Electronic Resource
English