AUTONOMOUS ROAD SIGN DETECTION USING DEEP LEARNING MODELS

Abstract

Detection of the object is an important technique for tracing and classifying objects in a picture or video. It is attained by sketching bounding boxes around the objects and allocating them consistent category labels. Earlier approaches, such as image filtering and training models on unambiguous weather situations, have proven insufficient. To overcome these problems, we propose a Deep Learning Model using YOLOv11 for robust traffic sign detection and identification. YOLOv11 suggests improvements in real-time object detection, surrounding tasks like instance segmentation, classification, and oriented object detection. The proposed model comprises an IoT input unit for image capture, a detection unit utilizing YOLOv11 for feature extraction and processing, and a voice activity detection unit to provide drivers with auditory alerts regarding detected signs. This integrated system aims to increase driving safety and efficiency by permitting vigorous and consistent traffic sign detection in real-world driving scenarios. The system is very efficient, achieving a weighted F1-score of 98.29%, weighted precision of 98.93%, weighted recall of 97.69%, weighted validation accuracy of 95.97% and macro validation accuracy of 96.17%.