A Deep Learning Approach for Traffic Sign Detection and Recognition Using Improved YOLOV5s

Abstract

Traffic sign detection is one of the most challenging tasks for autonomous vehicles, especially for the detection of different types of signs and real-time applications. Unmanned driving systems face a lot of problems to recognize traffic signs faster and more accurately. In this paper, we propose a model using improved YOLOv5 to detect Traffic signs and recognize them properly. Our dataset consists of 3500 pictures of the traffic sign and we have annotated all that pictures in YOLOv5 format. There is 39 classification of our dataset on all pictures based on the traffic sign. This system can be used for unmanned driving vehicles. Using this model a device can make which will help drivers who are driving a car. After the implementation of our dataset with the help of improved YOLOv5, the output shows an accuracy of 86.75% in different conditions such as low light, cloudy, rainy, and sunny.