Yolo V5 for Traffic Sign Recognition and Detection Using Transfer Learning

With the advancement in the field of autonomous cars, we are coming closer to reliable integration. However, in order for an autonomous vehicle to function in an urban environment, it has to abide by traffic rules. In this paper, we design a vision system based on our trained YOLO v5 models for both classification on the GTSRB dataset and detection on the GTSDB dataset using transfer learning from the classification to the detection model to optimise results. Our choice of the YOLO v5 algorithm is justified by its capability to combine accuracy and speed simultaneously, making it suitable for real-time applications.