Robust Traffic Sign Recognition Using CNN YOLOv5 Model

Traffic Sign Recognition System is important for communicating Sign information to drivers. To make driving safer and simpler, traffic signs offer drivers a wealth of useful information about the road. Throughout the past few decades, traffic sign detection has received significant research interest. This kind of mechanism is crucial to the advancement of driverless vehicles. It’s been a challenging task for self-driving cars to understand various traffic signs. Real-time and accurate detections are the primary requirements of a robust traffic sign detection system. We have implemented a Robust Traffic Sign Recognition System that is capable of assisting drivers. This system detects the traffic sign using a trained YOLOv5 Convolutional Neural Network. After detecting the specific traffic sign, it is recognized and displayed. Collected images are Pre-processed and Gray scaled. The image format is converted to YOLO format. The images are then trained using CNN Yolov5 algorithm in Python. The model is trained by setting the necessary parameters and constraints. Later, using the object-detection library, code is executed and the end result displays the respective name of the traffic signs detected with a good accuracy score. We used a dataset of 4 classes with 877 images each to train the neural network. The results showed that the recognition accuracy ranged between 90% to 97% using YOLOv5. The large area under the Precision-Recall curve indicates high performance, with a mean average precision of 0.989. The confusion matrix shows high true positive and true negative rates.