For training, I used YOLO (You Only Look Once) models, which are models for object detection tasks, you can find the data here: https://huggingface.co/datasets/ayoubsa/Sign_Road_Detection_Dataset/tree/main

I got the best performance using yolov8n training it on 30 epochs with 49.8% accuracy on the leaderboard (which I think there is something wrong with it, it must be more higher than this).