Advancing Traffic Sign Recognition: Real-Time Detection and Optimization with YOLOv8

For intelligent transportation networks to effectively manage traffic and increase road safety, traffic sign analysis is essential. This study uses YOLOv8, the most recent version the Ultralytics' You Only Look Once (YOLO) algorithm, to develop and assess a traffic sign categorisation technique. YOLOv8, which is well-known for its immediate recognition, remarkable accuracy, & efficiency, has been trained on a number of traffic sign datasets with a focus on performance optimisation through rigorous testing. To improve the model's detection capabilities, hyperparameters like batch number, learning pace, dropout rate, and initialisation iterations were changed. Mean the average precision (mAP), to commonly used statistic for assessing object identification algorithms, was used to assess the model's performance. Furthermore, the impact of several optimisers, such as Adam, SGD, & Auto, on the accuracy and convergence of the model was examined. The outcomes validate YOLOv8's efficacy for real-time traffic sign identification by demonstrating its excellent accuracy and recall. Through a thorough comparison analysis, this research further assesses the significance of YOLOv8's architectural improvements over its predecessors, including YOLOv4 and YOLOv5. In order to confirm its viability for intelligent transportation systems, the study also investigates real-world deployment issues by evaluating computing efficiency on edge devices, such as the Raspberry Pi and Jetson Nano. The robustness of the model is further enhanced by the incorporation of various traffic situations, such as motion blur, occlusions, and nighttime photos. The model can make a substantial contribution to safer and more effective transportation networks by accurately identifying and localising traffic signs. This article discusses the experiment's outcomes, methods, and lessons learnt from refining the YOLOv8-based highway sign identification algorithm.