DESIGN AND DEVELOPMENT OF A REAL-TIME TRAFFIC SIGN DETECTION AND RECOGNITION FRAMEWORK USING YOLOv8 ON EDGE AI
Pratiksha Madhukar Jadhav, Nilesh R. Wankhade, Sahebrao B Bagal, Ashwini V Sonawane, Pragati K Ghayvate, Vishal Ambhore
Abstract: The real-time traffic sign recognition and detection are essential elements of intelligent transportation systems and advanced driver assistance systems, especially in resource-limited edge systems. Traditional vision-based methods usually have a high level of computational complexity, they have latency problems, and they are less accurate in changing illumination, occlusion, and motion blur situations. The paper discusses the design and development of real-time traffic sign detection and recognition framework on the basis of YOLOv8 that will be optimized to work with edge AI devices. The suggested system combines light-weight backbone architecture, efficient feature pyramid network, and anchor-free detection mechanism to be highly accurate in detection and low inference latency. The optimization methods that are used to guarantee robustness and computational efficiency are model optimization, such as transfer learning, data augmentation, pruning and quantization. The structure is implemented on an edge computing platform to test the performance in the real world on dynamic traffic condition. Experimental evidence indicates that the suggested system has a high mean Average Precision (mAP), better F1-score, and lower inference time as compared to traditional CNN-based detectors. The model has a consistent performance with varying weather conditions and lighting effects with limited memory and power constraints. The results affirm that the optimized YOLOv8-based framework can be used to perform precise and low-latency traffic sign detection that can be utilized in autonomous driving and intelligent transportation systems. The research finds out that deep learning architectures with edges optimization can substantially improve real-time traffic monitoring performance and guarantee scalability, reliability, and energy efficiency in smart vehicular systems.
Keywords: Traffic Sign Detection; YOLOv8; Edge AI; Raspberry Pi; Intelligent Transportation Systems (ITS)
DOI: 10.24874/PES08.03.022
Recieved: 05.05.2026 Revised: 24.06.2026 Accepted: 21.07.2026
UDC:
Reads: 15 
