Development of an AI-Powered Waste Sorting System for Benghazi Using YOLOv8

Abstract

Benghazi faces limited challenges in waste management, primarily relying on traditional methods. This study aims to develop an intelligent system utilising the YOLOv8 (You Only Look Once) algorithm to enhance the efficiency and automation of waste sorting. The system was trained on a dataset collected from the Benghazi Waste Centre, initially comprising 1,000 images, which was later expanded to 3,000 images through data augmentation techniques such as rotation and brightness adjustments. The YOLOv8 algorithm was implemented using the Ultralytics library with optimized training configurations, including adjustments to the number of epochs, batch size, and learning rate, to achieve optimal performance. The system was tested in a simulated laboratory environment featuring a conveyor belt, a camera, and robotic arms for automated sorting. Initial simulation results showed an accuracy of 88%, which improved to 89.6% after data expansion and model retraining. These findings demonstrate that the system significantly enhances waste sorting, making it suitable for sustainable waste management strategies in Benghazi. Furthermore, the study highlights the promising potential of applying artificial intelligence to address real-world challenges and promote environmental sustainability.