An Automated Effective Waste Classification and Sorting Using YOLOv8

Increasing amounts of solid waste in developing nations such as Bangladesh require increasingly intelligent, sustainable solutions. Manual sorting is time-consuming and dangerous, particularly in the scenario of urbanization and heterogeneous waste. This article presents an end-to-end deep learning and IoT-enabled smart waste management system for waste classification, sorting, and monitoring in real-time. YOLOv8 object detection model was trained on the custom dataset of 900+ labeled images capturing urban, semi-urban, and rural environments with seven wastes: cable, e-waste, glass, medical waste, metal, plastics, and cans. The model was trained for 100 epochs and with enhanced augmentation attained $\text{mAP} {@} 0.5$ of 93.41% and an F1-score of 89.65%. The system is implemented in a smart bin prototype with a camera, sensors, and GSM/Wi-Fi modules for providing automatic detection, sorting, and cloud monitoring. The end-to-end, deployable solution is better than current solutions because it can handle real-world variation as well as autonomous waste management in low-resource environments.