Intelligent Solid Waste Management: A Smart Approach for Solid Waste Identification and Segregation

Abstract

Identifying solid waste is an essential component in achieving a sustainable community. In the case of Northwestern Mindanao State College of Science and Technology, it still relies on hand-sorting and open-burning solid waste methods for waste management despite the growing population.  This might result in negative effects on the environment and health among workers, employees, and students. Hence, this paper presents a smart approach to identifying solid waste types by utilizing the real-time object detection and image segmentation model, the YOLOv8. In this approach, the researchers gathered and used the solid waste datasets from Kaggle, GitHub, ShopMetro, and local solid waste images captured by a smartphone. The YOLOv8n model achieved 85.31% average validation accuracy at 50 epochs and 86.50% average validation accuracy at 100 epochs. The result shows that the model achieved an 86.67% multiclass accuracy in identifying solid waste by using 30 samples per waste type at 100 epochs approach. However, the study is still in its initial stages because the hardware for segregation was not developed. Future work will include on adding the hardware feature to complete the system.