YOLO-Based Waste Classification and Detection for Smart Recycling

The rapid growth of urban populations has led to a surge in waste generations, exceeding the capacity of existing waste management systems. Traditional manual sorting methods are labor-intensive, error-prone, and pose health risks to workers. To address these issues, this study explores the application of deep learning for automated waste classification and detection. We leverage the YOLO family of object detection models, specifically YOLOv8 and the newly introduced YOLOv9. We developed a comprehensive, annotated dataset comprising images of various waste categories, with single and multiple classes, creating Easy and Medium datasets for training and evaluation. YOLOv8s achieved a higher mean Average Precision (mAP) on the Easy dataset with single annotations, demonstrating its effectiveness for simpler tasks. However, YOLOv9, featuring innovations like Programmable Gradient Information (PGI) and Generalized Efficient Layer Aggregation Network (GELAN), outperformed YOLOv8s on the Medium dataset with multiple annotations. This highlights YOLOv9's ability to handle complex scenarios, making it a promising choice for real-world applications in automated waste management. Our findings provide valuable insights for developing efficient and accurate waste classification systems.