Optimizing Waste Classification Model using YOLOv11 Architecture

Abstract

Municipal solid waste management remains a critical challenge due to rapid urbanization and consumption patterns. This study proposed a image based waste classification model for organic, inorganic, and hazardous (B3) waste using the YOLOv11 architecture. To conduct the study, we gathered a huge dataset of 5,000 images across daylight, dusk, and night conditions. According to experimental results, the proposed model can achieve an [email protected] of 70%, a precision of 69%, a recall of 70%, and an F1-score of 0.70, operating at 43 frames per second (FPS) with 102 GFLOPs. It can confirm its suitability for real-time applications in resource-constrained environments. Compared to heavier deep learning models, this efficiency-performance balance highlights the practical advantage of YOLOv11 for continuous waste monitoring and automated sorting systems.