Strategic monitoring of improperly disposed urban waste using UAV imagery and object detection

• UAV imagery and YOLOv7 detected improperly discarded plastic bags in Lima. • The model reached 34% precision and 33% recall under constrained data conditions. • Detection accuracy varied by bag color due to contrast, light, and occlusion. • The approach proved feasible for integrating UAV monitoring into MSW operations. Urban areas in the Global South continue to face persistent municipal solid waste (MSW) management challenges driven by rapid urbanization, informal disposal practices, and limited monitoring capacity. This study assesses the operational feasibility of integrating unmanned aerial vehicle (UAV) imagery with YOLOv7 object detection to identify improperly discarded plastic bags in public spaces across Metropolitan Lima, Peru. A total of 370 high-resolution aerial images from three districts were processed to detect black, yellow, and white bags, representative of local disposal behaviors. The model achieved a recall of 33%, precision of 34%, and a mean average precision ([email protected]) of 21.3% at an operational point corresponding to a confidence threshold of 0.24, reflecting the challenges posed by data scarcity and high visual heterogeneity in the urban environment. Although performance was constrained by class imbalance and low inter-class contrast, detections remained adaptable to diverse urban conditions through threshold tuning. UAV-based detection provides spatially explicit data to prioritize cleanup operations, verify citizen reports, and optimize waste collection routes. This proof of concept demonstrates that UAV-assisted monitoring can complement municipal records and support incremental improvements in sanitation logistics within resource-limited urban contexts.