Uncollected Solid Waste Detection and Reporting Using Machine-learning and Geotagging

Uncollected solid waste has become an ever-growing threat to our livelihoods, posing many problems that span health risks, environmental pollution, and unpleasant aesthetics. Traditional methods of reporting uncollected solid waste have proven ineffective and we struggle to keep pace in addressing the growing volumes of waste. This study focused on developing a solid waste detection and reporting system based on machine learning and geotagging. The developed system sought to enable solid waste management entities to obtain verified solid waste reports and accurate location details of the uncollected waste instances. A ResNet model was trained on commonly used solid waste image datasets from online sources. The model achieved a provisional accuracy beforescore of seventy percent (70%) though this statistic can be improved by adjusting model parameters. A rapid iterative design approach was employed to facilitate the development of the crowdsourcing app. This enabled the researchers to swiftly build, evaluate, and refine functional prototypes of the system before finalization. The resulting system receives solid waste images together with their location data and verifies the correct image reports using the trained model, before finally visualizing the reports on a map. Overall, the system provides a platform for the general public to collaborate with solid waste management bodies in combatting uncollected solid waste. In the future, the system may be diversified to allow reporting other amenity issues besides uncollected solid waste.