Optimising Plastic Waste Management: YOLO-NAS-Based Robotic Sorting and Mobile App Prototypes for Adaptive Model Training

Effective waste management is essential for environmental sustainability and public health. Indonesia, the world's fourth most populous country, faces severe waste management challenges due to rapid urbanization and population growth. This study introduces a prototype robotic sorting machine powered by the YOLO-NAS model for classifying plastic waste. The system includes a mobile app that allows users to upload annotated images via FTP to a local database. This setup enables adaptive model training, continuously improving the model's accuracy over time. The model achieved a mean average precision (mAP) of 0.6, a recall of 1.0, and a validation precision of 0.018. Despite these promising results, the study identifies limitations and suggests enhancements for broader implementation in waste management systems.