Developpement of an Intelligent Waste Sorting System

Abstract

This project presents the development of an intelligent waste-sorting system that combines computer vision with robotic automation. A custom-trained YOLOv8 model was employed to detect and classify waste into four categories: plastic, paper, metal, and glass. The physical sorting was carried out by a robotic arm, operated using inverse kinematics and PID control algorithms. A labeled dataset containing over 7,900 images was created and used for training via Google Colab. The system achieved a detection accuracy of 89.4% (mAP@50) and a sorting success rate ranging from approximately 87% to 90%. These results highlight the system’s effectiveness in automating waste management processes and minimizing the need for manual intervention.