Real-time Sorting of Plastic Waste for Pre-Cleaning Recycle

Despite Thailand’s longstanding efforts in plastic waste sorting, a substantial portion remains unrecycled. This paper introduces an innovative approach utilizing the YOLOv8s deep learning algorithm for real-time plastic waste sorting, specifically targeting five categories: bottles (left and right), glass (left and right), and food containers. The dataset comprises 408 images of everyday plastic waste, including grayscale variations. Our experimental results demonstrate an average F1-score of 85%, precision of 96.6%, precision-recall of 89%, and a confusion matrix showing classification confidence levels of 100% for bottle left and food containers, 87% for bottle right and 86% for glass right.