An Efficient Automated System for E-Waste Identification and Classification Using Yolov8

The growing quantity and intricacy of electronic waste (e-waste) is proving to be an obstacle for efficient waste management and resource recycling. Identification and differentiation of e-waste plays an indispensable role in the recycling sequence. This study is aimed at filling this gap through development and testing of an automated e-waste identification model based on deep learning technologies. Particularly, the study uses YOLOv8 (You Only Look Once version 8) object detection framework with transfer learning for enhanced accuracy and effective training. For this study, a subset of images containing five categories of e-waste, including laptops, monitors, mice, mobile phones, and keyboards, were retrieved from the e-waste repository on Kaggle. The methodologies included data preprocessing, which comprised image standardization and augmentation, model training unto 100 epochs, and evaluation using standard object detection metrics on a separate test set. The results demonstrate the effectiveness of the proposed approach with an [email protected] of 0.950 and [email protected] of 0.808. Furthermore, the model is said to be reliable with precision of 0.936, recall of 0.941, F1 score of 0.939, constituting efficient identification and localization of e-waste elements. To showcase practical applicability, the trained model was successfully deployed into an interactive web application using Streamlit, enabling users to upload images and receive real-time e-waste identification results.