Development of Intelligent Waste Segregation System Based on Convolutional Neural Network

Abstract

As the world undergoes rapid development in all spheres, the phenomenon of globalization has led to rapid waste generation. Waste management poses a considerable challenge to the country's health, economy, and aesthetics. New techniques are being employed by waste management companies to automatically manage and segregate waste. The age-old method of manual sorting of waste is outdated and inefficient in terms of time and cost. Moreover, it also poses many risks to the health of those involved in the process. This research utilizes a Convolutional Neural Network (CNN) based on a Residual Network structure to develop an intelligent real-time waste segregation system that classifies the waste and automates the process of waste segregation with maximum efficiency. The development of the CNN model is performed by systematically optimizing the models' hyperparameters to achieve optimal performance. The proposed model was evaluated based on different metrics, including accuracy, precision, recall, f1-score, and confusion matrix. The testing evaluation results of the optimal model showed that the model achieved accuracy ranged from 93% to 98%, Precision score ranged from 81% to 97%, Recall scores ranged from 81% to 96%, and the F1-score ranged from 81% to 95%.  In order to facilitate the model implementation in the waste management industry, a prototype system was developed to simulate the real-time implementation.