Development of Automatic Waste Classification System using CNN-Based Deep Learning to Support Smart Waste Management

This research develops an automatic waste classification system using deep learning based on Convolutional Neural Network (CNN) to support the implementation of Smart Waste Management (SWM). The main objective of this research is to design and test a CNN model that is able to classify various types of waste, such as plastic, paper, organic, and other non-organic waste, with high accuracy and efficiency. The developed CNN model successfully achieved an accuracy rate of 94.86% on the training dataset. The system performed very well in classifying recyclable waste with a precision of 56.6% and recall of 63.5%, although it still faces challenges in the classification of organic waste with a precision of 45.7% and recall of 38.8%. This research also includes model validation using cross-validation techniques to ensure the generalizability of the model on different datasets. In addition, tests were conducted on external datasets to evaluate the robustness of the model under real-world conditions. Data preprocessing techniques such as image normalisation and data augmentation were used to improve the performance of the model. The results show that a CNN-based automated waste classification system has great potential to be implemented in SWM systems, enabling more efficient and automated waste management. However, there are still some challenges such as high variation in litter images and dataset limitations that need to be addressed for future development of a more robust system