A Technique Using Machine Learning to Anticipate and Differentiate Between Biodegradable and Non-Biodegradable Waste

Abstract

Urban waste has become a significant issue for planners due to the challenges of identifying and disposing of it. The rise in urban populations has resulted in a corresponding increase in waste and garbage. To address this issue, in this study, the researchers introduce a concrete approach that utilizes a Deep Learning (DL) framework to perform waste sorting at its basic level. In contrast to recognizing objects of a specific category, waste can have various characteristics such as color, shape, material or size making it challenging to detect. To overcome this, the authors proposed a material-based deep learning model called Smart-Bin, which employs an Improved Faster Recurrent Convolution Neural Network (IFRCNN) approach to differentiate between biodegradable and non-biodegradable waste. The aim of this study is to evaluate the performance of various IFRCNN models such as AlexNet, ResNet, InceptionNet, and VGG-16 together with the hardware system implemented for waste classification within the bin, the suggested technique demonstrated superior performance compared to other models. The InceptionNet Neural Network achieved remarkable precision rates of 98.15% and a training dataset loss of 0.10, while achieving 96.23% precision and a loss of 0.13 for the validation dataset.