Deep Neural Networks as a Tool for Enhancing the Efficiency of Plastic Waste Sorting

Abstract

In the recycling industry, there is an urgent need for high-quality sorted material. The problems of sorting centers related to the difficulties of sorting and cleaning plastic leads to the accumulation of waste in landfills instead of recycling, emphasizing the need to develop effective automated sorting methods. This study proposes an intelligent plastic classification model developed on the basis of a convolutional neural network (CNN) using architectures such as MobileNet, ResNet and EfficientNet. The models were trained on a dataset of more than 4,000 images distributed across five categories of plastic. Among the tested architectures, proposed EfficientNet-SED demonstrated the highest classification accuracy – 99.1%, which corresponds to the results of previous research in this area. These findings highlight the potential of using advanced CNN architectures to improve the efficiency of plastic recycling processes.