Sustainable Fabric Recycling using Hybrid CNN-LSTM Multi-Classification Model

Abstract

The textile industry is one of the largest contributors to environmental degradation; nevertheless, the implementation of recycling practices for textile waste has the potential to significantly reduce the severity of this impact. The current study addresses the challenge of multi-classification in fabric recycling by presenting a unique strategy that blends a convolutional neural network (CNN) with a long short-term memory (LSTM) network. This approach was developed as part of this research. Following the collection of a dataset that included 10,000 photographs of different types of cloth, the data was then sorted into four unique recycling categories, namely mechanical recycling, chemical recycling, upcycling, and downcycling. An overall accuracy of 92.63 percent was achieved by the hybrid model that was recommended. The category that displayed the best accuracy was the mechanical recycling category, while the upcycling category demonstrated the highest recall. On the other side, the downcycling category received the maximum possible score in the F1 competition. According to the data, the model that was presented demonstrates a high degree of efficacy in the categorization of waste textiles into various recycling groups. This is the case. Because of its ability to maximise the classification and reutilization of textile waste, the application of this strategy has the potential to make it easier to develop a textile industry that is environmentally responsible.