In response to the growing challenge of textile waste exacerbated by the fast fashion industry, our project leverages artificial intelligence to revolutionize the way we manage post-consumer textile waste. By integrating a Convolutional Neural Network (CNN), specifically Resnet-18, for feature extraction from images of second-hand textiles, and pairing these features with non-image data, we use a decision tree for the critical task of grading textile reusability. This innovative approach not only aims to enhance the efficiency and accuracy of textile sorting and grading processes but also to significantly reduce the environmental impact by promoting sustainable waste management practices.
Our methodology involved meticulous data collection and processing, photographing 190 clothing items to capture a wide range of necessary attributes for accurate grading. We streamlined the grading system for clarity and efficiency, addressing challenges such as data scarcity for certain textile types. The use of PyTorch's transforms module and one-hot encoding facilitated the preparation of our dataset for analysis with the ResNet-18 model and a decision tree classifier. This combination of AI technologies offers a promising solution to the textile waste problem, potentially transforming waste management practices in the fashion industry and beyond by improving sorting and grading processes and promoting sustainability.
Code and Report Available upon request.