Towards Improved Sustainability in The Textile Lifecycle with Deep Learning
Keywords: deep learning, object detection, textile recycling, textile sustainability, traceability
TL;DR: Combining low cost hardware and deep learning techniques to automatically sort textiles and track textiles through their lifecycle
Abstract: The garment industry is one of the world’s largest carbon and waste polluters. In the next decade, this industry is expected to produce 150 billion garments per year, while currently recycling about 1%. Garment landfills are growing large enough to be seen from space, while the water consumption and manufacturing side effects threaten both the environment and human health. Creating a sustainable circular economy for textiles in hampered by two key challenges – fabric identification and tracking. Without precise automatic fabric identification – scalable recycling measures cannot be put into effect. Without traceability, governments cannot enforce recycling laws and incentives. We propose two solutions to this problem – leveraging low-cost hardware and deep learning. Approach A – using microscope fabric images and Convolutional Neural Networks – demonstrates classification accuracy of over 90% for 14 fabric classes. Approach B, marking fabrics with a binary code visible only under black light and using YOLOv8 object detection to remain effective in the presence of unique fabric challenges such as creasing, wear, and light refraction, demonstrates an mAP 50 of over 0.98, retaining up to 0.93 after wash cycles. We outline a traceability system based on Approach B that can be implemented worldwide at a low cost to enable global fabric traceability for a functioning textile circular economy. Finally, we provide 3 open-source fabric datasets to encourage further research.
Submission Number: 12
Loading