
In a new technology project, researchers, second-hand, innovation and recycling stakeholders are helping to develop the sorting, valuation and marketing of second-hand clothes using artificial intelligence, machine learning and automation.
AI for circular fashion
A new green Swedish industry is developing with the potential to provide a wider and better range of second-hand clothing as textile collection increases significantly. If a mere five percent of Europe's textile consumption is replaced by second-hand garments, emissions in the industry would decrease by four percent, totaling 19 million tons of CO2 per year.
In a new technology project, researchers, second-hand, innovation and recycling stakeholders are helping to develop the sorting, valuation and marketing of second-hand clothes using artificial intelligence, machine learning and automation.
The collection and sorting of textiles in Europe faces major challenges. Today, between 50% and 75% of collected textiles are reusable, but handling is labor-intensive and highly manual. In addition, the sorting capacity is about a quarter of everything collected. This means that the amount of textiles collected is already so large that it cannot be handled responsibly and most of the clothes are burned or end up in landfills. The amount of clothing collected is also expected to increase dramatically in the coming years.
- When the new EU Waste Directive comes into force in 2025, all textiles from private individuals must be collected, which means an expected doubling of collection volumes compared to today. The time aspect is critical and to increase the sorting capacity, large scale and automation are required. says Susanne Eriksson, project manager at Wargön Innovation and project manager for AI for circular fashion.
The new technology development project AI for Circular Fashion is developing solutions for sorting and valuing textiles using artificial intelligence and machine learning.
Artificial intelligence is already being used as a technology for self-driving cars, facial recognition, e-commerce and the fashion industry. It is now moving into the growing second-hand clothing market. Among other things, the project explores the potential of AI to identify a garment's condition, brand, wear and damage alongside trends and demand in different markets. This is combined with robots and automation to develop second-hand handling into a large-scale, high-tech industry.
The project starts with the construction of a test pilot with a specially designed photo station in Wargön Innovation's facility in Vänersborg municipality.
The facts
Responsible at Wargön Innovation: Susanne Eriksson
Project coordinator: Wargön Innovation
Project participants:
RISE, Inimini, Sharetex, Myrorna, Red Cross, Björkå Frihet and Texaid.
Project duration: Oct 2021 - April 2024
Budget: 7 000 000 SEK
Funder: Vinnova together with the project partners.
Outcome
A key goal of the project was to create an open dataset of 30,000 used garments, which was successfully achieved with a final dataset of 31,997 garments. Each garment was documented with three images and several attributes, such as type of garment, color and material, which allowed for a thorough analysis and training of AI models.
The entire dataset is open and available to everyone on Zenodo.
- The project managed to conduct proof-of-concept tests where AI models proved to be faster than manual methods, although they did not reach their full potential due to time constraints.
- A life cycle assessment (LCA) was carried out to assess the environmental benefits of AI-assisted sorting. The results indicated that even a small increase in the reliability of sorting could lead to significant environmental benefits.
- The project demonstrated the potential of AI technologies to not only streamline sorting processes but also contribute to greenhouse gas emission reductions by promoting textile reuse and recycling.
The project laid the foundation for future developments in circular fashion and AI-assisted textile sorting, which will have a major impact on reducing the negative environmental impact of the textile industry. Here you can read the full final report with an in-depth description of the results.