World-first open AI database for second hand goods

The second-hand market is growing rapidly in Europe, but the capacity to collect and sort clothes is severely undersized. At the same time, the European Commission has taken further steps towards a more sustainable and circular textile industry, proposing new rules that make producers responsible for the entire life cycle of textile products and thus obliged to finance collection and sorting that enables reuse and recycling. Part of the solution is a world-first open AI database for intelligent sorting and valuation of second hand developed within the AI for Resource Efficient Circular Fashion project. The project is a collaboration between Wargön Innovation and RISE Research Institutes of Sweden.

Sorting capacity in Europe is only sufficient to deal with a quarter of what is currently collected. The rest is incinerated or landfilled, despite the fact that as much as 50 to 75% of the collected textiles can be reused. Using smart technology and automation opens the door to an efficient and high-tech multi-billion dollar second-hand industry.

Today, there are many open databases for newly produced clothing, for example, Artificial Intelligence (AI) is used to improve online sales of newly produced garments. There is no equivalent for second-hand clothing - until now.

- This database is groundbreaking! It is the first of its kind that can analyze multi-attributes of used clothing and includes everything from material composition to stains and tears. This is not just data, it is a catalyst for sustainable fashion!" says Farrukh Nauman, AI researcher at RISE.

Second-hand clothing requires manual sorting and valuation, which is a physically demanding and costly task. In addition, it is difficult for a human to evaluate each garment in a short time. With the help of AI and Machine Learning (ML), tens of thousands of different garments can be recognized and those who evaluate the garments get a good basis for deciding whether the garments should go to second hand or not.

- The use of AI in the sorting of second-hand goods means a significant relief for the sorting staff. Both the speed and accuracy of the automation have the potential to be far beyond what human handling can achieve. The hope is that the database can be used by the many actors who, in various ways, sell or rent out second-hand clothes today, says Susanne Eriksson, project manager at Wargön Innovation and responsible for the AI for resource-efficient circular fashion project.

- "By creating an open database, we want to support digital developments in textile sorting and thereby encourage the use of data and the development of new digital solutions," she continues.

Thousands of garments scanned

In the AI for resource-efficient circular fashionproject, thousands of photos were taken at specially designed photo stations located at Wargön Innovation, as well as at Björkåfrihet and Myrorna's sorting facilities. The photo stations have scanned second-hand garments to create a completely unique open database. In concrete terms, this means that garments have been photographed and information about the pattern, material, brand, how nubby a garment is, possible price and degree of fashion has been saved in the database.

- "Using AI, second-hand retailers can generate detailed descriptions of garments and receive recommendations on which sales platform or store is best suited to sell each unique item," says Farrukh Nauman.

- This project is a first step to test the functioning of the database and the AI tool. We have already seen several areas of further development and opportunities for future projects. Textile sorting is becoming large-scale and large-scale sorting requires automation. The many different variables in the properties of the textile material make AI a very good tool that can be adapted to the situation and needs. But the database will not only be helpful for large-scale actors but also for smaller actors who can build algorithms based on their specific needs, says Susanne Eriksson.

One of several companies participating in the project is the second-hand shop Inimini. Josefine Runquist, founder of Inimini, is positive about the open database:

- It is so exciting to see that second hand and AI are developing and this will be so good for the future. If AI can help second hand become profitable, I see no problem for more companies to jump on this sustainable trend.

ShareTex and its CEO Miguel Sanchis Sebastia are also positive about the use of AI in sorting and valuing second-hand goods:

- This project has been a perfect opportunity to strengthen our relationship with Wargön Innovation and align the development of sorting technologies with the needs of textile recyclers. The AI database can become something like a quality assurance of our raw material, as it can help us guarantee that we are not recycling any materials that can be repaired or resold.

Further development and refinement

The project, which runs until April 2024, will now continue to scan and analyze garments to build the database with more information. The AI algorithms and the user interface developed will also continue to be refined with the aim of making sorting, selling and shopping for second-hand items as easy as possible.

- During the fall, tests of the tool will be carried out by the need owners (second-hand actors) so that they can provide feedback on development opportunities. AI can be used to recognize many different characteristics, but this requires data and building up data for all the characteristics you want to use AI for is a challenge. A challenge that we are happy to take on! concludes Susanne Eriksson, project manager at Wargön Innovation.

The AI database is developed in the project AI for resource-efficient circular fashion, which is a collaboration between Wargön Innovation, RISE Research Institute of Sweden, ShareTex, Inimini and the second-hand industry through Myrorna, Red Cross, Björkå Frihet and Texaid (Switzerland). The project is funded by Vinnova . The database is being further developed for large-scale sorting within the CISUTAC project.