Master’s Project Makes Sorting and Recycling Textiles Easier
Refiberd Tag Reader: Automated Label Collection for Textile Sorting Start-Up is the winner of the 2024 Sarukkai Social Impact Award
Master of Information Management and Systems (MIMS) alums Erin Jones, Isidora Rollan, Abdullah Azhar, Prashant Sharma, and Mustafa Hameed are the winners of the Sarukkai Social Impact Award, for their project, Refiberd Tag Reader: Automated Label Collection for Textile Sorting Start-Up.
The Sarukkai Social Impact Award was established by Sekhar and Rajashree Sarukkai in 2021. It recognizes capstone and final projects with the greatest potential to solve important social problems.
The MIMS team worked with an Oakland-based start-up to streamline the process of identifying and sorting textiles for recycling. The team created an application that captures an image of a garment tag and displays the fabric composition in the format required by the company’s processes.
We spoke to the team to learn more —
Tell us about your project.
Refiberd, an early-stage start-up in the textile recycling space, is using its proprietary supervised learning algorithm to automate and improve the accuracy of end-of-life fabric sorting processes. Refiberd’s model is trained on a dataset consisting of features derived from hyperspectral images and a set of fabric composition labels captured from non-standard tags attached to each training fabric sample. The model is only as good as their training set but gets better with each new sample it is trained on. However, the label capture process is manual and requires 50% of their workforce to conduct tedious dictation into an Excel spreadsheet.
Our team built a web-based application that automates the label capture process, which expedites and reduces the resources required to expand this portion of their training dataset by 50%. Our intervention applies information technologies to improve both the speed and scale at which a real-world start-up can reduce harm caused by textile waste.
What inspired your project? How did you decide on the concept?
Our journey began with an interest in learning more about the use of computer vision in the recycling industry, with hopes of specifically working on textile waste. The textile industry contributes more greenhouse gas (GHG) emissions than both aviation and shipping combined, and of all textiles produced, only 1% are estimated to be recycled. This is largely due to the extreme difficulty associated with each garment’s journey after being tossed in the bin. After visiting several waste sorting facilities and working with our mentors, it became clear that the best way to make an impact on such a large problem would be to work alongside brilliant individuals who already had a foothold in the space.
We serendipitously came across Refiberd, an Oakland-based firm, that was eager to work with us to find a mutually beneficial project. We met and chatted more with their team and identified goals and pain points: they wanted to expand their algorithm’s training dataset, but the process for doing so was manual and required significant manpower. Out of our conversations grew an idea to create an intervention that would automate, expedite, and improve the quality of this process using our I School skills.
This award is a wonderful reminder that everyone is positioned in some way or another to make a difference.
What was the timeline or process like from concept to final project?
Our research began in November, and we began our collaboration with Refiberd in January. After properly understanding their needs, we were already in crunch time, which is to be expected with a 6-month project timeline. From late February to early May, we went from an idea to a fully functional minimum viable product. This would not have been possible without the full commitment and dedication of our team members.
We had to learn many AWS services to enable the creation of a web-based application that could be accessed via mobile device, leveraged the use of a device camera, and ran inference on the image captured based on a fine-tuned computer vision model. We built application programming interface (API) endpoints, an entire front-end, and captured raw label images to fine-tune our model and put it all together. The result was an application that was ready for live demo and presentation to our client by graduation, presenting the possibility of significantly improving the speed and scale at which Refiberd could expand its training dataset.
How did your I School curriculum help prepare you for this project?
The project was end-to-end and would not have been possible without the interdisciplinary nature of the I School’s curriculum. Notable courses included product design, data engineering, natural language processing, computer vision, and front and back-end web architecture, which enabled all team members to contribute, regardless of their varied backgrounds and skill sets.
Beyond this, the continued mentorship of fellow students and faculty was crucial to the success of this project. We couldn’t have done it without Prof. John Chuang, with his Socratic guidance, Prof. Hany Farid, with his interest and initial guardrails for completing a successful exploration, and Prof. Kay Ashaolu, with his advice on building a cloud-based application.
Do you have any future plans for the project?
We are fine-tuning a delivery packet that will fully enable Refiberd’s team to implement the solution. We are also currently researching a few alternative options for their team to explore, such as batch processing of images instead of processing one image at a time as we currently do.
How could this project make an impact, or, who will it serve?
The application of machine learning in the energy transition and sustainability space is actively helping to improve the speed, scale, and accuracy of teams working in the field. While our solution won’t reduce emissions directly, it could help start-ups as they capture new datasets without the manual or tedious burden associated with the job. Better data leads to better models, which lead to stronger insights that stand to make an impact. In this case, freeing up 50% of the team that was initially manually capturing textile content from labels can help them deploy their solution to clients more broadly. The more accurately and expeditiously a technology like Refiberd’s is adopted, the sooner we will see fewer textiles blindly being tossed into landfills.
You were recognized with the Sarukkai Social Impact Award. How does that feel?
Amongst a cohort of truly awesome projects, our team feels honored to be recognized with this award. At times it felt that our intervention was small; however, with extremely difficult problems like climate change, the journey towards sustainability starts with baby steps.
This award is a wonderful reminder that everyone is positioned in some way or another to make a difference. We are lucky to have had the opportunity to work with the amazing folks at Refiberd with the support of Berkeley’s School of Information.
