‘What are your clothes made of?’ is a deceptively difficult question. AI can help answer it.

• AI can help enhance garment-sorting accuracy in the fashion industry.
• The company Refiberd uses AI and sensors to identify recyclable textiles that can be repurposed.
• This article is part of “Build IT,” a series about digital tech trends disrupting industries.

For decades, the fashion industry has had an unlimited-growth mindset, fueled by cheap production and opacity. Even a seemingly simple question like “What are your clothes made of?” can flummox brands because fashion supply chains are so lengthy and far-flung.

It doesn’t have to be this way, and AI could become a major force for a more sustainable industry.

In the past few years, brands have been accused of mislabeling their garments’ fabric composition and country of origin. A study conducted in the Netherlands involving about 7,000 garments found that 41% of the garments had been mislabeled. Mislabeling poses an issue not just for curious consumers but for textile recyclers, who need to know exactly which fabrics they’re dealing with in order to process them effectively.

Recycling technologies have improved over the past few decades, but vast piles of clothes need to be sorted first. This can be hard for workers to do accurately, and errors can be costly.

With their backgrounds in textile engineering and AI, Sarika Bajaj and Tushita Gupta, the cofounders of Refiberd, spotted this accuracy problem in the late 2010s and understood that AI could help solve it.

To identify textiles, Refiberd uses AI and a sensor called a hyperspectral camera. “It’s almost like a video camera which pulls out near-infrared data,” Bajaj told Business Insider. “Light gets reflected onto our sensor, and that tells us what materials we’re dealing with. What you end up with are spectra, graphs that tell you there’s an extra 1% here of elastane or an extra 2% there of cotton.”

Refiberd tech uses a regression-based model, which estimates fabric composition and is more precise than a classification model that sorts materials into broad categories.

Bajaj said the regression-based model works with the help of a materials library made up of “thousands upon thousands of samples,” allowing Refiberd to “detect textile material composition within 2% of the actual material composition.”

The problems with fast fashion

To understand how we got here, it’s worth breaking down exactly how fast fashion — the leading cause of the industry’s textile-waste problem — is made. Bajaj described the bulk of consumer-bought clothes as a “notorious mix of blends and contaminants” often manufactured in various factories worldwide.

“What we have found is that there’s variability at every stage of the process,” she said.

“We might see a garment made of organic cotton and think, ‘Organic, amazing,'” Bajaj added. “Yet the process of dying organic, virgin cotton at a certain level just makes it a plastic.”

Synthetic dyes in garments require enormous amounts of water and contain contaminants. “At what point do we still call this pure?” Bajaj said. “Then that garment might be sent off to a different factory for cut and sew, but those workers added polyester thread everywhere, and none of that has been accounted for on the label.”

A lack of regulation in the fashion industry means creative solutions are needed to tackle the problem. Leading in this field is the Hong Kong Research Institute of Textiles and Apparel, or HKRITA, known for its sustainability-focused tech.

In addition to a smart garment-sorting system — which also uses hyperspectral cameras — HKRITA has several innovations designed to address what its CEO, Edwin Keh, sees as a deeper-rooted issue in the fashion industry: a refusal to consider product lifespan.

“There’s a lack of thoughtfulness about reuse and second use,” Keh said.

He said that while aluminum beverage cans’ high recycling rates could be attributed partly to their use of a single material, the fashion industry’s reliance on material blending to cut costs had complicated recycling. “We use all sorts of blends and coatings,” Keh said, “because we just didn’t think about second use.”

Keh described garment sorting as the “big bottleneck,” given how manual it is. But there are other remedies. HKRITA’s projects include an AI-powered mechanical recycling system designed to identify reusable indigo-dyed warp yarns in denim and separate them from nonrecyclable synthetic yarns.

Why securing investments can be difficult

Both Refiberd and HKRITA have received cash awards from the H&M Foundation, a nonprofit funded by the founders of the H&M Group. The foundation, which says it aims to create a more “socially inclusive and planet positive textile industry,” has partnered with HKRITA on a program called Planet First.

The novelty of such innovations means securing investment can be a challenge. “As a philanthropic organization, we can support these ideas when they’re at a very critical stage and need funding to further develop, but they’re too early for many investors,” Anna Gedda, the global manager of the H&M Foundation, told BI. “We can absorb that risk and support them to develop and scale.”

The necessity of these solutions can’t be underestimated. Tons of unwanted clothes are sent to landfills daily, and tons more are sold to traders who sort through clothes to find items of value and discard the rest. Rejected items are dumped out of sight, often in countries like Ghana, whose secondhand-clothes dealers have taken a stand against this practice, or in remote locations like Chile’s Atacama Desert, recently dubbed fast fashion’s great garbage patch.

“Very few people in North America and Europe are seeing where these offshore deliveries are going,” Bajaj said. “They’re going to the magical clothing pile in the sky.”

Transitioning to AI-powered tech

These offshore deliveries aren’t wholly bad; many resellers rely on them. “The thought that I’ve heard is if these clothes are still going to go there, it should be the clothes that are actually going to get used, not trash,” Bajaj said. “The question is how you sort for usability in those markets so that whatever garments these resellers are buying are actually high quality and actually have a resale life.”

Transitioning to AI-powered tech could preserve jobs, with appropriate workplace protections and government regulations, while maximizing recycling rates, Bajaj said: “In the fashion industry, AI is being seen more as a supplement than a replacement.”

Bajaj views these tech innovations as a pragmatic response to human accuracy’s limitations. “Our goal is just to support the textile-recycling industry as much as we can,” she said. Complex, blended fabrics pose huge challenges for textile recyclers, not to mention the environment. Bajaj said Refiberd aims to remedy those difficulties to see recyclers “continue to grow.”