Skip to content
Artificial intelligence. Clearly. RSS feed
AI NEWS AI

News. Evidence. Perspective.

AI textile sorting moves into the mixed-waste stream

Tecnalia’s plant trials combine three kinds of imaging to find discarded textiles among other waste, connecting identification to a wider recycling project.

Tecnalia has taken AI textile identification into a waste-management plant, testing it with Valoriza on separated textiles, non-textile rubbish and mixtures of the two. Its English-language update, published on 23 October, says footwear was added to make the trials more demanding.[1]

The system combines ordinary colour images, near-infrared sensing and three-dimensional imaging. The intended gain is finding recoverable textiles inside a mixed stream, as well as sorting material already collected separately. The update gives no accuracy or recovered-tonnage figures.[1]

This is part of CIRTEXCAM, which Valoriza lists as an active 2026–2028 project. Its broader task is to return discarded textiles to production and reduce reliance on virgin material.[3]

A classification has to lead somewhere

Treelogic, another participant, describes models that combine RGB and infrared information to identify garment composition. It is developing and validating multispectral classification: using information from different wavelength ranges to distinguish material. Those descriptions connect AI’s job to the composition of the recycling feedstock, rather than simply recognising that an object is clothing.[2]

Consider a sorting line that successfully recognises a jacket but sends it to a process unsuited to its fibres. Recognition has worked; the recovery task has not. That hypothetical example explains why a useful classifier needs an onward route for the material it identifies.

UDIT’s project account supplies that downstream picture. Planned work spans thermomechanical, thermochemical, chemical and enzymatic recycling of difficult textile mixtures. Recovered material is to become new yarn, blended with merino wool from native sheep breeds in the Madrid region, then made into knitted outer garments for evaluation in working environments. These are project plans, not reported production results.[4]

UDIT is also developing an ecodesign model incorporating life-cycle assessment and circularity. In that arrangement, decisions about what to manufacture and how to recover it belong in the same project. Better information at the sorting stage can help make the link between an unwanted garment and a usable next product concrete.[4]

The plant trials now give that larger programme a harder starting point: textiles encountered alongside the other things people throw away.[1]

Follow the evidence

Sources & further reading

  1. Treelogic joins the CIRTEXCAM consortium Treelogic · 3 Apr 2026 · Accessed 28 Oct 2026
  2. CIRTEXCAM Valoriza · Accessed 28 Oct 2026