Can AI textile sorting give wearable clothes another chance?
A worn shirt might still have years left as a shirt. I'd love a machine that could spot it in a mountain of used clothing and help it find another wearer. NVIDIA's November 10 profile describes reverse.fashion using high-resolution cameras and AI to assess garments, including small defects, and sort them into different categories. This is a supplier's account of a startup using its hardware, not an independent test of how much clothing gets worn again. The startup's own product page makes a useful distinction: line.sort chooses the most profitable category under the customer's saved rules; co.sort recommends categories while a human sorter decides. Neither description, by itself, tells us how many wearable shirts miss a second life. I'd want to open a sample of the clothes routed away from resale and ask an experienced sorter which could still be worn or repaired. Were they rejected for damage, or because there wasn't a buyer for that style? Those call for different remedies. That's a check I'd ask for, not a failure these sources document. Sorting faster could make looking through low-value clothing worthwhile. What would convince you that it gives more garments another wearer, rather than just getting them through the building faster?
Comments
I'd follow a sample from the resale pile to the buyer, too. A garment assigned to resale hasn't necessarily found another wearer. Ask the receiving shop which pieces it accepts, which it rejects and why; keep unsold stock separate from confirmed sales. That would show whether better sorting is helping shops stock clothes people want, or passing the sorting job downstream. Even a sale wouldn't prove how long a garment stays in use, but it's firmer evidence than a category label. This is a follow-up I'd propose, not a result reported by reverse.fashion or NVIDIA. It would also give the sorter useful feedback: a missed stain and a style that doesn't sell need different fixes.
I'd separate what the camera missed from what it couldn't sense. A shirt can look ready to sell but smell of smoke, or have elastic that no longer springs back. In that shop feedback, leave room for ‘needed a smell or stretch check’ rather than labelling every rejection a visual sorting mistake. I'd ask where those hands-on checks happen and how much clothing staff still have to handle. That's a question about the process, not a fault documented in this system. Saving someone from inspecting every seam would still be useful even if a camera can't finish the whole job.