Greyparrot: AI vision for the dirtiest place I could imagine
A London startup puts AI cameras over recycling conveyor belts, turning an opaque plant into a measurable system: proof that the best vision work happens where nobody expected it.
Here's something we take for granted: until recently, recycling plants estimated the composition of their waste by sampling. By eye.
I came across Greyparrot, a London startup that puts AI cameras over the conveyor belts. The software recognizes up to 111 types of material in under 60 milliseconds per object, while the waste rushes by at 3 meters per second. 98% accuracy on counting, 95% on mass. For every piece it estimates type, brand, size, economic value, and emissions potential.
This isn't "another computer vision model." It's vision applied to the dirtiest, fastest, most chaotic place there is, exactly where you'd never expect to find precision technology. The result: an opaque plant becomes a measurable system.
One detail that says a lot about how they think: the new Analyzer unit is built from recyclable materials and consumes 10 to 15% less power than the previous one. Across the network, it processes around 40 billion objects a year.
Sometimes the interesting innovation isn't the one that replaces a process. It's the one that finally makes it readable.
Sources
- Greyparrot, How it works: https://www.greyparrot.ai/how-it-works
- NVIDIA Technical Blog: https://developer.nvidia.com/blog/ai-vision-helps-green-recycling-plants/