Recycleye uses AI to detect and sort WEEE for the “first” time


Recycleye AI

Recycleye and SWEEEP Kuusakoski have announced the “first successful commercial application” of AI computer vision to detect and sort WEEE (Waste Electrical and Electronic Equipment) in the UK.

The companies used an optical sorter that utilises AI and machine learning to sort e-waste for recycling. Recycleye explained that while the use of AI for automated sorting is increasingly being used to sort household waste, it has not yet been widely applied to WEEE or metals.

Waste management technology company Recycleye said that existing optical sorters in waste facilities already use pneumatic ejection but the integration of the technology with AI rather than near-infrared (NIR) is “novel to this application”.

Using AI to detect objects means they are identified by a range of visual features, just like a human eye, rather than by purely colour and light-based sensors, Recycleye explained.

The company said the new technology can detect and extract precious metals for recovery. Similarly, the multi-material nature of batteries makes them difficult to detect with NIR, Recycleye said; however, AI has the potential to detect and sort batteries based on visual features.

This is an important milestone for the use of our AI and sorting technology.

Recycleye installed its AI-powered optical sorter at the back end of WEEE collection and recycling company SWEEP Kuusakoski’s plant in Sittingbourne.

Recycleye explained that the technology sorts between higher-value items with precious metal content, such as copper, PCBs, cables and brass, and lower-value materials, such as aluminium, plastics, steel, ferrous metals and batteries.

Commenting on the new partnership, Zoe Cook, Technical Sales Manager (UK) at Recycleye, said: “This is an important milestone for the use of our AI and sorting technology.

“This successful application in the sorting of WEEE demonstrates that AI-powered sorting automation can be utilised to tackle even more waste categories, due to the flexibility to adapt machine learning models to different streams.”

Send this to a friend