AI identifies Finnish waste, but people still do the most effective sorting

An artificial intelligence system is learning to identify Finnish waste, but people remain essential to sorting it, Yle reports. A camera at a waste-to-energy plant in Mustasaari analyses hundreds of images a second to find recyclable or unsuitable materials in mixed waste before it is burned.

The system was developed through the partly EU-funded Ekoälyä project, involving the University of Turku, Haaga-Helia University of Applied Sciences and waste company Westenergy. Project assistants manually classified thousands of images to train the AI. A Finnish dataset was needed because waste packaging differs between countries: the article notes, for example, that a Finnish milk carton looks different from an American one.

Westenergy’s plant receives 40 full-trailer loads of mixed waste each day and burns 190,000 tonnes a year from 70 municipalities. Manual spot checks cover only about a day’s worth of waste over the course of a year. The camera and a small computer now analyse test batches tipped onto the facility floor.

Westenergy business intelligence manager Juha Ripatti says the system could help identify unsuitable materials and link them to the company that delivered a load. Knowing the waste composition could also help the plant adjust its combustion process and the chemicals used to clean flue gases. Recyclable materials, including plastic, could be diverted from incineration.

The pilot system has been trained to recognise about 20 selected types of waste. Each type requires around 1,000 manually checked images for reliable identification. The technology works when waste is visible to the camera, but mixed household waste is a challenging target. The system can also struggle to distinguish wood from cardboard, particularly when items overlap or are covered.

The camera cannot see inside piles or rubbish bags, and Westenergy does not open bags at the facility because the aim is to identify waste quickly without adding manual work. University of Turku postdoctoral researcher Pouya Jafarzadeh says the more waste is sorted at home or at work, the easier it is for AI to complete the sorting.

Ripatti says people and businesses remain the most important part of the process. Sorting packaging at home, such as putting a plastic package in the plastic recycling bin, is the most cost-effective way to sort waste.

Haaga-Helia project manager Sami Nykter says the data could help meet the EU’s circular economy targets and support cross-border trade in recycled materials.