How Rematics aims to make sorting recyclables smarter through AI-powered computer vision

Every morning, in towns and cities across the world, refuse collection trucks make their rounds through neighbourhoods, emptying bins and hauling waste to processing plants. But as rubbish is tossed into the vehicle, one question lingers—how are recyclables separated from the rest?

While we try our best to separate the recyclables from the waste, nobody is perfect, and intermixing seems inevitable. This causes a problem because a stray leaking battery may contaminate the entire recycling batch and endanger waste management employees.

Refuse collection trucks typically transport collected materials to a sorting facility. There, advanced technologies like optical scanners, magnetic separators and air classifiers— which sort materials by density using air flow— often work alongside manual sorting to separate recyclables by type, such as metal and plastic, while filtering out non-recyclable waste.

However, most truck drivers do not have a clear idea of what is in their vehicle before they reach the sorting facilities, creating significant inefficiencies.

Enter Rematics, a Belgian company that has built a camera powered by artificial intelligence (AI) to improve the efficiency of waste collection. The camera scans the waste and analyses its properties in real time as it is dumped into the truck. The AI then analyses the footage to distinguish recyclables from general waste.

“We need to optimise sorting at the ground level, as close as possible to the end consumer, and the collection moment is the right time to do that,” said Rematics CEO Bram Peetermans.

Building the technology

To achieve maximum performance, Rematics is deploying machine learning, continuously training a neural network with massive datasets gathered from sources ranging from publicly available photographs to synthetic data built using different training techniques. The company’s method ensured a reliable algorithm that has consistently generated accurate data of all collected waste material. 

Belgium is an ideal testing ground for Rematics’s technology. Despite the country achieving impressive recycling rates of over 90% for materials like glass and metals, the recycling rate for plastics lags behind at 54%.  

Additionally, new European Union rules that significantly limit the exporting of plastic waste is set to be rolled out by 2027. “So we have to monitor what comes in and what comes out accurately,” said Peetermans.

That reporting process has typically been extremely manual (and time-consuming) and often falls on the responsibility of truck drivers. It is a complicated task that adds quite a burden to an already difficult job, so the reporting is often ignored or dismissed.

By providing accurate and real-time data on the waste that goes into the trucks, Rematics’s camera could also make it easier for waste collectors to sort recyclables from other waste at the collection points, before they reach waste processing plants.

“We need to relieve the staff from administrative burdens while providing insights to the collection companies that allow them to optimise their operations,” said Peetermans. “And, of course, we have a bigger purpose. We do believe that if we make it easier to sort and to collect recyclables the proper way, it will improve the sorting rates.”

The road to commercialisation

In 2024, Rematics launched and installed its first AI-powered camera prototype on a waste collection truck.  The camera can identify and help sort plastic materials with a 70+% accuracy rate.  

Peetermans cited Alliance to End Plastic Waste’s crucial role in supporting the technology’s development. Besides providing a US$300,000 grant, the Alliance provided advice on overcoming technology hurdles.

“I really believe that we are evolving the solution from a proof-of-concept phase into an industrial prototype, and that’s really where the Alliance helped us,” said Peetermans.

‍

Peetermans said that Rematics’s current priority is on refining the final commercial product so that it is resilient and can endure the demanding conditions of a waste collection truck. The company is also focusing on increasing the system’s accuracy to up to 90%.

“We’ve received a lot of demand for our cameras from the market because every point in the value chain wants an accurate analysis of what is in the waste,” said Peetermans.

To date, Rematics has been in discussions with six recycling companies to validate its technology’s use case and potential return on investment. Peetermans said that it was able to showcase up to 10% savings in operational costs. “We are showing that we can help these companies optimise routing and logistics by delivering insights about the collected waste as well as improving their commercial discussions with clients.”

Besides making recycling more efficient, another of the camera’s key selling point is safety. By being able to identify and remove hazardous waste materials quickly, recycling companies can mitigate the risk of an accident, such as a collection truck catching fire because of flammable waste materials.  

And it’s not just Belgium that’s in Rematics’s sights. Over the coming years, the company hopes to expand into neighbouring France, Germany and the Netherlands, spreading its vision to other countries with similar waste management challenges.

The company has become a good example of the types of ingenuity required to help recycling industries improve so that they can ensure the plastics stay within the circular economy. The problem demands we find inefficiencies or places where small enhancements can make a big difference.

Rematics is trying to perfect the recyclable sorting process and, in doing so, will ensure that some of the plastic destined for landfills or incineration will be recycled for future use.