Urban waste recycling: smart sorting with AI

The disposal and recycling of municipal waste is one of the most pressing environmental challenges of our time. According to the US Environmental Protection Agency (EPA), approximately 75% of urban materials can be recycled, but only 35% of it is actually recovered. That means over 68 million tonnes of recyclable materials end up in landfills or incinerators every year. And even when waste is correctly sent for recycling, there is still the issue of contamination: non-recyclable materials find their way into recyclable waste streams at a rate of around 25%, compromising the quality and economics of the process, according to the EPA. 

In this context, automation and smarter sorting are crucial steps towards making the circular economy a practical reality.


Scalable, affordable, replicable model 
AI is a key tool in improving this process in a number of ways: it can reduce contamination in waste streams, cut operational costs and risks, and lay the groundwork for a scalable, cost-effective and replicable model of smart sorting. A new project developed by Stony Brook University with the support of the AI Innovation Seed Grant is proof of this: it is a system based on video footage and artificial intelligence algorithms capable of analysing, monitoring and potentially automating the recognition of recyclable materials in plants on Long Island, across from New York City and Connecticut.  

Researchers collaborated with the Waste Data and Analysis Centre and the New York State Department of Environmental Conservation (NYSDEC) to collect high-resolution images from actual sorting processes, with the goal of creating one of the industry's first public datasets. <<We are not just building stand-alone tools, but collecting real data from multiple stages of the process and working with recyclers to understand the trouble points as well as potential solutions to improve speed, safety and operational awareness,>> said Ruwen Qin (Department of Civil Engineering), who coordinates the team. 

To classify and quantify recyclable materials in motion, the project uses innovative computer vision technologies, including object detection (You Only Look Once – YOLO), segmentation (Segment Anything Model – SAM) and tracking (Deep Simple Online and Realtime Tracking – DeepSort) systems. The anticipated outcome is a system that can support operators, facilities and administrations in managing flows more efficiently, with the aim of fostering an open ecosystem of collaboration between research institutes, industries and the public. 
 

• You may also be interested in: The future of small WEEE: reuse, repair, and recycling

Towards cleaner, smarter recycling 
The Stony Brook University initiative is part of a broader wave of research exploring how artificial intelligence can revolutionise waste management. In several countries, universities and companies are experimenting with systems capable of recognising materials in real time, rectifying sorting errors and learning from data collected daily in facilities. 

According to a study published in Sustainability Global, the use of AI can increase recycling rates by up to 50%, with 90% accuracy in material recognition, compared to 60% for manual sorting (Waste Management Review, 2025). The future of recycling will depend more and more on the ability to render processes fast, accurate, and transparent, minimising contamination and manual intervention. To be fully exploited, however, artificial intelligence will need good data, adequate infrastructure, and reliable algorithms that offer comprehensible and verifiable results. 

Despite these challenges, the advantages are clear:

• more efficient facilities,
• lower costs,
• greater safety for operators
• reduced environmental impact. 

As also highlighted at Ecomondo 2025, the direction is therefore towards an integrated approach that combines real data, human expertise and intelligent algorithms, with the aim of transforming sorting centres into true hubs of the circular economy. 

 

• You may also be interested in: Ecomondo looks to the future of the circular economy and sets a date for 2026

 

Written by Maria Carla Rota 

This blog is a joint project by Ecomondo and Renewable Matter