Illegal dumping is a growing anti-social and environmental problem, but a Trinity Hall researcher is turning everyday bin lorries into unlikely crime‑fighters.

Using artificial intelligence and the cameras already installed on refuse collection vehicles, new research aims to spot fly‑tipped waste automatically and report it before it becomes a persistence nuisance.

The project, Deep Learning for Fly‑tipped Waste Detection, has just received a £25,000 grant through ai@cam’s Local Government AI Accelerator, a University of Cambridge initiative designed to help councils deploy practical AI solutions in public services.

The work is led by Tyler Holderness, a second year PhD student at Trinity Hall, and his supervisor Dr Florian Urmetzer, from the University’s Institute for Manufacturing.

Tyler said: “It’s exciting to do something that is immediately beneficial to our city, and applies my research in a practical way.”

Funded by the Ministry of Housing, Communities and Local Government, the accelerator pairs researchers directly with local authorities to tackle real‑world challenges. In this case it is being delivered in partnership with South Cambridgeshire District Council and Greater Cambridge Shared Waste.

The team will develop a proof‑of‑concept system that uses deep learning to detect fly‑tipped items as bin lorries go about their normal routes – helping councils act faster and more efficiently.

Originally from Cape Town, South Africa, Tyler’s research specialises in computer vision and deep learning. While his core research focuses on measuring human attention using subtle physiological signals, he is particularly motivated by applications with immediate local impact.

“ It’s exciting to do something that is immediately beneficial to our city, and applies my research in a practical way. ”

— Tyler Holderness

Tyler was a recipient of the Lee-Yung Family Fund for Entrepreneurship this academic year. For more on entrepreneurship funds see our alumni network webpages.

For more on the accelerator programme see their webpages.