Challenge

Eyya took part in Innovate UK’s BridgeAI programme and benefited from expert advice through The Alan Turing Institute’s Independent Scientific Advisor (ISA). With this support, the team created a technology that can automatically identify dirt, scratches, and graffiti on trains’ exteriors. EYYA’s TRACK system captures real-time images and video as trains pass through key inspection points, presenting the results through intuitive dashboards that give rail operators a clear, live overview of each train’s condition.

“Train operators can be fined for sending out unclean trains, so it’s an important issue for those companies… But until now there has been no automated system to assess train conditions. We are building one using advanced computer vision and AI techniques.”
EYYA CEO Dr Niazy Kioufi.

Approach

TRACK is now being piloted at a major rail depot in London, capturing over 4,000 videos per month of trains as they pass through. The team’s AI and machine learning specialists, including Evelyn Wu, Jingxin Du, and Negin Riazati, are developing a series of deep-learning models to process this footage – producing cleanliness scores and detecting damage.

Early models showed promise but faced limitations. For example, graffiti detection performed well, while scratch detection was much weaker due to an imbalanced dataset. That’s where Professor Diwei Zhou of Loughborough University was able to help, as EYYA’s assigned ISA.

“Professor Zhou gave us really practical, detailed advice,” says Evelyn. “She suggested technical improvements such as artificially generating more scratch images – a process known as data augmentation – as well as refining our image labelling process and splitting the graffiti and scratch models to avoid a phenomenon called overfitting, in which a model becomes too closely aligned to its training data. We have already implemented many of these suggestions, which has significantly improved the performance of our models.”

Niazy adds: “Taking part in the ISA offer has been a game-changer for us. It has given our team confidence, structure, and the reassurance that we are – if you’ll excuse the pun – on the right track.”

Reflecting on her work with EYYA, Diwei says: “What stands out about EYYA is their readiness for AI adoption. They know exactly what they want from AI, and they have a team with the right mix of expertise and a strong willingness to put guidance into practice. Their work is not only commercially valuable – it also contributes positively to society.”

Diwei’s support has helped EYYA move from feasibility to deployment, and has strengthened the company’s ability to engage new markets. The EYYA team is already exploring how the same technology could be used to assess the condition of ambulances in NHS depots, potentially reducing downtime for critical emergency vehicles. And EYYA’s long-term, international commercial partnership with the innovative Orange Train Wash system means there is scope to roll out the system to hundreds of train depots in the UK and overseas.

In addition to Professor Zhou’s guidance, EYYA has also received support from BridgeAI partner BSI’s Digital Trust Consultancy unit who undertook a review of the organisation’s approach to the management of AI, which included a high-level gap analysis of the start-up’s approach against ISO 42001 (the world’s first AI management system standard).

Working with Professor Zhou has been a hugely valuable experience – not just in terms of improving the performance and reliability of our machine learning models, but in helping us see where this technology can take us next. For a company like ours, operating in a niche space, that combination of technical insight and strategic guidance is invaluable.

– Dr Niazy Kioufi, CEO, EYYA Group Ltd

Next steps

Acutely aware of the amount of data that the organisation currently handles and the tools it employs, EYYA is determined to ensure it has in place a robust approach to AI, comprising policies, processes, and procedures compliant with the latest standards, ultimately to help the business to grow, and further develop AI-powered solutions for its clients based on responsible data management practices. BSI’s consultancy support, consisted of a one-day workshop to understand EYYA’s existing AI management system controls, and the subsequent provision of a high-level gap analysis assessment report against ISO 42001 as well as templates to support its implementation, elevated EYYA’s awareness and understanding of the standard, including in terms of its benefits, and provided the start-up with clarity on areas of non-compliance and the steps it will need to take to become compliant (including the order in which to do so).

According to EYYA CEO Dr Niazy Kioufi: “Using ISO 42001 and working towards certification will benefit us as an organisation by increasing our knowledge, confidence, and competence in how we manage our AI systems and ensure we do so in a more structured and professional way. It will also give us credibility with both existing and potential clients.”

Dr Kioufi goes on to add: “Moving forwards, customers may question the trustworthiness of TRACK’s dashboard and the data it comprises. Having standards in place seriously adds to our credibility with them, essentially providing comfort to customers, and linking what we’re practically doing and how we’re presenting it, with how we’re handling the results of that practical work.”

Over the next year, EYYA plans to invest in ISO 42001 and use the report and accompanying templates provided by BSI’s Digital Trust Consulting unit to help it create the documentation that the standard requires and to implement the associated processes and procedures with the view to acquiring certification (to the standard).

Dr Kioufi finishes with: “Implementing ISO 42001 will enable us to formalise what we’re doing and provide us with a clear view of where all our data is, how it flows, and how we interact and behave with it.”