This is where the University of Edinburgh came in. The Creative Informatics programme launched a Challenge Project offering funding to individuals or SMEs that could develop a data-driven solution to The List’s challenge. The winning responder was Viapontica AI, a boutique machine-learning consultancy based at the Bayes Centre. It was selected because the company showed a strong understanding of the complexity of the task at hand, and of the shortcomings of available cropping tools.
Viapontica AI developed a single AI automated image cropping pipeline that automates aesthetic and text-aware image cropping. Its team worked closely with The List in order to gain an understanding of the key issues and most frequent edits made to images. With time and exposure to more image data, and feedback from users, the AI model has improved its accuracy and initial results show significant improvements on previous automated cropping tools and significant cost savings.
“It is really exciting every time we collaborative with the creative sector,” Viapontica AI’s Executive Director Vesko Cholakov says. Over the last few years, Viapontica AI has assisted a range of organisations in the UK, from high growth startups to government departments, to deliver a range of cutting-edge AI innovations, from the early workshop to the final deployment in production.
Cholakov says: “A lot of our work is mission critical and we put a massive focus on reliability and scalability, protecting data and respecting privacy. It is always exciting to step outside of the highly regulated industries in which we usually operate to work with an agile creative company like The List which has such a massive recognition in the art and entertainment circles and is taking bold steps to innovate.”
Viapontica AI and The List are now embarking on further discussions to take the idea forward, both within existing systems at The List and as a product for the wider market. As the cropping tool enables cost savings and automation in a sector that still often relies on manual efforts, there is significant potential for such a product. Working in partnership, Viapontica AI and The List are leveraging their engineering and commercial expertise and ever-expanding network of partners and clients. Viapontica AI has also identified several potential future extensions of the data processing pipeline to significantly add to the capabilities and flexible use of the tool.
In the course of this project Viapontica AI was able to forge a closer relationship with Creative Informatics, and become acquainted with the synergies between its own organization and the wider Challenge Project.
Brendan Miles, Director Data and Content, The List, says:
As The List grows its numbers of data clients, it must supply images in an ever wider range of sizes and ratios. With complex imagery received in many different sizes, formats and resolutions, currently available tools were not of sufficient quality to solve the problem to a high enough degree of satisfaction. The Creative Informatics project allowed us to partner with experts in AI image processing in Viapontica AI and develop new software that will help our business scale.
Vesko Cholakov, Executive Director, Viapontica AI, says:
The synergies rest at the intersection of the AI engineering experience which we bring as a private sector R&D partner, and the visibility which Creative Informatics has to challenges in the creative sector. The current project has established a strong foundation for further ideation and uncovering of potential opportunities now that we have been through a project together, understand better each other’s working style, strategic organizational objectives and ways in which we work best together.
If you have a business challenge in this sector, or if you want to explore how Edinburgh Innovations can support innovation in your organisation, please get in touch.
We make it easy to access to the University of Edinburgh’s multi-disciplinary expertise by matching your needs to the latest research, new technologies and world-class facilities.