The water surrounding the UK is teeming with interesting creatures. ‘Sea pens’, for example, are otherworldly structures that often resemble old-fashioned quill pens. They live in the mud on the ocean floor and are actually made from lots of smaller animals called polyps. Sea pens are also particularly sensitive to changes in their environment, which makes them a good potential indicator of the health of these muddy ecosystems.
To start building a picture of long-term changes in sea pen distribution and abundance, researchers at the government’s Centre for Environment, Fisheries and Aquaculture Science (Cefas) are looking back at over 15 years of video footage of the ocean floor off the north-east coast of England. This was collected by the research ship Cefas Endeavour, towing an underwater sledge fitted with cameras and lights. Sea pens can be identified manually in the footage, but with hundreds of hours of recordings to process, Cefas recognised the need for an automated method. So it engaged with the Turing via a Data Study Group (DSG) – our collaborative ‘hackathons’ that allow organisations to pose real-world challenges to multidisciplinary teams of researchers.
Over two weeks in December 2022, the DSG team developed a prototype machine learning algorithm that Cefas estimates can detect and classify sea pens with up to 95% accuracy compared to manual identification. The team also developed a second algorithm that enhances the image quality of the footage by automatically altering the contrast, frame by frame. This can be used on any video footage of murky marine environments and is currently being explored by other teams at Cefas, including a project called LobsterCAM that observes the underwater behaviour of lobsters.
“Studying the creatures living on the seafloor is essential to advancing our understanding of this fragile habitat. Our collaboration with the Turing has given us a potentially game-changing automated tool for this task, which we hope to deploy in support of new research in the near future.”
Anna Downie, Ecologist at Cefas and Challenge Owner of the Turing-Cefas Data Study Group
This piece first appeared in The Alan Turing Institute’s Annual Report 2022-23
Top image: vodolaz