Skip to content 
Imagine turning hours—or even months—of underwater footage into actionable insights in minutes. That’s what SeaFrame, an AI-powered computer vision software, aims to achieve.
For fishers, this means pinpointing critical moments like gear movement or catch behaviour without sifting through endless video. For scientists, it enables faster, smarter biodiversity assessments. SeaFrame harnesses artificial intelligence to revolutionise how we understand and manage the ocean, making data analysis faster, easier, and more effective.
Funded by Innovate UK, this five-month project brought together the latest technology and real-world challenges to develop a system that simplifies underwater monitoring. SeaFrame demonstrates how AI can transform how we process underwater data, solving marine challenges such as species detection, biodiversity assessment, and gear performance analysis.
Turning Hours of Footage into Clear Ocean Data
Underwater cameras, like our CatchCam, are invaluable for fishing and marine research, but the large amount of footage they generate can be overwhelming.
For fishing vessels, CatchCam footage offers a deeper understanding of how their fishing gear behaves below the waves, helping troubleshoot issues related to accidental bycatch and seabed impact. Fishers often record over 40 hours of footage per trip, yet only a small portion contains actionable data.
Marine researchers face even larger datasets. While these cameras are invaluable for monitoring ocean biodiversity, less than 1% of footage typically captures significant activity.
SeaFrame changes this by automating and speeding up the review process. But how does it work?
Machine Learning and AI for Marine Monitoring
The key to SeaFrame’s success is the extensive library of training videos. The software has been trained on hours of footage recorded by the CatchCam camera, collected from deployments all around the world. All this footage provides a rich and diverse dataset, with real-world environments and ocean conditions.
By analysing these scenarios, SeaFrame’s algorithm is finely tuned to detect events like fish swimming, gear movement, or lighting changes. Even in challenging conditions, such as murky waters or chaotic underwater environments, this training ensures accurate and reliable results tailored to practical applications.
Using machine learning and deep learning, SeaFrame rapidly analyses large datasets. It generates an ‘interestingness’ graph, scoring video sections based on activity. This approach automatically highlights key moments while skipping over hours of uneventful footage. That way, you can focus on decision-making, instead of reviewing endless hours of footage.
Accurate Detection in Complex Environments
One of SeaFrame’s standout features is its ability to detect trawl contact with the seabed—an essential insight for minimising environmental impact. This was achieved by training the software with user-generated CatchCam footage, recorded by fishers deploying cameras on their trawl gear. These practical inputs have enhanced SeaFrame’s accuracy and effectiveness for industry use.
The software performs well even with low-quality video. In fact, frame rates as low as 1fps or recorded in turbid waters, can still be analysed by SeaFrame. So, whether its moving fish species, underwater objects, gear movement, ground contact, or changes in lighting – they can all be detected.
By addressing real-world scenarios, SeaFrame showcases its potential as a reliable and versatile tool for both sustainable fishing practices and marine ecosystem monitoring.
Collaboration for Innovation
SeaFrame is a great example of what we can achieve through collaborative projects. It also shows the power of AI in advancing sustainable fishing and marine research.
We’re proud of the work our team and partners—Aran Dasan, Tom Rossiter, Chris Lewis, Dylan Van Bramer, Pranali Dhane, Lachlan Stibbard Hawkes, Sarah Ready and Darren McClements—put into this project.
Although the SeaFrame project has ended, the journey doesn’t stop here. At CatchCam Technologies, we develop tailored tools for the fishing and marine research industries. From integrating AI to addressing specific operational needs, we’re here to help.
SeaFrame shows how targeted projects can solve real problems. It’s a great example of how we tackle challenges, one project at a time.
For more updates on our work or to discuss your project needs, contact our team here.
