ReefCloud: Diving into deep learning to protect coral reefs
Harnessing the power of human collaboration and artificial intelligence, ReefCloud allows the world's coral reef monitoring community to work together in real time to improve the monitoring, reporting and conservation of our reefs.
Find out how Accenture has been working with the Australian Institute of Marine Science and their partners across the Pacific to help protect the future of coral reefs.
CALL FOR CHANGE
Modernizing marine monitoring
Coral reefs are sometimes called the rainforests of the sea—a fitting nickname, as both are ecologically diverse, structurally complex and critically endangered. Scientists estimate that 75% of the globe’s coral reefs are severely threatened1 by climate change and chronic pollution. It’s a dismal data point that becomes a dire reality when you consider that reefs support at least 25% of marine species, not to mention the food source and livelihood for half a billion people worldwide.
But the scientific community hasn’t lost hope. Global conservationists are now able to conduct more regular in-water monitoring for the majority of major coral reef regions of the world thanks to modern camera technology. However, limited resources and inconsistent data formats have made it difficult to synthesize the millions of data points collected from the photos. Without a modern and efficient way to integrate global reef-monitoring efforts, conservation research and decision-making is incredibly slow, and data can easily become outdated. In fact, it currently takes up to four years to collate all that data—and with the climate situation changing so rapidly, this is time that the world’s reefs simply don’t have.
The Australian Institute of Marine Science (AIMS) is the country’s tropical marine research agency tasked with protecting and preserving Australia’s marine ecosystems. In 2019, the institute started a project to improve reef conservation. Around the same time, The Dock (Accenture’s global R&D and flagship innovation center located in Ireland) ran an environmental–themed AI hackathon. The winning idea? Using AI to monitor, analyze and protect coral reefs.
AIMS teamed up with The Dock to tackle the challenge together, kicking off a journey to transform how the scientific community approaches the conservation of this fragile marine ecosystem.
The team aligned to bring together the speed of cloud computing, the automation provided by AI and the decision-making capabilities of data analytics – all into a single solution built with scientists’ needs and uses in mind. Over the course of 18 months, they worked with the Queensland University of Technology (QUT), coral reef scientists, conservation practitioners and technology experts across various organizations to design an open access tool called ReefCloud.
ReefCloud is a cloud-based web platform hosted on Amazon Web Services that uses computer vision to analyze reef photos, and automatically extract, annotate and share data points like coral cover or reef composition. After analyzing a photo, the platform automatically adds the data to a dashboard, designed with a user-friendly interface. The dashboard translates the data into valuable insights and recommendations so conservationists can quickly understand what the data says about reef and coral health. And perhaps most importantly, the platform enables democratized knowledge. It standardizes the data for rapid interpretation, reporting and communication on reef conditions across languages, geographies and scientific methodologies.
The tool’s AI model was trained using data from AIMS’ own reef-monitoring program, which contains thousands of images and data points from the Great Barrier Reef. But the computer vision model was just one part of the equation—ReefCloud also needs people.
ReefCloud is built for collaboration. It allows the world’s reef-monitoring community to upload photos and work together in real time. And the algorithm learns with each new image added—so the more people who use ReefCloud, the bigger the impact it can have.
Diving into deep learning to protect coral reefs
In this blog post, Accenture’s Melanie Angerer describes how her team is applying machine learning to protect coral reefs.
ReefCloud has already improved the efficiency of conservation work, producing estimates of coral reef composition with 80%-90% accuracy at a speed 700 times faster than manual efforts. Because it accomplishes in seconds what used to take reef scientists 15 minutes, the platform has freed up months of skilled labor and precious reef management resources. This means marine scientists can now focus on influencing policy instead of spending time wading through floods of data.
Now with 200+ users across 24 countries, ReefCloud is enabling scientists to share data and collaborate faster, strengthening global conservation efforts. Collaborators from around the world (including Australia, Palau, Fiji, Solomon Islands, Vanuatu and the Maldives) have shared a collective 1.3 million images.
But ReefCloud is not just about reefs. The conservation community can scale the technology across other ecosystems like mangroves, seagrass and even national parks. It fully demonstrates the unlimited potential of technology, design and expert knowledge. Which, when fused together, can transform how we solve the world’s greatest challenges—even those at the bottom of the sea.
ReefCloud is a “partnership that’s bringing science and AI together with indigenous knowledge to [...] inform future policy making.”
— Jamie Isbister, Australian Ambassador of Environment