Learn how our Blue w(AI)ve Accelerator program supported SeaDeep, which develops AI to map, monitor, and explore the ocean, in this case study produced by our partners at Northeastern Institute's AI Solutions Hub.
The biggest obstacle preventing many companies from building powerful AI tools is messy, unstructured data.
The problem exists in every industry, but perhaps nowhere is it more acute than in our vast, underexplored oceans, where imaging and sensor data are crucial for understanding complex, interconnected ecosystems over large swaths of the planet.
SeaDeep is using AI to help energy companies derive insights from that data. The company is using advanced AI techniques to map, monitor, and explore our oceans and underwater infrastructure from raw data.
Photo Credit: Austin Bumpus, Institute for Experiential AI at Northeastern University.
By the end of the program, SeaDeep had developed an unsupervised learning model for sorting and annotating raw hyperspectral data, a type of imaging data that includes information from across the electromagnetic spectrum to make it easier to detect objects and identify materials.
The tool will help SeaDeep revolutionize subsea exploration and monitoring.
The underwater environment is so complex, and there are a lot of factors that affect the data: how much sunlight is reaching the area you're imaging, what flashlight you’re using, the turbidity of the water. In order for your AI to scale across the globe in different ocean environments, you have to collect and process a lot of different kinds of data.
Shishir RaoCo-Founder and CTO, SeaDeep
Accelerating Ocean Tech
The Blue w(AI)ve program was created to drive innovation in ocean technology. It supported seven early-stage companies leveraging artificial intelligence in ocean-based solutions.
GMRI designed and launched the program, providing each company with a scientists-in-residence and investors-in-residence who provided deep industry knowledge along with access to blue economy research, expertise, and networks.
GMRI also provided weekly workshops to each company to compliment the consultations and build on their work.
We recruited from a stellar international group of startups who all had a vested interest in the Gulf of Maine and the unique resources we offer to help move their companies to the next phase of growth. With each company, we experienced a pull toward the AI projects, and also the ability to gain access to scientists, sector-specific investors, industry, and the opportunity to engage with ecosystem partners and test products and services on the water.
Lucy BarrettSenior Manager, Innovation and Entrepreneurship, GMRI
Experts from Northeastern’s AI Solutions Hub (AISH) contributed up to $25,000 in consulting projects as part of Blue w(AI)ve’s AI-in-residence program. Companies also received guidance from investors-in-residence from Bold Ocean Ventures and partner venture funds.
The Blue w(AI)ve program showcases how Northeastern University's Roux Institute and Institute for Experiential AI can partner with high-growth startups to solve climate challenges. The significant progress made during the program was a direct result of the AI development projects facilitated by the AI Solutions Hub.
Benjamin CheslerDirector of Venture Creation and Acceleration, Roux Institute
The work was nothing new for the AI Solutions Hub, which works with organizations of all sizes to modernize their AI strategies and build transformational, AI-powered solutions. But it did give AISH experts an opportunity to apply their skills to our planet’s vast, underexplored oceans.
The Blue w(AI)ve program exemplifies the power of partnership with internal and external teams, showcasing Northeastern’s commitment to core innovation. It blends state-of-the-art AI techniques, business strategies, and sustainability practices to develop next-generation marine technology, positioning Maine at the forefront of this Blue Economy w(AI)ve (pun intended).
James Shanahan, Ph.D.Director of AI Solutions Hub
Photo Credit: Austin Bumpus, Institute for Experiential AI at Northeastern University.
SeaDeep’s challenge
Large energy companies have collected millions of hours of underwater footage of offshore infrastructure like oil and gas rigs. Since its founding in 2020, SeaDeep has helped those companies gain insights about their infrastructure, addressing issues like corrosion and marine growth.
The company’s AI tools can improve imaging data by restoring color, improving lighting, removing marine snow, and more. It can then perform object detection, make characterizations of the seafloor, and collect other information to create visualizations and reports.
When people talk about AI challenges, they often talk about architectural challenges, but one of the biggest challenges is access to quality data. Without quality data, you can't build powerful AI solutions.
Shishir RaoCo-Founder and CTO, SeaDeep
Most of SeaDeep’s work so far has dealt with vision data, but the company wanted to aggregate other kinds of data from things like sonar and hyperspectral cameras. For that data to be useful for the company’s AI models, it needed to be categorized and labeled. To do that at scale, SeaDeep needed to leverage AI.
A lot of the AI that you see deployed now are trained using ground truth data, or there is an easy way to create ground truth data. But that isn’t true in the underwater environment.
Shishir RaoCo-Founder and CTO, SeaDeep
Characterizing Data at Scale
Blue w(AI)ve’s experts, including Director of AISH James Shanahan and AISH Director of Strategic Research Projects Ardeshir Contractor, collaborated closely with SeaDeep’s team, providing guidance on AI tool development.
It was a constant back and forth of us talking with them and trying to think about the challenges we are facing and how we can come up with solutions. They thought deeply about our challenges and what we could achieve with AI. It was really hands-on and a great experience.
Shishir RaoCo-Founder and CTO, SeaDeep
Focusing on hyperspectral data, which collects and processes information from across the electromagnetic spectrum, AISH experts worked with SeaDeep to develop an unsupervised learning model that allowed the company to automatically label clusters of raw hyperspectral data that offered information about the objects and conditions around offshore infrastructure. For instance, one cluster was automatically associated with coral, another sand.
We started by thinking about unsupervised learning with respect to hyperspectral images: what's out there, and what's the state of the art? Then we considered the solutions SeaDeep had already developed, and we combined those things to make something novel.
Shishir RaoCo-Founder and CTO, SeaDeep
Rao describes the tool as a way to accelerate the creation of ground truth data that’s so vital for training AI models.
The AISH team helped us accelerate our ability to ground truth the data, because you can imagine if we had to manually go and label every single pixel and say 'this is sand, this is coral,' that's a very daunting task.
Shishir RaoCo-Founder and CTO, SeaDeep
A Growing Collaboration
SeaDeep’s team was thrilled with the progress they made in the Blue w(AI)ve program and plans to continue its collaboration with the Roux Institute and the Institute for Experiential AI by applying to research grants with the organizations.
Northeastern has extensive resources, and being able to pull from all of those resources for such a complex task was huge. The team they can bring to the table, along with the product development that SeaDeep can bring, paved the way for a natural collaboration.
Shishir RaoCo-Founder and CTO, SeaDeep
Rao credits the Blue w(AI)ve’s team with helping SeaDeep think more broadly about where the company’s technology could solve problems, which is what led to the ongoing grant application.
When you're only working with your clients on one problem, you tend to get tunnel vision. The early work discussing what we could build together gave us an opportunity to think about adjacent applications to what we're currently working on and how we can leverage Northeastern's expertise to build something really important. That is what working with an external group brought to the table.
In early 2024, we announced a new partnership between Gulf of Maine Ventures and The Roux Institute at Northeastern University to drive blue technology innovation.
