Form the Nature Conservancy: More than 300 teams and individuals have joined a two-month competition to use artificial intelligence and machine learning to help New England fishermen provide accurate catch information in a cost-effective manner on groundfishing trips. The technologies developed in the competition will be applied to video coming from commercial fishermen in Maine, Massachusetts, New Hampshire, and Rhode Island who are using video cameras to meet their federal fisheries monitoring requirements. The fishermen are collaborating with The Nature Conservancy, the Gulf of Maine Research Institute, the Cape Cod Commercial Fishermen’s Alliance and the Maine Coast Fishermen’s Association to develop and test the video systems.
Participants in the “N+1 fish, N+2 fish” competition hosted by DrivenData will develop tools to analyze video footage to count, identify, and measure discards of undersized groundfish—such species as cod and haddock—as the fish are returned to the sea. Participants represent a global community of data experts, from professional data scientists to academic researchers to students and engineers looking to hone their algorithmic skills. The competition is only the second-ever fish image recognition competition worldwide and the first to use video and focus on United States fisheries. Entries are due by midnight on Oct. 30.
Competition organizers seek to spark advances in data science and image processing—innovative tools that are just beginning to be applied in the ocean. The goal: to make it possible to tackle fisheries management and conservation challenges that a decade ago would have required a prohibitive amount of staff time and computer analysis.
Video monitoring systems are now being tested on about 20 New England groundfish boats by participating fishermen interested in improving fisheries science with better catch data. Currently, the collected video is reviewed by trained observers which takes hours per fishing day to complete. Automating the video review can make it faster to get catch estimates and reduce costs so that every boat that needs to carry a camera system can afford one.
For boats that don’t use video, catch monitoring requires human observers to ride onboard during 16 percent of fishing trips—an approach that is costly for fishermen and taxpayers.
“Video collected at sea can be used to verify fishermen’s observations, and improve conservation and sustainable fisheries management,” said Chris McGuire, marine director for The Nature Conservancy in Massachusetts and part of the project team. “Applying machine learning to extract these data efficiently could be a game changer.”
This project builds on The Nature Conservancy’s data science competition to bring AI to Pacific tuna fisheries, which closed this spring.
The four-person project team includes experts in machine learning and in fisheries who joined together for the first time to take on this project.
This project is supported by grants from the National Fish and Wildlife Foundation, the Kingfisher Foundation, the Walton Family Foundation, and The Nature Conservancy.