competition
complete
$35,000

About the project


The Cook Inlet beluga population has declined by nearly 75% since 1979—from about 1,300 whales to less than 300 today. Concern about the high level of human impact on this population drove NOAA Fisheries, a U.S. federal agency within the National Oceanic Atmospheric Administration (NOAA), to designate Cook Inlet belugas as depleted under the U.S. Marine Mammal Protection Act in 1999. According to scientists, it is suspected that this population is still struggling to recover for a variety of reasons, including pollution, ocean noise, dredging, seismic surveys, oil exploration and drilling, and reduced prey. Cook Inlet borders Anchorage, the most populous city in the state of Alaska.

To gather accurate population data, the Marine Mammal Laboratory at the NOAA Alaska Fishery Science Center conducts aerial surveys to count groups of beluga whales as they collect around river mouths in the spring and summer months, using uncrewed aerial systems. This is process is called wildlife photo-identification, a noninvasive technique for identifying and tracking individuals of a wild animal population over time.

Survey images of Cook Inlet belugas are sent to a platform known as Flukebook for auto-detection and annotation. Flukebook applies computer vision and deep learning algorithms to detect and track individual whales. Following initial cropping and pre-processing, NOAA Fisheries then applies a manual photo-identification and matching procedure to attempt to identify each individual beluga in a photograph based on color, marks and scarring, and dorsal ridge features.

Whale Watching

Scientists use a variety of visual techniques to answer the fundamental question: "Which animal is this?"
Image Credit: Wild Me

New and improved computer vision and machine learning methods will help to expedite and improve the accuracy of this matching process, which is critical to the effective measurement, management, and protection of the Cook Inlet beluga population. The goal is to have a performant and scalable human–machine continual curation system for animal photo-identification, and solutions from this challenge will be used as the ranking component of such a system.

About the project team


The Bureau of Ocean Energy Management (BOEM) manages the development of U.S. Outer Continental Shelf energy and mineral resources in an environmentally and economically responsible way. BOEM focuses on understanding the potential impacts of offshore energy activities while minimizing impacts to marine mammals. To ensure that a species is not jeopardized or that a critical habitat is not adversely modified by BOEM-permitted activities, BOEM monitors marine ecosystems and population dynamics within these ecosystems.

NOAA Fisheries, also known as the National Marine Fisheries Service, is responsible for the stewardship of the nation’s ocean resources and their habitat. It provides scientific, ecosystem-based services to manage productive and sustainable fisheries, safe sources of seafood, recovery and conservation of protected resources, and healthy ecosystems. Under the Marine Mammal Protection Act and the Endangered Species Act, NOAA Fisheries works to recover protected marine species while allowing economic and recreational opportunities.

The Alaska Fisheries Science Center of NOAA Fisheries studies Alaska’s marine life to ensure the sustainable use of living marine resources in federal waters. It monitors fish and marine mammal populations that have supported Alaska Native communities for centuries and have provided food, income, and recreational enjoyment for millions of people around the world.

Wild Me develops open-source platforms for identifying and tracking wildlife, combining the strengths of artificial intelligence and citizen scientists to fight extinction. Using computer vision and deep learning algorithms, Wild Me created a platform called Wildbook, which scans millions of crowdsourced wildlife images at scale to identify species and individual animals to inform conservation decisions.

You can learn more here: