Conservation Tech: Monitoring Humpback Whales Using Image Recognition

Mark Altaweel

Updated:

As artificial intelligence gains increasing traction in our society, it has also increasingly been applied in areas such as conservation and research. Conservation technology is now being used not only by professional organizations and scientists but also by ordinary people and advocacy groups to monitor and safeguard specific species.

The hope is that artificial intelligence (AI) and similar technologies will aid in countering the ongoing ecological decline impacting our most endangered species, with initiatives like the application of AI to recognize and monitor humpback whales serving as one of the latest examples.

Tracking a global population of whales using AI

Humpback whales have a range that covers all of the world’s oceans, making individuals sometimes difficult to track over the course of their lifetimes. Automated image recognition software potentially enables the identification of objects from almost any type of picture, including normal optical camera pictures.

Using citizen science and AI to track humpback whales

Happywhale is a citizen science group set up to understand and care for marine environments through conservation and education.[1] The aim of the group is to enable the public to participate in this conservation using technology and the power of AI.

Screenshot showing the HappyWhale website with pictures of whale tails.
Screenshot from the Happywhale website explaining how their whale tracking works.

The website’s fundamental feature allows users to upload photos, which are then analyzed by an automated image recognition system designed to identify specific body features of whales. This tool compares the uploaded image with a database of previously submitted photos to determine if the whale has been documented before.

Using this technique, over 30,100 individual humpback whales in the North Pacific Ocean basin have been identified. Overall, 70,500 individual whales having been sited over the lifetime of the effort. As the North Pacific has far more sightings in general, the humpback whales in this area have been sighted about six times on average.

How WhaleID works

The tool, called WhaleID, looks for key features particularly in the tails (flukes) of whales, which distinguish one whale from another. Matches of nearly 100% indicate a match and then using geotagged data from images or GPS enables the spotted location to be mapped. The individual whale can then be tracked through repeated uploading of images so that a network showing the range and locations that individual whales have traveled to throughout their life can be built.

Benefits of tracking whales

This technology benefits not only conservation efforts but also enables scientists to better study humpbacks, which are naturally shy animals. Scientists now do not need to approach the whales closely and can simply monitor them using this approach, which enables the whales to be less disturbed.

One example of the benefits from compiled this database is that scientists have learned that juvenile whales are far more likely to die from entanglements in fishing nets. This has led to calls to limit the use of large fishing nets spread over wide areas by fishing vessels. Of course, not all whales are seen multiple times, so many are simply kept as point data until a future observation is made.

The website also shows statistics of uploaded images from individuals or groups contributing, giving credit to those who contribute to the conservation and scientific effort. Individual whale’s statistics can be seen, including the number of times it has been spotted. One can see which whale was spotted the most or even the longest period when an individual whale was spotted between pictures (43.8 years).

The website takes financial contributions and enables users to name whales as a way to sustain and encourage the effort. Overall, this is just one of many increasing websites using citizen science approaches to promote research and conservation.[2]

Using WhaledID to track other marine mammals

The success of WhaleID and Happywhale has led to calls for efforts to scale the effort to other marine mammals. In fact, Happywhale does collect data for other marine mammals such as bottlenose dolphins who can be identified through their faces.

The Happywhale project greatly benefited from individual research efforts boosting the number of photos taken between 2004-2006 and using some historic photos before digital cameras. This may not be as easy for marine mammals less studied, but given the technology works by simply focusing on parts of bodies that help identify animals, such as tails, the approach can similarly be applied by focusing efforts on unique identifying features of marine mammals.[3]

A non-invasive way to study individual marine animals

Using data science and AI techniques to identify and track marine mammals has great potential for enabling non-invasive techniques to track many vulnerable species. By empowering scientists and non-scientists to contribute, the public can feel more attached and involved in conservation efforts, which can aid in advocating for efforts to minimize harm such as calls for greater care in fishing that harms humpback populations. Scientists and others can also build on such efforts and expand to marine and other creatures potential using such technology so that there is better tracking of vulnerable animals and aid in their conservation.

References

[1]    The Happywhale website can be accessed here:  https://happywhale.com/.

[2]    A story on Happywhale can be seen here:  https://wildlabs.net/article/happywhale-ai-powered-whale-identification.

[3]    For more on WhaleID statistics and how it has been built up over the years, see:  Cheeseman, T., Southerland, K., Acebes, J.M. et al. A collaborative and near-comprehensive North Pacific humpback whale photo-ID dataset. Sci Rep 13, 10237 (2023). https://doi.org/10.1038/s41598-023-36928-1

Related

Photo of author
About the author
Mark Altaweel
Mark Altaweel is a Reader in Near Eastern Archaeology at the Institute of Archaeology, University College London, having held previous appointments and joint appointments at the University of Chicago, University of Alaska, and Argonne National Laboratory. Mark has an undergraduate degree in Anthropology and Masters and PhD degrees from the University of Chicago’s Department of Near Eastern Languages and Civilizations.