ARTIFICIAL INTELLIGENCE TO IDENTIFY CETACEANS

Morigenos participated in the development of new artificial intelligence tool for improved cetacean research. We contributed data to the development of a new deep learning model for cetacean identification in a global international study published in the renowned scientific journal Methods in Ecology and Evolution.
The ability to identify individuals plays a key role in cetacean research. Knowing who is who among dolphins or whales allows us to, among other things, track their migrations, determine their social structure and, most importantly, know how many of them there are. Many cetacean species carry various natural markings on their bodies that can be used to distinguish them from each other – like a fingerprint or a face in humans. Many species of dolphin, including bottlenose dolphins in the northern Adriatic, can be identified by natural markings on their dorsal fins. Some other species, such as the humpback whale, can be identified by the natural patterns on the underside of the tail fluke. This method is called photo-identification, because researchers take photographs of the animals and use the natural markings to identify them.

Dorsal fins of dolphins from the Gulf of Trieste, by which Morigenos researchers identify individuals. In 20 years of research, we have photo-identified more than 400 individuals. Around 150 of them are permanent residents in the area.

A Morigenos researcher photographing dolphins for individual identification.

But carefully examining thousands and thousands of photographs is a painstaking and time-consuming task. Scientists need to make sure they have as large a set of relevant photographs as possible. At the same time, the identification of individuals must also be precise and accurate if the data are to be reliable and useful. In recent years, there have been major advances in the use of artificial intelligence to identify human faces and different animals individually, but new models usually need to be developed for each species. In this new study, 56 researchers from 6 continents joined forces to develop a new model for multi-species identification of cetaceans. The study involved 25 cetacean species from around the world, including bottlenose dolphins of the Gulf of Trieste, which Morigenos has been studying for more than 20 years.
»From a conservation standpoint it is really useful to be able to recognize the same individuals over time because you can see what areas the individuals use« said Philip Patton, a PhD student at the University of Hawaii and lead author of the study. »You can also use this information to estimate population size and population trends.«
The new model is based on human face recognition technology. It has proven to be very reliable, not just for one species, but for several different species of cetaceans.
»This new approach to cetacean identification will have an impact on the efficiency of photo-identification data processing on a global scale,« said Dr Tilen Genov of Morigenos, co-author of the study. »It will also help our work here in the northern Adriatic, as it will make our work more efficient and faster by automating the processing of the data.«

Humpback whales are identified mainly by natural patterns on their tail flukes.

This model will also be applicable to other species, both marine and terrestrial. The model is freely available as code on GitHub and as a graphical user interface on Happywhale.com.
Morigenos has been studying dolphins in the northern Adriatic since 2002, focusing on research on population size and distribution, behaviour, social and genetic structure, and the impact of human activities on dolphins, and has also been involved in the study of cetaceans elsewhere in the world.
The research article is freely available at: https://doi.org/10.1111/2041-210X.14167