Artificial intelligence to distinguish between individuals of other animal species

Estimating the size of the populations of each animal species is vital to determine if there is a risk of the species becoming extinct and what measures should be applied.

Some methods of estimating the population size of an animal species are based on the individual identification of animals. Traditionally, this has been done by physically marking captured animals. But this is a strategy that consumes many resources and is very laborious to implement.

It is increasingly common to study animals non-invasively, using cameras. However, animals of the same species often look alike, making it difficult to distinguish individuals even when there are photographs of them. Reidentifying animals from photographs is usually done through close observation by human specialists, which is expensive, laborious and requires considerable skill.

This is where deep learning (a form of artificial intelligence) can come in handy.

A team led by Emmanuel Kabuga, from the Center for Statistics on Ecology, Environment and Conservation (CSEEC), attached to the University of Cape Town in South Africa, has developed an automated method, based on deep learning, through which a computer equipped with This artificial intelligence tool can determine whether the same individual appears in a pair of photographs or not.

Given photographs of dolphins, whales, seals and toads, obtained from various databases, this artificial intelligence system successfully identified the individuals on a high percentage of occasions (between 83 percent and 96 percent), even when a few years had passed between when the first photo of the individual was taken and when the second was taken.

The new artificial intelligence system recognizes individuals among dolphins, among whales, among seals and among toads, and not necessarily by their face. (Image: Happywhale.org/Sea Mammal Research Unit, University of St Andrews/ToadNUTS)

It is clear that using artificial intelligence can improve estimates of the population size of a species and, ultimately, help adopt the most appropriate conservation and management strategies for each case.

Kabuga and his colleagues present the technical details of their strategy using artificial intelligence and the results obtained with it in the academic journal Ecosphere, under the title “Similarity learning networks uniquely identify individuals of four marine and terrestrial species.” (Fountain: NCYT by Amazings)