A Switzerland-based research team has developed a new method to track and analyse animal behaviour using Artificial Intelligence. The technology is currently being tested at the Zurich Zoo.
Can a machine sense when an animal is happy, curious, or afraid? Researchers studying animal behaviour typically rely on hours of video footage, meticulously observing and documenting animal movements and interactions. A team at the Federal Institute of Technology Zurich (ETH) and the University of Zurich have now come up with an automated way to analyse these recordings.
The image-analysis algorithm they developed uses computer vision and machine learning to distinguish individual animals and identify specific behaviours. This includes behaviours that signal curiosity, fear, or harmonious social interactions with other members of their species.
The algorithm is highly sensitive, enabling it to identify subtle behavioural changes that develop very gradually over long periods of time. “Those are the kinds of changes that are often tricky to spot with the human eye,” said Markus Marks, a postdoc researcher at ETH and lead author of a study applying the technology in primates and mice. The results of the study were published in Nature Machine IntelligenceExternal link in April.
The method is expected to cut down on the time spent watching video footage and make it easier to compare results because researchers using the algorithm would rely on the same standard. The researchers have made the algorithm available on a public platform so that other researchers can use it. It’s already being tested by the Zurich Zoo, which is using it for animal husbandry and behavioural research. Researchers are also using it to study chimpanzees in the wild in Uganda.
While there are other machine-learning algorithms to track animal behaviour, Marks says that this technology is unique in the way it captures “social behaviour in complex settings”. In addition to enhancing basic understanding of animal behaviour, the tool can help improve animal welfare by detecting abnormal behaviours early on that could be signs of disease onset or danger.
In compliance with the JTI standards