A group of experts in artificial intelligence (AI) and animal ecology from the Ecole Polytechnique Fédérale de Lausanne has developed a new big data approach to improve wildlife research and improve wildlife preservation.
The new study was published in Nature Communication.
Wildlife data collection
The field of animal ecology is now driven by big data and the Internet of Things, with massive amounts of data collected on wild animal populations through technologies such as satellites, drones and automatic cameras. These new technologies are driving faster research developments while minimizing disturbance to natural habitats.
Many AI programs are used to analyze large datasets, but they are often general and not precise enough to observe the behavior and appearance of wild animals.
The team of scientists developed a new approach to circumvent this problem, and they did so by combining advances in computer vision with the expertise of ecologists.
Leveraging the Expertise of Ecologists
Conservationists are currently using AI and computer vision to extract key features from images, videos and other visual forms of data, allowing them to perform tasks such as classifying wildlife species and counting. individual animals. However, the generic programs that are often used to process these data are limited in their ability to take advantage of existing animal knowledge. They are also difficult to personalize and are subject to ethical issues related to sensitive data.
Professor Devis Tuia is the director of EPFL’s Laboratory for Computational Environmental Sciences and Earth Observation and the lead author of the study.
“We wanted to interest more researchers in this subject and pool their efforts to move forward in this emerging field. AI can serve as a key enabler in wildlife research and environmental protection more broadly,” says Professor Tuia.
In order to reduce the margin of error of an AI program trained to recognize a specific species, computer scientists should be able to take advantage of the knowledge of animal ecologists.
Professor Mackenzie Mathis is the director of the Bertarelli Foundation Chair in Integrative Neuroscience at EPFL and co-author of the study.
“This is where the fusion of ecology and machine learning is key: the field biologist has immense knowledge about the animals being studied, and we as machine learning researchers need to work with them. to create tools to find a solution,” she said. .
This is not the first time that Tuia and the research team have tackled this question. The team previously developed an animal species recognition program based on drone images, while Mathis and his team developed an open-source software package to help scientists estimate and track animal poses.
As for the new work, the team hopes it can capture a wider audience.
“A community is gradually taking shape,” says Tuia. “Until now, we have used word of mouth to build a first network. We started two years ago with the people who are now the other main authors of the article: Benjamin Kellenberger, also at EPFL; Sara Beery of Caltech in the United States; and Blair Costelloe at the Max Planck Institute in Germany.