Image tool to help AI scour wildlife shots for climate change clues

This development could help scientists create new artificial intelligence algorithms that can perform rapid and in-depth analysis of the millions of wildlife images uploaded to the Internet every year.

Researchers say these could help reveal key insights into the impact of climate change, pollution, habitat loss and other stresses on tens of thousands of plant and animal species.

Citizen science websites are a potentially rich source of information about how plants and animals are responding to climate change. However, while existing artificial intelligence algorithms can automatically identify species in uploaded images, it is unclear whether they can also reveal other information.

Now, an international team of scientists has created a new tool to test how artificial intelligence algorithms can mine image libraries for additional information. This may include details such as what species are being eaten, how healthy they are, and which other species they interact with.

The tool, called INQUIRE, measures the ability of artificial intelligence to draw conclusions from an image library of 5 million wildlife photos uploaded to the iNaturalist citizen science website.

The research team found that current artificial intelligence algorithms can answer some of these types of questions, but not more complex ones. These include content that requires reasoning about small features in images and content that contains detailed scientific terminology.

The team said the findings highlight opportunities to develop new artificial intelligence algorithms that can better help scientists explore large image collections efficiently.

The peer-reviewed research results will be presented at NeurIPS, one of the leading conferences in the field of machine learning.

The team includes researchers from the University of Edinburgh, University College London, UMass Amherst, iNaturalist and the Massachusetts Institute of Technology (MIT). This work was partially supported by the Generative Artificial Intelligence Laboratory at the University of Edinburgh.

Dr Oisin Mac Aodha from the School of Information Science at the University of Edinburgh said: “Thousands of wildlife photos uploaded online every day provide scientists with valuable insights into where different species are found on Earth. However, understanding which species exist on Earth is just a matter of photos. The tip of the iceberg.

“These images can be an incredibly rich resource, but are largely untapped. Being able to quickly and accurately tease out the wealth of information they contain could provide important clues about how species are responding to challenges ranging from climate change to many others.” “

“This careful curation of data, with a focus on capturing real examples of scientific inquiry in the fields of ecology and environmental science research, has proven critical to expanding our understanding of current capabilities,” said Sarah Beery, Ph.D., assistant professor at MIT. important.

“It also outlines gaps in current research that we can now work to address, particularly for complex combinatorial queries, technical terminology, and fine-grained, subtle differences that delineate classes of interest to our collaborators.”