AI for Wildlife? A sound approach

Tilburg University's Department of Cognitive Science and AI has a strong emphasis on AI for social issues such as sustainability. A new research line in the department is AI for Nature and Environment, led by Dan Stowell, Associate Professor for AI and Biodiversity.

This includes a deep collaboration with Naturalis Biodiversity Center, with whom Stowell is developing AI methods for automatically monitoring wildlife across the whole of Europe - through images and now also through sound. By enabling ecologists and citizens to identify and monitor animals in this way, we aim to help society to maintain the sustainability of our natural environment.

Recent examples of this work:

• A funded Lorentz Centre workshop which attracted a full house of bioacoustics researchers in August 2022, spending a week developing methods for applying AI to natural sound recordings (you can read the workshop report);
• A published paper by MSc CSAI student Alberto Garcia Arroba Parrilla, presented at a conference in France, evaluating how sound recognition systems could work when many birds sing at the same time (read the paper online);
• A thesis by MSc DSS student Dario Gandini on classifying insect sounds despite having only small datasets (read the thesis online).
• A survey paper by Stowell covering the state of the art and future directions in computational bioacoustics using deep learning.

AI for wildlife also inspires industrial collaborations. Stowell's group is collaborating with Heijmans BV on acoustic analysis in construction sites; Iconize BV on acoustic monitoring of cattle farms and bee hives; and Vaart BV on fish image recognition. The group also works with the Jheronimus Academy of Data Science, and national nature organisations.

Through the "AI4Bio" seminar series, staff and students learn about AI for wildlife from international experts. Recent guests include Sara Beery (Google/MIT) and Daniele Silvestro (University of Fribourg).