Les rencontres de STMS welcome Vincent Sirius Kather, a PhD student at the Naturalis Biodiversity Centre in Leiden. He will present his work,in english entitled « Deep Learning in Bioacoustics: Structuring large amounts of bioacoustic recordings in search for unknown species.

Abstract:

Bioacoustics describes the study of sounds produced by organisms. It has advantages over other forms of ecological monitoring due to its low interference with the subjects of interest. However, the large datasets created through passive acoustic monitoring demand computational processing strategies to handle the large volumes of data. Like in many other research domains, deep learning is impacting computational bioacoustics by enabling researchers to train classification models on large datasets and infer animal presence. My research interest lies in evaluating the feature spaces created by bioacoustic deep learning models and how they can guide us to address ecological questions of unknown species detection. Join my talk to get a brief introduction into computational bioacoustics and how current trends could help us identify previously unheard species and sounds.

Biography:

My name is Vincent Sirius Kather. I have studied mechanical and acoustic engineering at the Technical University of Berlin. My interest in computational bioacoustics began during my masters thesis, when I developed a deep learning classifier for humpback whales. I am currently doing a PhD at the Naturalis Biodiversity Center, which is a natural history museum in Leiden, The Netherlands. As part of my PhD I am doing a 6 month research visit at the Museum Nacionale d'Histoire Naturelle under the supervision of Sylvain Haupert. In my research I investigate how deep learning models trained on bioacoustic data can be used as feature extractors to find novel species.