Flamingos Don’t Count Themselves: How Many Birds Are in That Noise?

Aysenur Arslan-Dogan is a Research Assistant in Data Science and AI at DACS, FSE, and an alumna of Maastricht University. Her work focuses on applying machine learning techniques to real-world problems, including bioacoustics and ecological monitoring

This talk introduces how modern AI systems can recognize bird species from audio recordings and explains why estimating the number of individuals is a much more complex task. Using real -world examples from zoo aviaries, it highlights the challenges of overlapping calls, silent individuals, and collective vocal behavior. Unlike the wild, where true population sizes are uncertain, zoos keep exact records of every individual. By recording the soundscapes of aviaries housing known numbers of birds, a dataset can be created where algorithms can be honestly evaluated against ground truth.

This talk explores the core challenges of counting birds from sound, focusing on complex acoustic environments like flamingo flocks. It discusses why standard detection approaches fail and how open-ended methods can push research forward in ecological.

Emre Argın is a second-year MSc student in Artificial Intelligence at DACS, FSE. His interests lie in machine learning, audio analysis, and developing innovative approaches to challenging real-world AI problems.