Merlin Sound ID: Challenges in Real-time Analysis of Bird Vocalizations

The Merlin Bird ID app integrates Merlin Sound ID, a sophisticated system designed for real-time classification of bird vocalizations on mobile devices. With a global user base numbering in the millions, we’re not only achieving our initial goal of connecting people with their neighborhood birds but also extending the utility of Sound ID as a versatile monitoring tool. This presentation will start with an overview of the Merlin Sound ID architecture and functionality. I will then delve into the technical challenges surrounding noisy annotations and how we use them for training and evaluation. Finally I will explore our benchmarking procedures and ongoing efforts to enhance performance. This talk aims to shed light on the challenges and solutions associated with deploying real-time auditory classification systems in ecological applications.