NEWS & EVENTS
Unlocking Biodiversity Insights: How GUARDEN EU Research Benefits the Xeno-Canto Community
The GUARDEN EU project has brought a transformative wave of innovation to Xeno-Canto (XC), a renowned global platform for sharing bird sound recordings. For nearly two decades, XC has been a vital hub for collecting and curating avian sounds. Now, through advancements in machine learning (ML) powered by GUARDEN and related Horizon Europe projects, XC is taking a significant step forward in integrating cutting-edge automatic sound recognition.
At the heart of this work is AvesEcho v1, an ML algorithm designed to identify European bird species from audio recordings. Developed by the Naturalis Biodiversity Center, AvesEcho v1 builds upon BirdNET, a sophisticated model created by the Cornell Lab of Ornithology. What sets AvesEcho v1 apart is its fine-tuning with European data derived primarily from XC itself. This integration underscores XC’s dual role: As a repository of invaluable data and as a driving force behind technological advances in bioacoustics.
The implementation of AvesEcho v1 addresses a longstanding XC ambition—seamless and meaningful ML integration. Currently, the algorithm analyzes “mystery recordings” uploaded by XC users, offering a list of potential species present. These suggestions assist recordists in labeling species more efficiently and accurately, enhancing the overall quality of the XC database. While the ML suggestions remain advisory, they provide an exciting starting point for identifying species in complex soundscapes.
The GUARDEN project’s contribution goes beyond tool development; it exemplifies the potential of collaborative research to tackle limitations in ML models. For instance, a persistent challenge in bioacoustics is “weakly labeled” data—recordings labeled for entire audio files rather than specific segments. GUARDEN highlights the importance of generating “stronger labels,” where bird vocalizations are meticulously annotated with start and stop times, sound types, and frequency ranges.
For XC’s vibrant community of bird enthusiasts and researchers, these advancements signal a new era. Automatic sound recognition not only accelerates the process of identifying species but also enriches the platform’s role in biodiversity monitoring. As the GUARDEN project bridges science and citizen engagement, it reaffirms the value of open, collaborative initiatives in addressing ecological challenges. The journey of sound recognition may have just begun, but with projects like GUARDEN leading the way, the future of biodiversity research sounds more promising than ever.