A birdsong recognition AI algorithm is now running live online, in Xeno-canto, the world's largest open access collection of animal sounds. When "mystery" bird recordings are uploaded from anywhere in Europe, the algorithm - developed at Naturalis - gives suggestions of which European bird is likely.
Bird soundsonline
The algorithm is developed by Naturalis researcher Dr. Burooj Ghani as part of the "GUARDEN" project. It is now accessible online via Xeno-canto. A website known to many birdwatchers, but not to the wider world, Xeno-canto has since 2006 been the place to share bird sounds as open data. This makes it an important source of recordings for scientific research, and for acoustic AI as well. If you have an app on your phone that recognizes bird song, its AI is powered by Xeno-canto.
Openand European
AvesEcho
What’s different about this new AI, called AvesEcho, is that it is specialised in European birds. It is also completely open: anyone can review the code and use it. Also, it is integrated into Xeno-canto, and will be made public through the ARISE research infrastructure in the future. Ghani: "We developed this new birdsong recognition algorithm by investigating many of the latest deep learning techniques. These methods are powerful enough to be applied worldwide. For this project we tailored the system to cover all the bird species you are likely to hear in Europe."
Share withvolunteers
Naturalis develops AI methods for sounds, images, DNA and more. The audio research is led by Dr. Dan Stowell, Associate Professor in AI and Biodiversity. Stowell said: "I'm excited to put our research into the hands of the many volunteers who record bird sounds and share them freely through the Xeno-canto collection. We want to use these tools to monitor biodiversity, but also to help people engage with nature. I hope more people can be fascinated by the animal sounds we hear every day."
As next steps in their research, the team will develop methods to recognise all animal sounds on land - not just birds! They also work on methods to analyse finer details, such as automatically identifying call types and individual signatures.
Moreinformation
- Read more about this on the Xeno-Canto website.
- Check out our AI for Nature page if you want to know more about how we develop Artificial Intelligence for species identification, and how this might help your work.