AI-supported bird call recognition with BirdNET
For acoustic biomonitoring – the identification of bird species through sound recordings – the latest developments in data analysis with artificial intelligence brings great advantages. Whereas previously all recordings had to be painstakingly analysed by bird call experts, it is now at least possible to gain an initial impression through automated recognition. Conspicuous results always need to be checked by experts, but for a rough assessment or as a quick indicator, AI brings great advantages for research and nature conservation.
In the Dawn Chorus app, we use bird call recognition from the BirdNET app.
BirdNET is an app from Chemnitz University of Technology and the Cornell Lab for Ornithology that uses machine learning to recognise and classify bird calls and provides an innovative tool for nature conservation, biologists and birdwatchers.
This algorithm is based on the algorithm of the BirdNET app and can already recognise around 6000 bird species based on their songs and calls. At least in theory; unfortunately, it is not yet error-free. Especially in the polyphonic Dawn Chorus, when many bird species sing “in a jumble”. Bird calls are also very varied and different; there are bird species whose songs are very easy to recognise (even for the AI) – and there are bird songs that are very difficult to identify (especially for the AI).
Are you an expert on bird calls?
Help us improve the Dawn Chorus AI!
To improve the AI in the long term, we are working with volunteer bird call experts to annotate the data – your Dawn Chorus recordings. Volunteers are welcome to contact us at: vogelstimmen@lbv.de. (If you would like to test your knowledge, you can try out our annotation quiz: https://annotate.dawn-chorus.org/quiz.)