Aquila Ecology is constantly working to improve and expand its AI techniques. Recognising different types of bat sounds associated with different types of behaviour is essential for some studies. Therefore, we come up with the first model that distinguishes not only species but also echolocation, feeding buzz and social call for the 7 most common species. Moreover, it can distinguish multiple species in a single recording.
To create this AI model, we scoured and 'annotated' a lot of data: manually indicating what type of noise was involved. Then, as is often the case with Machine learning, it turned out that the algorithm was no longer optimal on this slightly different task and data set. Therefore, the algorithm was then tinkered with extensively and AI knobs were turned, resulting in a better-functioning model.
An example of a prediction from the model is shown below.
