FrogSong

What are we trying to do?

We are a research team focused on understanding frog calling patterns and behaviours. By monitoring frog calling patterns in long-term recordings we hope to understand: the environmental drivers of frog chorusing and other group behaviours, how frog calls and other noise sources interact to structure frog communities, as well as monitor how frog communities change through time in response to environmental stressors.

A common method for collecting data on animals that vocalise (e.g. frogs, birds) has been to take recordings in natural environments that can later be analysed to determine species presence and calling behaviour. Traditionally researchers have worked on shorter recordings (1-30 minutes) taken over relatively short time periods (weeks-months). Much of this has been dictated by storage and battery constraints of recorders. This also allowed most of the recordings to be analysed directly by listening to all or a large subset of the audio collected.

With improvements to the storage capabilities and battery-life of acoustic recorders, the amount of data researchers can collect has ballooned. While all this extra data will allow researchers to understand communities of vocalising animals on a scale not before possible, there is simply no way to manually listen to all the audio collected. Developing computer models to detect and classify calls in long recordings is one way for researchers to process large amounts of acoustic data.

Why do you need our help?

To train models to detect and classify animal calls researchers require very large labelled datasets (i.e. example audio with confirmed species presences), much larger than a small group of researchers can produce on their own. Your classifications of our audio data will help us collect the information necessary to train our models.