ARIEL: Arthur Rylah Institute Ecoacoustic Listener

Passive acoustic recorders have become an important tool for field ecologists as they provide considerable cost efficiencies and provide the ability to sample many locations simultaneously. Recorders can generate large volumes of acoustic data which can be slow to review for the identification of target species even for species experts. The task of identifying fauna from acoustic data is readily suited to the use of artificial intelligence analysis methods, however the performance of these methods requires large amounts of validated and labelled training data that can be used to develop the AI models. Validating and labelling the training data for AI models can be a laborious task that is essential before any analysis can take place. To facilitate the workflow of AI analytics we have developed custom software for the rapid validation and labelling of acoustic data. The software is fully customisable and presents users with audio and visual cues for the rapid labelling of acoustic data, and automates the validation process while retaining ancillary information (e.g. site, survey etc). By using the common data format of .csv files, this allows the software to be flexibly embedded into most existing ecoacoustic workflows as well providing a format to readily share validated species detections. The software also supports the rapid validation of predictive model outputs therefore supports complete analytic workflow for full reporting from large acoustic datasets.