The collection and annotation of bioacoustic data presents a number of challenges to researchers, often constraining analysis to highly vocal species. Computational tools allow monitoring to be extended to rare and cryptic species but data limitations remain an issue. We present a human-in-the-loop approach that combines the efficiency of computational tools with the accuracy of human analysis. This methodology uses a prototypical few-shot active learning approach that only requires a single example per class to provide analysis recommendations to a user. We present an ablation study evaluating segmentation and classification approaches using an invasive species detection case study to analyse field data of both rare and cryptic common brushtail possum (Trichosurus vulpecula) vocalisations. This approach was able to achieve 98.8% validation accuracy on a binary classification task using a limited dataset of 200 5-minute recordings and 81.3% validation accuracy for 2-shot 2-way learning before fine-tuning. We implement this methodology into a publicly available web application https://github.com/Listening-Lab/Annotator.

This content is only available via PDF.