DeepSqueak

DeepSqueak is a MATLAB application for bioacoustics researchers studying animal vocalizations, including ultrasonic calls from mice and rats. It uses neural networks to detect calls in audio recordings and provides a graphical interface for reviewing detections. The code is open source under the BSD-3-Clause license.

Researchers can move through an entire recording to adjust detected call regions or add boxes around calls the detector missed. That manual review also supports training: you can retrain an existing network with your own recordings, or annotate calls to build a detector for another species. This makes custom detection a substantial part of the tool's scope, alongside analysis of mouse and rat recordings.

Its detection architecture uses YOLO V2, a machine vision approach applied to audio spectrograms. DeepSqueak automatically removes horizontal noise bands and ties image scaling to the trained networks, so the detector's handling of spectrogram images is part of the analysis process.

For comparing detected calls, DeepSqueak includes a separate clustering interface and uses variational autoencoders for contour-invariant grouping. Researchers can review individual detections and explore groups of vocalizations within the same MATLAB application.

DeepSqueak runs locally in MATLAB 2020a or later and requires additional MATLAB toolboxes. Its BSD-licensed source code does not include licences for those proprietary dependencies.