Bioacoustics Cookbook

This repository contains various "recipes" for bioacoustics workflows and problems (sometimes more broadly applicable to quantitative ecology). These recipes include tutorials, discussions, and demonstrations, often relying on our open-source bioacoustics package opensoundscape. They are written as Juptyer Notebooks to facilitate the integration of text, visuals, and code.

Machine Learning

• set up a model training project that uses a config.yml file to set hyperparameters and other options by copying the cnn_project_with_cnfig_yml/ subdirectory
• use the VGGish pre-trained audio feature extractor with OpenSoundscape with vggish.ipynb
• generate embeddings with an OpenSoundscape CNN object with cnn_embed.ipynb
• train shallow classifiers on pre-generated embeddings using sklearn with shallow_classifier.ipynb
• copy weights from an sklearn MLPClassifier to a torch/opensoundscape model with copy_shallow_classifier_sklearn_to_torch.ipynb

Classifier-guided listening

• Review and annotate audio clips based on classifier scores and metadata with classifier-guided_listening.ipynb

Data selection and review

• calculate sunrise and sunset times at a set of coordinate locations across a range of dates with sunset_sunrise_calculator.ipynb
• filter a set of AudioMoth files to include only specific date ranges and (start) time ranges with filter_files_by_dates_and_times.ipynb
• review grids of spectrograms using viwe_spec_grid.ipynb

GIS

• Convert between UTM and Lat/Lon coordinates, and create KML files from tables/csvs of waypoints, in gis/coordinates_and_kml.ipynb
• Create a grid of points as a KML file in gis/create_points_grid_kml.ipynb