Example fixtures
Maintained audio fixtures
mouse_1.wav and mouse_2.wav and their matching *_outputs/ directories are the maintained example fixtures. They serve as the ground truth for CI baseline smoke checks — any time the pipeline runs, results are compared against these stored outputs to catch regressions.
Each fixture includes:
Detection and classification outputs (CSV,
.vocalpyfiles)Autoencoder-based segmentation masks pre-generated by SqueakOut, stored under
cnn_mask/
Key paths:
Manifest:
examples/audios/baselines.yml— defines which fixtures exist and what counts as a passing comparisonRefresh:
python scripts/rewrite_example_vocalpy_fixtures.py— regenerates the serialized output files when expected outputs intentionally changeSegmentation masks:
examples/audios/mouse_1_outputs/cnn_mask/andexamples/audios/mouse_2_outputs/cnn_mask/
Maintained notebooks
These notebooks are supported and kept up to date. They demonstrate the main analysis features using the bundled fixtures and models.
class_classifier_demo.ipynb— loads bundled spectrogram images and runs the type classifier, showing per-call class predictions and probability distributionsnoise_classifier_demo.ipynb— demonstrates noise vs. vocal filtering, showing which candidates get removed and which are keptvisualization_plots_demo.ipynb— produces group-level visualization plots using themouse_1andmouse_2fixtures; useful for understanding theVizAPI
To run any notebook:
pip install -r requirements-dev.txt # includes jupyterlab
jupyter lab examples/
Archived notebooks
These notebooks are kept for historical context only. They are not maintained, may not run against the current API, and are not part of the supported surface:
candidate_vocalization_identifier_demo.ipynbunsupervised_embedding.ipynbunsupervised_embedding-Copy1.ipynb
Legacy demo data
examples/audios/example_outputs/ is legacy image-only classifier demo data from an earlier version of the pipeline. It is not registered in the CI baseline manifest and does not represent canonical pipeline output. Ignore it for any current analysis.