CLUSTERING UNSELECTED BIO-ACOUSTIC RECORDINGS

SUMMARY:

• 🚧 Still under development 🚧
• Spectrogram clusters were obtained via auto-encoders, dimensionality reduction and unsupervised clustering.
• This page currently only features the last step, clustering with DBSCAN.
• Once ready, this tool could automatically create pre-labelled dataset.
• This could speed-up the annotation of datasets to train supervised models.

DEPENDENCIES:

• Developed under Python 3.12.8
• First make a venv, then:
• pip install -r requirements.txt
• pip install umap-learn

USAGE:

• Clone the repo
• go to the repo's root dir
• To start the Streamlit dashboard do streamlit run stmain.py

DATA LICENSE:

• Please find details in ./license_xc_data/xeno_canto_files_list.csv