LEAVES is a powerful ecoacoustics tool designed to streamline the annotation and visualization of large-scale natural soundscape datasets. By leveraging advanced machine learning, real-time analysis, and a user-friendly interface, LEAVES empowers researchers and citizen scientists to process and label their data efficiently.

• Efficient Labelling: Reduces annotation time for large datasets with cluster-based workflows.
• Interactive Visualisations: Explore your data in 3D scatterplots, spectrograms, and waveforms.
• Customizable Configuration: Flexible settings for preprocessing, clustering, and visualization.
• Multi-Format Support: Works with .WAV, .MP3, .FLAC, and more.
• Real-Time Spectrograms: Analyze acoustic features while annotating.
• Cluster Filtering: Focus on specific sound groups for detailed examination.

Get the most out of LEAVES with the LEAVES User Guide, which includes:

• Step-by-step setup instructions.
• Detailed explanations of features like File Uploading, Annotation, Cluster Filtering, and more.
• Visual examples and troubleshooting tips.

Follow these steps to install and start using LEAVES:

git clone https://github.com/thomasnapier/LEAVES.git

Navigate to the project directory and install the required Python libraries:

pip install -r requirements.txt

Start the web-based interface by running:

python app.py

Open your browser and navigate to http://127.0.0.1:8050.

Use the Upload Module to import your audio recordings and visualize them in 3D scatterplots.

Process, annotate, and save your data using LEAVES' advanced clustering and annotation tools.

1. Data Ingestion: Upload audio recordings in .WAV, .MP3, .FLAC, or other supported formats.
2. Signal Processing: Denoising, short-term windowing, and feature extraction (e.g., MFCCs) prepare your data for analysis.
3. Dimensionality Reduction: Techniques like UMAP and t-SNE reduce data complexity for intuitive visualization.
4. Clustering: Algorithms like DBSCAN and k-means group similar sounds for efficient labeling.
5. Annotation: Assign labels to clusters or individual points, with propagation features to save time.
6. Visualisation: Interactive tools like waveforms, spectrograms, and 3D scatterplots aid exploration and analysis.

• Technologies Used: Python 3, Dash, Plotly, Librosa, Scikit-Learn.
• Current Status: Alpha Version.
• Documentation: User Guide

• Project Site: Comprehensive guide and instructions for LEAVES features.
• Installation Guide: Detailed instructions for setup.
• Video Demonstration: Talk and live demonstration.
• Paper: Paper on the methodology and experimental results.

If you use LEAVES in your work, please cite it as:

@article{napier2025leaves,
    title={LEAVES: An open-source web-based tool for the scalable annotation and visualisation of large-scale ecoacoustic datasets using cluster analysis},
    author={Napier, Thomas and Ahn, Euijoon and Allen-Ankins, Slade and Schwarzkopf, Lin and Lee, Ickjai},
    journal={Ecological Informatics},
    volume={87},
    pages={103026},
    year={2025},
    publisher={Elsevier},
    doi={https://doi.org/10.1016/j.ecoinf.2025.103026}
}						

For questions or feedback, feel free to reach out:

• Author: Thomas Napier
• GitHub: Thomas Napier