This repository provides the BirdNET system for detecting and classifying avian vocalizations in soundscape recordings, aimed at ornithologists and citizen scientists. It leverages machine learning to assist in bird monitoring and protection efforts.

How It Works

BirdNET is a Theano/Lasagne implementation in Python that uses artificial neural networks to identify hundreds of bird species from their vocalizations. It processes audio recordings by extracting spectrograms and applying trained models for classification. The system offers adjustable sensitivity and confidence thresholds for tailored analysis.

Quick Start & Requirements

• Install: pip install -r requirements.txt (after git clone)
• Prerequisites: Python 2/3, NumPy 1.14.5, SciPy 1.0.0, Librosa 0.7.0, Future 0.17.1, FFMPEG, Lasagne/Theano, BLAS/LAPACK.
• GPU Support: Optional, requires CUDA 9.2+ and cuDNN.
• Model Download: sh model/fetch_model.sh (300 MB)
• Docker: Provided for CPU and GPU modes.
• Docs: https://birdnet.cornell.edu/

Highlighted Details

• Supports analysis of single files or directories.
• Output formats compatible with Raven and Audacity.
• Allows specifying recording location (latitude, longitude) and week of year for context.
• Offers parameters for overlap, species probability combination, sensitivity, and minimum confidence.

Maintenance & Community

This repository is deprecated and no longer maintained. Users are directed to the BirdNET-Analyzer repository for the latest versions.

Licensing & Compatibility

Licensed under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. Not suitable for commercial use or linking with closed-source projects due to the non-commercial clause.

Limitations & Caveats

The repository is deprecated and will not receive updates or support. The original implementation relies on Theano/Lasagne, which are older deep learning frameworks. Users seeking current functionality should migrate to the recommended BirdNET-Analyzer repository.