Avian biodiversity and vocalization behavior in Washington's Olympic Experimental State Forest
- Visual Studio Code with Microsoft extensions Python, Python Debugger, and Jupyter
- Install dependencies to a virtual python environment (.venv) that is used exclusively for this project. Create the environment manually, or via VS Code.
- Python 3.9+ 64-bit (3.10 recommended, ensure "Add path to environment variables" is checked during install)
From a terminal shell within the virtual environment, navigate to the root directory of this repository (.../avian-bioacoustics-oesf
), and run:
git submodule update --init --recursive
pip install --upgrade pip
pip install -r requirements.txt
This should install all dependencies you need to run, for example,
analyze_file.py
. For further reference, see setup instructions for BirdNET-Analyzer and birdnetlib.
This repository has a fork of bird_mixit as a submodule, which is used to perform sound separation.
- Follow steps to install gsutil
- Install bird_mixit model checkpoints via a terminal shell (see below)
gsutil -m cp -r gs://gresearch/sound_separation/bird_mixit_model_checkpoints data/models/
analyses
– ecological analyses and models (e.g. species accumulation curves)annotation
– extracting samples for annotation and clusteringclassification
– run classifier on audio/directory, evaluate performancedata
– data associated with the study area and project
- Manually annotate training examples with Raven Pro
- Run
training_extract_audio_examples.py
to extract audio examples for training- Manually add any additonal class examples (e.g. "Background") to the
audio
subdirectory
- Manually add any additonal class examples (e.g. "Background") to the
- Run
training_assemble_datasets.py
to assemble datasets for training - Train custom classifiers with TODO
cd src/submodules/BirdNET-Analyzer/
python3 train.py --i /Users/giojacuzzi/repos/avian-bioacoustics-oesf/data/training/Custom/audio --o /Users/giojacuzzi/repos/avian-bioacoustics-oesf/data/models/Custom/Custom_Classifier.tflite
- Return to top directory with
cd ../../../
, then runtest_compare_validation_performance.py