/HumBugDB

Acoustic mosquito detection code with Bayesian Neural Networks

Primary LanguageJupyter NotebookMIT LicenseMIT

HumBugDB

Acoustic mosquito detection with Bayesian Neural Networks.

General use instructions

This code is complementary to the paper: "HumBugDB: a large-scale acoustic mosquito dataset" and the dataset on Zenodo.

See documentation in the paper supplement for:

  • Section A: Licensing
  • Section B: Code use, feature and model engineering
  • Section C: Description and visualisations of metadata in data/metadata/*.csv

Additional documentation for:

Installation instructions

You may choose to use the Colab environment which is natively compatible with all of our code. Alternatively, see the instructions for manually configuring an environment to run the Jupyter notebooks.

Google Colab

  • Installation and use with Google Colab here.

Jupyter notebook

  • Installation instructions and requirements for PyTorch: InstallationLogPyTorch.txt.
  • Keras min. requirements: condarequirementsKeras.txt and piprequirementsKeras.txt. Compatible with Tensorflow 1.X and 2.X.

After installation of requirements:

  • git clone https://github.com/HumBug-Mosquito/HumBugDB.git
    
  • Extract audio from four-part-archive to Zenodo to /data/audio/.

Contact

Developed by Ivan Kiskin of MLRG University of Oxford. Contact ivankiskin1@gmail.com for enquiries or suggestions. Follow our Twitter on @OxHumBug and visit our HumBug website for updates on the overall HumBug project.