audb

audb manages your annotated media files.

Databases used in machine learning should be easily exchangeable and versioned for reproducibility. audb allows both as it stores all databases in a common format and manages different versions of a database. Databases are stored in repositories on local file systems, MinIO, S3, or Artifactory servers.

You can request resampling or remixing of audio content and filter the downloaded data, e.g. if you just want to download the test set.

Have a look at the installation and quickstart instructions.

Citation

If you want to cite audb, you can refer to our paper:

@article{wierstorf2023audb,
  title={audb -- Sharing and Versioning of Audio and Annotation Data in Python},
  author={Hagen Wierstorf
    and Johannes Wagner
    and Florian Eyben
    and Felix Burkhardt
    and Bj{\"o}rn W. Schuller},
  journal={arXiv preprint arXiv:2303.00645},
  year={2023},
}