Soundata is a Python library for loading and working with audio datasets in a standardized way,
removing the need for writing custom loaders in every project, and improving reproducibility by providing
tools validate data against a canonical version. It speeds up research pipelines by allowing you to quickly
download a dataset,
load it into memory in a
that the dataset is complete and correct, and more.
Soundata is based and inspired on mirdata, and was created following these desig principles:
Easy to use: Soundata is designed to be easy to use and to simplify the research pipeline considerably. Check out the examples in the Getting started page.
Easy to contribute to: we welcome and encourage contributions, especially new datasets. You can contribute following the instructions in our Contributing page.
Increase reproducibility: by providing a common framework for researchers to compare and validate their data, when mistakes are found in annotations or audio versions change, using Soundata the audio community can fix mistakes while still being able to compare methods moving forward.
Standardize usage of sound datasets: we standardized common attributes of sound datasets such as
tagsto simplify audio research pipelines, while preserving each dataset’s idiosyncracies: if a dataset has ‘non-standard’ attributes, we include them as well.
To install Soundata simply do:
pip install soundata
We recommend to do this inside a conda or virtual environment for reproducibility.
import soundata # learn wich datasets are available in soundata print(soundata.list_datasets()) # choose a dataset and download it dataset = soundata.initialize('urbansound8k', data_home='choose_where_data_live') dataset.download() # get annotations and audio for a random clip example_clip = dataset.choice_clip() tags = example_clip.tags y, sr = example_clip.audio
We welcome and encourage contributions to soundata, especially new dataset loaders. To contribute a new loader, please follow the steps indicated in the Contributing page and submit a Pull Request (PR) to the github repository. For any doubt or comment about your contribution, you can always submit an issue or open a discussion in the repository.
Fuentes, M., Salamon, J., Zinemanas, P., Rocamora, M., Plaja, G., Román, I.R., Miron, M., Serra, X. and Bello, J.P., 2021. Soundata: A Python library for reproducible use of audio datasets. arXiv e-prints, pp.arXiv-2109.
When working with datasets, please cite the version of
soundata that you are using (given by the
AND include the reference of the dataset, which can be found in the respective dataset loader using the