Skip to main content

Download data and software

How to start

We suggest starting with the installation of the software and then proceeding to follow the instructions for downloading and extracting the data.

A. Software

All the necessary software to run the recipes and make your own submission is available on our Clarity-Cadenza GitHub repository.

The official code for the ICASSP 2024 Cadenza Challenge was released in version v0.4.1. To avoid any conflict, we highly recommend for you to work using this version and not with the code from the main branch. To install this version you have three options

1. Download the files of the release v0.4.1

From inside the directory, run:

pip install -e .
2. Clone the repository and checkout version v0.4.1
git clone
git checkout tags/v0.4.1

cd clarity
pip install -e .
3. Install pyclarity version 0.4.1 from PyPI
pip install pyclarity==0.4.1

B. Data

B.1 Download the packages

The data is available in five packages, please complete this form to request access.

All participants will require the core data packages.

Participants wanting to extend the training data can use either of the augmentation data packages.

  • cad_icassp_2024_core.v1.1.tgz [230 KB] - metadata and hrtf signals.
  • cad_icassp_2024_train.v1.0.tgz [14.8 GB] - audio data for training systems.
  • cad_icassp_2024_validation.v1.0.tgz [5.9 GB] - audio and metadata data for validation.
  • cad_icassp_2024_evaluation.v1.0.tgz [1.4 GB] - audio and metadata data for evaluation (only mixtures).
  • cad_icassp_2024_medleydb.v1.0.tgz [38.1 GB] - optional audio data for training augmentation.
  • [125 MB] - optional audio data for training augmentation.
  • [7.2 GB] - optional audio data for training augmentation.

B.2 Unpack the packages

After downloading the packages Unpack the packages under the same root directory as:

tar -xvzf <PACKAGE_NAME>