Download data and software
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
- download from https://github.com/claritychallenge/clarity/releases/tag/v0.4.1
- unpacked the package
From inside the directory, run:
pip install -e .
2. Clone the repository and checkout version v0.4.1
git clone https://github.com/claritychallenge/clarity.git
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 here.
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_complete.v1.0.tgz
[6.3 GB] - audio and metadata data for evaluation (mixtures and stems).
For data augmentation, you can request from their original source:
- MedleyDB version 1: https://zenodo.org/records/1649325 (access upon request)
- MedleyDB version 2: https://zenodo.org/records/1715175 (access upon request)
- BACH10: https://labsites.rochester.edu/air/resource.html (free access)
- FMA_small: https://os.unil.cloud.switch.ch/fma/fma_small.zip (free access)
B.2 Unpack the packages​
After downloading the packages Unpack the packages under the same root directory as:
tar -xvzf <PACKAGE_NAME>