CEC3 Download
25th July update. The eval data is now available for all tasks. See details below.
Software​
All the necessary software tools are available as part of the PyClarity software package available on GitHub. The challenge is using version 0.5.x of the package.
We recommend that you download and install the latest packaged release in the 0.5.x series from GitHub. Doing this will make it easy to access the command-line recipes for the baseline system.
The software can be installed as a library using
pip install "pyclarity~=0.5.1"
Further installation instructions are available in the repository's README.
Data​
The data is available on MyAirBridge at https://mab.to/BndoGwtUhjTL3.
If you have any issues with the download, please contact the challenge organisers.
On the download site you will see the following packages are available,
clarity_CEC3_data.v1_0.tar.gz
[74 GB] - data from Task 1 and Task 2clarity_CEC3_task3_data.v1_0.tar.gz
[111 GB] - data from Task 3clarity_CEC3_data_eval.v1_0.tar.gz
[12 GB] - eval data for all tasks.
To unpack the data, use the following command:
tar -xvzf <PACKAGE_NAME>
e.g.,
tar -xvzf clarity_CEC3_data.v1_0.tar.gz
If unpacking more than one package, please unpack them in the same directory so that the unpackaged data is merged into the same root directory.
Additional Training Data​
For Task 1, participants are encouraged to use the training data set from CEC2.
The CEC2 data is available on MyAirBridge at https://mab.to/zU7TS8jJelkoD.
On the download site you will see the following package,
clarity_CEC2_train.v1_1.tgz
[69 GB] - scenes for training systems