This project contains various scripts that can assist in the process of preparing datasets. To have a broad overview of the tools please refer to the Detailed Tools Description.
Tools in this repository were used to create the SC2ReSet: StarCraft II Esport Replaypack Set, and finally SC2EGSet: StarCraft II Esport Game State Dataset, citation information Cite Us!.
Note
To run this project there are some prerequisites that you need to have installed on your system:
- Docker
- make
Our prefered way of distributing the toolset is through DockerHub. We Use the Docker Image to provide a fully reproducible environment for our scripts.
To pull the image from DockerHub, run the following command:
docker pull kaszanas/datasetpreparator:latest
If you wish to clone the repository and build the Docker image yourself, run the following command:
make docker_build
After building the image, please refer to the Command Line Arguments Usage section for the usage of the scripts and for a full description for each of the scripts refer to Detailed Tools Description.
When using Docker, you will have to pass the arguments through the docker run
command and mount the input/output directory. Below is an example of how to run the directory_flattener
script using Docker. For ease of use we have prepared example directory structure in the processing
directory. The command below uses that to issue a command to flatten the directory structure:
docker run \
-v "./processing:/app/processing" \
datasetpreparator:latest \
python3 directory_flattener.py \
--input_path /app/processing/directory_flattener/input \
--output_path /app/processing/directory_flattener/output
Note
Instructions below are for reproducing the result of the SC2EGSet dataset. If you wish to use the tools in this repository separately for your own dataset, please refer to the Detailed Tools Description.
We provide a release image containing all of the scripts. To see the usage of these scripts please refer to their respective README.md
files as described in Detailed Tools Description.
The following steps were used to prepare the SC2EGSet dataset:
- Build the docker image for the DatasetPreparator using the provided
makefile
command:make docker_build
. This will load all of the dependencies such as the SC2InfoExtractorGo. - Place the input replaypacks into
./processing/directory_flattener/input
directory. - Run the command
make sc2reset_sc2egset
to process the replaypacks and create the dataset. The output will be placed in./processing/sc2egset_replaypack_processor/output
directory.
Each of the scripts has its usage described in their respective README.md
files, you can find the detailed description of the available tools below.
- Directory Packager (dir_packager): README
- Directory Flattener (directory_flattener): README
- File Renamer (file_renamer): README
- JSON Merger (json_merger): README
- Processed Mapping Copier (processed_mapping_copier): README
- SC2 Map Downloader (sc2_map_downloader): README
- SC2EGSet Replaypack Processor (sc2egset_replaypack_processor): README
- SC2ReSet Replaypack Downloader (sc2reset_replaypack_downloader): README
If you want to report a bug, request a feature, or open any other issue, please do so in the issue tracker.
Please see CONTRIBUTING.md for detailed development instructions and contribution guidelines.
@software{Białecki_2022_6366039,
author = {Białecki, Andrzej and
Białecki, Piotr and
Krupiński, Leszek},
title = {{Kaszanas/SC2DatasetPreparator: 1.2.0
SC2DatasetPreparator Release}},
month = {jun},
year = {2022},
publisher = {Zenodo},
version = {1.2.0},
doi = {10.5281/zenodo.5296664},
url = {https://doi.org/10.5281/zenodo.5296664}
}
@article{Bialecki2023_SC2EGSet,
author = {Bia{\l}ecki, Andrzej
and Jakubowska, Natalia
and Dobrowolski, Pawe{\l}
and Bia{\l}ecki, Piotr
and Krupi{\'{n}}ski, Leszek
and Szczap, Andrzej
and Bia{\l}ecki, Robert
and Gajewski, Jan},
title = {SC2EGSet: StarCraft II Esport Replay and Game-state Dataset},
journal = {Scientific Data},
year = {2023},
month = {Sep},
day = {08},
volume = {10},
number = {1},
pages = {600},
issn = {2052-4463},
doi = {10.1038/s41597-023-02510-7},
url = {https://doi.org/10.1038/s41597-023-02510-7}
}