Pysc2_DataExtract

A tutorial of data extraction for starcraft II replays using Deepmind's starcraft II RL environment pysc2.

Environment

Please follow Pysc2 repo to install pysc2 and other dependencies. I am currently running and testing data_extraction.py at starcraft II version 4.1.1. Please note that sc2 replays are version dependent, execution will fail if replay was made with a different version of the game.

Getting started

Clone or download this repo, and cd to the directory, run

python -m data_extraction --replay "path-to-replay" --norender --map_path "path-to-map" --step_mul 10 --fps 30

Additional flags:

--disable_fog   # disable fog of war, default False.
--fps   # frames per second to run the game, default 30.
--step_mul   # game steps per observation, default 10.
--observed_player   # which player to observe, default 1.
--helpfull   # see help on flags.

Data

• The data will be download to ./replay_data/, screen and minimap features use sparse matrix to store only the nonzero elements and their position indices. Please check replay_data_visualize.ipynb for more details.
• Replay_example.7z is an example of replay data used in replay_data_visualize.ipynb.

Troubleshooting

If you have any problem running script, feel free to leave a comment.