Configuration Issues

[aabinks]

First, this looks like a really great SC dataset generating tool. A colleague and I are trying to use it to generate data from the AIIDE Starcraft competition but are having some configuration issues getting all the pieces together. We want to use it as is for now and were wondering if you have a running ScExtractor and would consider generating data for us? We'd be happy to include you as a collaborator with you on our paper too if that is of interest.

[aabinks]

Bryan - Most of what we were looking for was in the output from ScExtractor (https://www.cs.auckland.ac.nz/research/gameai/projects/Robertson14DatabaseDiagram.pdf). Namely state information from replays that capture player visibility/fog-of-war and units/buildings throughout the game.

[BryanSWeber]

Hi Adam, I can give getting this particular extractor a try and see if I can get the information output in the SQL databases as listed. It looks like it has sat for awhile (7 years) so it may be difficult to get working and extracting the data. Otherwise, I think it is actually *easier* to write something from scratch that will dump all of the unit and player data you are asking for to a text file (.csv) for each game. You mention that "most" of what you are looking for is in there. What else did you need? Bryan Weber

…

On Thu, Sep 24, 2020 at 10:51 AM Adam Amos-Binks ***@***.***> wrote: Bryan - Most of what we were looking for was in the output from ScExtractor ( https://www.cs.auckland.ac.nz/research/gameai/projects/Robertson14DatabaseDiagram.pdf). Namely state information from replays that capture player visibility/fog-of-war and units/buildings throughout the game. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AHBOJD6KDVW6NTUVSSOVQRTSHNMGJANCNFSM4RYLTSEQ> .

[phoglenix]

Hi Adam, As Brian pointed out, the project hasn't been updated in a while since I finished my research. I would expect it to still work because StarCraft hasn't been updated either. Unfortunately I don't have a set up and running version of it now. I might be able to help figure out the setup issues if you can elaborate? Otherwise, if Bryan is able to dump out the relevant info for you by writing something from scratch then great (thanks for being so helpful!).

…

-Glen

On Fri, 25 Sep 2020, 01:40 Bryan S Weber, ***@***.***> wrote: Hi Adam, I can give getting this particular extractor a try and see if I can get the information output in the SQL databases as listed. It looks like it has sat for awhile (7 years) so it may be difficult to get working and extracting the data. Otherwise, I think it is actually *easier* to write something from scratch that will dump all of the unit and player data you are asking for to a text file (.csv) for each game. You mention that "most" of what you are looking for is in there. What else did you need? Bryan Weber On Thu, Sep 24, 2020 at 10:51 AM Adam Amos-Binks ***@***.*** > wrote: > Bryan - Most of what we were looking for was in the output from > ScExtractor ( > https://www.cs.auckland.ac.nz/research/gameai/projects/Robertson14DatabaseDiagram.pdf ). > Namely state information from replays that capture player > visibility/fog-of-war and units/buildings throughout the game. > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > < #1 (comment)>, > or unsubscribe > < https://github.com/notifications/unsubscribe-auth/AHBOJD6KDVW6NTUVSSOVQRTSHNMGJANCNFSM4RYLTSEQ > > . > — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAIYF2ECU3AKUONPDLWXMGDSHNSALANCNFSM4RYLTSEQ> .

[aabinks]

Glen - thanks for the prompt response. Main issue we are having is getting a tcl/tk for BWAPI 3.7. Any pointers on that would be helpful and if you think ScExtractor could run with a newer version of BWAPI (4.2 installs without issue).

[aabinks]

Bryan - I could see it being easier to start from scratch and a .csv would be great. As for data, we want to make sure we have enough game state info to determine the enemy tiles/units/buildings an AI agent can 'see' in any frame (perhaps by aggregating unit visibilities from the extracted data) and know the location of its own units (though it seems like that may be there already). Does that sound feasible? Do you have interest in collaborating on a paper?

[BryanSWeber]

I just realized responding to githubs update posted here instead of emailing privately. Sorry for cluttering the chat with private correspondence.