目前平台处于试运行阶段,如有任何问题或建议,请发送邮件至 service@dataju.cn 或加入QQ群 565136792 联系管理员。

满足不同角色需求:领域专家 数据管理者 数据科学家 科研人员、高校教师及研究生 数据分析爱好者
星际争霸2录像数据集
1136次浏览 dataju 于 2017-10-14 发布
该内容是由用户自发提供,聚数力平台仅提供平台,让大数据应用过程中的信息实现共享、交易与托管。如该内容涉及到您的隐私或可能侵犯版权,请告知我们及时删除。
数据集概述

MSC: A Dataset for Macro-Management in StarCraft II

https://github.com/wuhuikai/MSC

MSC

MSC: A Dataset for Macro-Management in StarCraft II.

@article{wu2017ms,
  title={MSC: A Dataset for Macro-Management in StarCraft II},
  author={Wu, Huikai and Zhang, Junge and Huang, Kaiqi},
  journal={arXiv preprint arXiv:1710.03131},
  year={2017}
}

Download

FTP: ftp://surveillance.idealtest.org/ (User Name: msc; Password: msc)

  • [Stat]: ftp://msc:msc@surveillance.idealtest.org/Stat.tar.gz
  • [TRAIN|VAL|TEST]: ftp://msc:msc@surveillance.idealtest.org/TRAIN-VAL-TEST.tar.gz
  • [Global]: ftp://msc:msc@surveillance.idealtest.org/GlobalFeatureVector.tar.gz
  • [Spatial]:TvT: ftp://msc:msc@surveillance.idealtest.org/TvT.tar.gzTvP: ftp://msc:msc@surveillance.idealtest.org/PvT.tar.gzTvZ: ftp://msc:msc@surveillance.idealtest.org/TvZ/TvZ.tar.gzPvP: ftp://msc:msc@surveillance.idealtest.org/PvP.tar.gzPvZ: ftp://msc:msc@surveillance.idealtest.org/PvZ.tar.gzZvZ: ftp://msc:msc@surveillance.idealtest.org/ZvZ.tar.gz

For Linux and MacOS:

wget ftp://msc:msc@surveillance.idealtest.org/[FILE_NAME]

NOTE:

  • Global features are also available HERE.
  • [TRAIN|VAL|TEST] split is also available HERE.
  • [Stat] is also available HERE. The stat files with postfix _human.json are human-readable.

Baselines

Global State Evaluation

MethodTvT:TTvZ:TTvZ:ZTvP:TTvP:PZvZ:ZZvP:ZZvP:PPvP:P
Baseline[Global]61.0958.8960.6157.2160.9559.9159.9559.3551.36
Baseline[Spatial]50.8552.3559.8254.9059.1554.6555.0258.7857.76

Build Order Prediction

MethodTvT:TTvZ:TTvZ:ZTvP:TTvP:PZvZ:ZZvP:ZZvP:PPvP:P
Baseline[Global]74.1273.0173.8970.2979.2876.0772.0278.0876.28
Baseline[Spatial]73.0773.7175.9264.1575.0974.8872.3276.1274.22

Dataset: Global Feature Vector

Each replay is a (T, M) matrix F, where F[t, :] is the feature vector for time step t.

Each row of F is a M-dimensional vector, with M varying as [RACE] v.s. [RACE].

The M-dimensional vector is orgnized as follows:

  1. [0]: reward, i.e. final result of the game. 0: DEFEAT, 1: WIN.
  2. [1]: ground truth action, ranging from [0, #ACTION].
  3. [2-15): cumulative score [NOT NORMALIZED], which is defined in Here.
  4. [15-M): observation feature vector, which is normalized into [0, 1].[15]: frame id.[16-27): player info, including various resources and n_power_source.[27-#1): alerts, boolean.[#1-#2): upgrades, boolean.[#2-#3): research count.[#3-#4): friendly units info, which is defined in Here.[#4-M): enemy units info, where M = #4 + #[ENEMY RACE].V.S.TvT:TTvZ:TTvZ:ZTvP:TTvP:PZvZ:ZZvP:ZZvP:PPvP:PM75311311121663653149910311031563RACE#1#2#3#4#ACTION#RACETerran29608141775336Protoss29557131761246Zerg29557178574714

Code for loading F:

import numpy as np
from scipy import sparse
F = np.asarray(sparse.load_npz(PATH).todense())

Dataset: Spatial Feature Tensor

Each replay contains a (T, 13, 64, 64) tensor S and a (T, 26) matrix G.

The specifics for S[t, :, :, :] is as follows:

  1. S[t, 0:8, :, :]: screen features, roughly normalized into [0-1], which is defined in Here.
  2. S[t, 8:13, :, :]: minimap features, roughly normalized into [0-1], which is defined in Here.

WARNING[Cheat Layer]: The last layer S[t, 12, :, :] refers to unit_type, which could only be obtained in replays.

Code for loading S:

import numpy as np
from scipy import sparse
S = np.asarray(sparse.load_npz(PATH).todense()).reshape([-1, 13, 64, 64])

The specifics for G[t, :] is as follows:

  1. [0-11): frame id + player info, normalized into [0, 1], which is defined Here.
  2. [11-24): cumulative score [NOT NORMALIZED], which is defined in Here.
  3. [24]: reward, i.e. final result of the game. 0: DEFEAT, 1: WIN
  4. [25]: ground truth action, ranging from [0, #ACTION].

Code for loading G:

import numpy as np
from scipy import sparse
G = np.asarray(sparse.load_npz(PATH).todense())

Build the Dataset Yourself Step by Step

Install SC2LE

  1. Download and unzip (Password: iagreetotheeula) StarCraft II Linux Packages 3.16.1 into $STAR_CRAFT$.
  2. Download and unzip (Password: iagreetotheeula) Replay Packs (3.16.1 - Pack 13.16.1 - Pack 2 [Currently not used]) into $STAR_CRAFT$.

After step 1 and step 2, the folder structure is as follows:

$STAR_CRAFT$
    ├── Battle.net
    ├── Libs
    ├── Maps
    ├── Replays
    ├── SC2Data
    └── Versions
  • NOTE:$STAR_CRAFT$/Replays contains all *.SC2Replay files from 3.16.1 - Pack 1 and 3.16.1 - Pack 2 [Currently not used]$STAR_CRAFT$/Battle.net contains all contents from the folder Battle.net in 3.16.1 - Pack 1 and 3.16.1 - Pack 2 [Currently not used]

Step-by-Step Instructions

The Easy Way

The Hard Way [Step-by-Step in Details]

Requirements

future == 0.16.0

numpy == 1.13.0
scipy == 0.19.0

python_gflags == 3.1.1

tqdm == 4.14.0

protobuf == 3.4.0
pystream_protobuf == 1.4.4

PySC2 == 1.0
s2clientprotocol == 1.1


数据集详情
暂无
数据集元数据
暂无
概念层次
领域场景: 未指定
领域问题: 未指定
领域应用: 未指定
应用案例: 未指定

目前平台处于试运行阶段,如有任何问题或建议,请发送邮件至 service@dataju.cn 或加入QQ群 565136792 联系管理员。