MSC
收藏资源简介:
MSC数据集是中国科学院自动化研究所模式识别国家重点实验室创建的,专注于星际争霸II游戏中的宏观管理问题。该数据集包含36619条高质量的游戏回放数据,通过标准化的预处理、解析和特征提取过程构建。数据集内容涵盖精心设计的功能向量、预定义的高级动作空间以及每场比赛的最终结果。创建过程中,数据集被分为训练、验证和测试集,以便于评估和比较。MSC数据集主要应用于宏观管理任务,如建造顺序预测和全局状态评估,旨在通过模仿专业玩家的行为来提高AI的表现。
The MSC Dataset was created by the State Key Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, focusing on macro-management problems in StarCraft II games. This dataset contains 36,619 high-quality game replay data, and is constructed through standardized preprocessing, parsing and feature extraction processes. The dataset covers carefully designed feature vectors, predefined high-level action spaces, and the final results of each match. During the creation process, the dataset was split into training, validation and test sets to facilitate evaluation and comparison. The MSC Dataset is mainly applied to macro-management tasks such as build order prediction and global state assessment, aiming to improve AI performance by imitating the behaviors of professional players.

- 1MSC: A Marine Wildlife Video Dataset with Grounded Segmentation and Clip-Level Captioning香港科技大学 · 2025年



