Atari Grand Challenge Dataset
收藏arXiv2017-05-31 更新2024-06-21 收录
下载链接：
http://atarigrandchallenge.com/data
下载链接
链接失效反馈官方服务：
资源简介：
Atari Grand Challenge Dataset是由德国亚琛工业大学视觉计算研究所创建的大型数据集，包含约970万帧（约45小时）的人类Atari 2600游戏重播，是目前公开发布的最大和最多样化的此类数据集。数据集涵盖五款游戏，旨在通过分析人类演示的质量与模仿学习性能之间的关系，推动强化学习在复杂虚拟环境中的应用。该数据集不仅支持强化学习研究，还可能对机器学习、游戏AI和认知科学等领域产生影响。
Atari Grand Challenge Dataset is a large-scale dataset created by the Institute for Visual Computing at RWTH Aachen University. It contains approximately 9.7 million frames (equivalent to around 45 hours) of human-played Atari 2600 game replays, making it the largest and most diverse publicly released dataset of its kind to date. The dataset covers five game titles, and is designed to advance the application of reinforcement learning in complex virtual environments by analyzing the relationship between the quality of human demonstrations and imitation learning performance. This dataset not only supports reinforcement learning research, but may also exert influences on fields such as machine learning, game AI, and cognitive science.
提供机构：
视觉计算研究所亚琛工业大学
创建时间：
2017-05-31
