Atari-HEAD
收藏资源简介:
Atari-HEAD数据集是由德克萨斯大学奥斯汀分校计算机科学系创建的,专注于人类在玩Atari视频游戏时的行为和凝视数据。该数据集包含117小时的游玩数据,覆盖20种不同的游戏,记录了800万次行动演示和3.28亿次凝视样本。数据收集过程中,采用了半帧逐帧的游戏模式,确保了数据的高质量和行动与状态的匹配。Atari-HEAD数据集主要用于视觉注意力、模仿学习和强化学习领域的研究,旨在通过分析人类的游戏行为和凝视模式,提高机器在这些领域的学习和决策能力。
The Atari-HEAD dataset was created by the Department of Computer Science at The University of Texas at Austin, focusing on human behavioral and gaze data while playing Atari video games. This dataset contains 117 hours of gameplay data spanning 20 distinct games, recording 8 million action demonstrations and 328 million gaze samples. During the data collection process, a semi-frame-by-frame gameplay mode was adopted to ensure high data quality and accurate alignment between actions and game states. The Atari-HEAD dataset is primarily utilized for research in the fields of visual attention, imitation learning, and reinforcement learning, aiming to improve machine learning and decision-making capabilities in these domains by analyzing human gameplay behaviors and gaze patterns.




