Human Gameplay Trajectories in Bleeding Edge
收藏arXiv2025-09-30 收录
下载链接:
https://adaptingworldmodels.github.io/
下载链接
链接失效反馈官方服务:
资源简介:
该数据集包含了71,940条在Sky Garden地图上记录的人类玩家游戏轨迹,这些数据来自2020年2月9日至2022年10月19日期间记录的8788场比赛。数据集不仅包括视觉和遥测数据,还包含了玩家在游戏中所采取的行动。此外,该数据集被用于评估在3D游戏环境中轨迹跟随模型的性能,其中包括了来自代理摄像机的视觉信号和符号化的遥测数据。规模上,数据集涵盖了71,940条轨迹,来自8788场比赛。该数据集的任务是对3D游戏环境中的轨迹进行跟随。
This dataset contains 71,940 human player trajectories recorded on the Sky Garden map, sourced from 8,788 matches logged between February 9, 2020 and October 19, 2022. The dataset includes not only visual and telemetry data, but also the in-game actions taken by players. Furthermore, this dataset has been utilized to evaluate the performance of trajectory-following models in 3D game environments, where the models take visual signals from agent cameras and symbolic telemetry data as input. Collectively, the dataset encompasses 71,940 trajectories derived from 8,788 matches. The dedicated task of this dataset is trajectory following in 3D game environments.



