DorianAtSchool/robocasa-trajectories-1k

--- pretty_name: RoboCasa Trajectories Single configs: - config_name: default data_files: - split: train path: data/train-* --- # RoboCasa Trajectories Single This dataset contains flat RoboCasa step rows with sidecar episode JSON. ## Structure Each row is one tool step. Episode-level JSON is not duplicated into parquet. Instead, each row carries repo-relative references: - `adapted_trajectory_path` - `original_trajectory_path` - `execution_metadata_path` Image columns stay inline and viewable in the dataset table: - `room_view` - `top_view` - `map` - `agentview_center` - `agentview_left` - `agentview_right` - `wrist` ## Summary - rows: 39732 - episodes: 1008 - tasks: HotDogSetup, PrepareCoffee, PrepareSandwichStation - average steps per episode: 39.4 ## Load ```python from datasets import load_dataset ds = load_dataset("DorianAtSchool/robocasa-trajectories-1k", split="train") ``` ## Notes - Camera renders are stored as JPEG. Maps remain PNG. - MP4 videos are not included in the dataset. - Debug `initial` / `pre_initial_state` camera frames are not part of the dataset. - Episode JSON sidecars are available in the repo files at the paths referenced by `*_path` columns.--- pretty_name: RoboCasa单轨迹数据集（RoboCasa Trajectories Single） configs: - config_name: 默认配置 data_files: - split: 训练集 path: data/train-* --- # RoboCasa单轨迹数据集（RoboCasa Trajectories Single） 本数据集包含扁平化的RoboCasa步骤行，并附带附属情节JSON文件。 ## 数据结构 每一行对应一个工具执行步骤。 情节级别的JSON并未重复存储至Parquet文件中，取而代之的是，每一行都携带了相对于数据集仓库的相对引用路径： - `adapted_trajectory_path`：适配轨迹路径 - `original_trajectory_path`：原始轨迹路径 - `execution_metadata_path`：执行元数据路径 内嵌于数据集表格中的图像列可直接预览： - `room_view`：房间视角图像 - `top_view`：俯视视角图像 - `map`：地图图像 - `agentview_center`：智能体中心视角图像 - `agentview_left`：智能体左侧视角图像 - `agentview_right`：智能体右侧视角图像 - `wrist`：腕部视角图像 ## 统计摘要 - 总数据行数：39732 - 总情节数：1008 - 涵盖任务：热狗搭建（HotDogSetup）、咖啡制备（PrepareCoffee）、三明治工作站搭建（PrepareSandwichStation） - 单情节平均步数：39.4 ## 加载方式 python from datasets import load_dataset ds = load_dataset("DorianAtSchool/robocasa-trajectories-1k", split="train") ## 注意事项 - 相机渲染图像以JPEG格式存储，地图图像采用PNG格式。 - 数据集中未包含MP4格式视频文件。 - 调试用的`initial`/`pre_initial_state`相机帧未纳入本数据集。 - 附属情节JSON文件可通过各`*_path`字段引用的仓库路径获取。