RoboChallenge/Table30

RoboChallenge数据集包含30个不同的操作任务，涵盖4种机器人实现。每个任务包含多个演示片段，每个片段包括元数据、机器人状态和来自不同摄像头视角的视频。数据集结构按任务组织，每个任务文件夹包含任务描述、元数据和数据子文件夹。数据子文件夹中存储了单个演示片段的信息，包括元数据、机器人状态和视频文件。机器人状态数据以JSONL格式存储，记录了每帧的关节角度、末端执行器位姿、夹持器状态和时间戳。数据集还提供了将数据转换为LeRobot格式的脚本，以便利用LeRobot的数据处理和加载工具。

The RoboChallenge Dataset includes 30 diverse manipulation tasks across 4 embodiments. Each task contains multiple demonstration episodes, with each episode comprising metadata, robot states, and videos from various camera views. The dataset is organized by tasks, with each task folder containing task descriptions, metadata, and data subfolders. The data subfolders store individual episode information, including metadata, robot states, and video files. Robot state data is stored in JSONL format, recording joint angles, end-effector poses, gripper states, and timestamps per frame. The dataset also provides a script to convert the data into the LeRobot format for leveraging LeRobots data processing and loading utilities.