rollout_pick_n_place_dagger_r9
收藏资源简介:
该数据集使用LeRobot工具创建,属于机器人学领域,包含一个完整的episode,共计2142帧数据,帧率为30fps。数据以Parquet文件格式存储,总数据文件大小约为100MB,关联的视频文件大小约为200MB。核心特征包括:机器人的动作指令(action),为6维浮点数数组,控制肩部平移、肩部抬升、肘部弯曲、腕部弯曲、腕部旋转和夹爪的位置;观测状态(observation.state),同样是6维浮点数数组,反映六个关节的实时位置;两个视觉观测通道(observation.images.front和observation.images.annotated),均为分辨率480x640的RGB视频流。此外,数据集还包含时间戳、帧索引、episode索引、全局索引和任务索引等元数据字段。机器人类型标注为so_follower。数据集仅划分了训练集,适用于机器人模仿学习、策略学习或行为克隆等任务的训练,采用Apache-2.0许可证。
This dataset is created using the LeRobot tool and belongs to the field of robotics. It contains a complete episode with a total of 2142 frames at a frame rate of 30fps. The data is stored in Parquet file format, with a total data file size of approximately 100MB and an associated video file size of about 200MB. Key features include: the robots action command (action), which is a 6-dimensional float array controlling shoulder translation, shoulder elevation, elbow flexion, wrist flexion, wrist rotation, and gripper position; the observation state (observation.state), also a 6-dimensional float array reflecting the real-time positions of the six joints; and two visual observation channels (observation.images.front and observation.images.annotated), both being RGB video streams with a resolution of 480x640. Additionally, the dataset includes metadata fields such as timestamp, frame index, episode index, global index, and task index. The robot type is labeled as so_follower. The dataset is only divided into a training set and is suitable for tasks like robot imitation learning, policy learning, or behavior cloning, licensed under Apache-2.0.
数据集概述
该数据集是一个机器人学的数据集,使用 LeRobot 框架创建,专注于机器人“拾取与放置”任务。
基本信息
- 许可证:Apache-2.0
- 任务类别:机器人学 (robotics)
- 机器人类型:so_follower
数据规模与结构
- 总集数 (Episodes):1
- 总帧数 (Frames):2142
- 总任务数 (Tasks):1
- 数据集分割:仅包含训练集 (train: "0:1")
- 数据文件大小:约 100 MB
- 视频文件大小:约 200 MB
- 帧率 (FPS):30
特征 (Features)
数据集包含以下特征:
- 动作 (action):6维浮点数向量,包含机器人各关节位置 (shoulder_pan, shoulder_lift, elbow_flex, wrist_flex, wrist_roll, gripper)。
- 观察状态 (observation.state):与动作相同的6维关节位置。
- 观察图像 (observation.images):
- front:前视摄像头视频,分辨率为 480x640,3个通道 (RGB),编码为 AV1。
- annotated:带标注的视频,同样分辨率 480x640,3个通道,编码为 AV1。
- 时间戳 (timestamp):浮点数 (float32)。
- 帧索引 (frame_index):整型 (int64)。
- 集索引 (episode_index):整型 (int64)。
- 索引 (index):整型 (int64)。
- 任务索引 (task_index):整型 (int64)。
数据格式
- 数据以 Parquet 文件存储,路径为:
data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet - 视频以 MP4 文件存储,路径为:
videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4
可视化
数据集可通过 LeRobot 可视化工具进行查看,在线链接为:https://huggingface.co/spaces/lerobot/visualize_dataset?path=anikitakis/rollout_pick_n_place_dagger_r9




