haneda-20260529-urinal-right-local-wipe-1s-12s
收藏资源简介：
该数据集是一个用于机器人学习的数据集，由LeRobot工具创建。它包含50个训练集，总计16500帧数据，帧率为30fps。数据以parquet格式存储，总数据文件大小为100MB，视频文件大小为200MB。数据集的特征包括：动作数据（8维浮点数组，表示机器人的关节位置和旋转模式）、观测状态数据（21维浮点数组，包括关节位置、速度和扭矩）、来自末端执行器相机的图像观测（480x640x3的视频数据，编码为av1，无音频）、点云数据（包括XYZ坐标、掩码和RGB信息，每个点云有4096个点），以及时间戳、帧索引、剧集索引、索引和任务索引等元数据。机器人类型为cleank-2-0_follower-w-lift，适用于机器人控制、模仿学习或强化学习任务。
This dataset is designed for robotic learning and was created with the LeRobot toolkit. It contains 50 training sets, with a total of 16500 frames at a frame rate of 30 fps. The data is stored in Parquet format, with an overall data file size of 100 MB and a video file size of 200 MB. The dataset includes the following features: 8-dimensional floating-point action arrays representing the robot's joint positions and rotation modes; 21-dimensional floating-point observation state arrays covering joint positions, velocities, and torques; image observations from the end-effector camera, which are 480×640×3 video data encoded in AV1 format without audio; point cloud data containing XYZ coordinates, masks, and RGB information, with 4096 points per point cloud; as well as metadata including timestamps, frame indices, episode indices, indexes, and task indices. The robot platform for this dataset is cleank-2-0_follower-w-lift, and it is applicable to robotic control, imitation learning, and reinforcement learning tasks.
数据集概览
- 数据集名称: haneda-20260529-urinal-right-local-wipe-1s-12s
- 创建者: CleanK-07
- 许可证: Apache-2.0
- 任务类别: 机器人学 (robotics)
- 标注: LeRobot
- 数据集配置: 默认配置，数据文件位于
data/*/*.parquet
数据集结构
- 代码库版本: v3.0
- 机器人类型: cleank-2-0_follower-w-lift
- 总集数 (Episodes): 50
- 总帧数 (Frames): 16,500
- 总任务数 (Tasks): 1
- 块大小 (Chunk Size): 1,000
- 数据文件大小: 100 MB
- 视频文件大小: 200 MB
- 帧率 (FPS): 30
- 数据集划分: 训练集 (train) 包含 0 到 49 集 (共50集)
特征 (Features)
数据集包含以下特征，每个特征的类型和维度如下：
动作 (action)
- 数据类型: float32
- 维度: [8]
- 特征名称:
- yaw.pos
- pitch1.pos
- pitch2.pos
- roll1.pos
- pitch3.pos
- roll2.pos
- lift.pos
- roll2.spin_mode
观测状态 (observation.state)
- 数据类型: float32
- 维度: [21]
- 特征名称:
- yaw.pos
- pitch1.pos
- pitch2.pos
- roll1.pos
- pitch3.pos
- roll2.pos
- lift.pos
- yaw.vel
- pitch1.vel
- pitch2.vel
- roll1.vel
- pitch3.vel
- roll2.vel
- lift.vel
- yaw.tor
- pitch1.tor
- pitch2.tor
- roll1.tor
- pitch3.tor
- roll2.tor
- lift.tor
观测图像 (observation.images.cam_ee)
- 数据类型: video
- 维度: [480, 640, 3] (高x宽x通道)
- 视频信息:
- 高度: 480
- 宽度: 640
- 编码: av1
- 像素格式: yuv420p
- 是否为深度图: false
- 帧率: 30
- 通道数: 3
- 是否有音频: false
点云数据 (observation.point_cloud.cam_ee.xyz)
- 数据类型: float32
- 维度: [4096, 3]
- 特征名称: point, xyz
点云掩码 (observation.point_cloud.cam_ee.mask)
- 数据类型: uint8
- 维度: [4096]
- 特征名称: point
点云RGB (observation.point_cloud.cam_ee.rgb)
- 数据类型: uint8
- 维度: [4096, 3]
- 特征名称: point, rgb
其他特征
- timestamp: float32, 维度 [1]
- frame_index: int64, 维度 [1]
- episode_index: int64, 维度 [1]
- index: int64, 维度 [1]
- task_index: int64, 维度 [1]
数据存储路径
- 数据文件:
data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet - 视频文件:
videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4
引用信息
当前未提供 BibTeX 引用格式。
