openarm-cube-distractor-lang-v3
收藏资源简介:
该数据集使用LeRobot框架创建,是一个用于机器人学的模拟数据集,特别关注拾放任务。数据集包含500个训练episodes,总计222,451帧数据,采样频率为50 FPS。数据以parquet文件格式存储,总数据文件大小为100 MB,视频文件大小为200 MB。数据集结构包括多个特征字段:观测状态(9维浮点数组,表示7个关节和2个手指关节的位置)、动作(8维浮点数组,表示7个关节目标和1个夹持器目标)、三个视角的图像观测(来自左腕、右腕和基座摄像头,均为480x480 RGB视频)、时间戳、帧索引、episode索引、全局索引和任务索引。数据集适用于语言条件、多物体场景的机器人控制任务,如模拟环境中的强化学习训练。标签表明其与MuJoCo模拟器、开源机器人平台(OpenArm)和强化学习生成数据相关。
This dataset is created using the LeRobot framework and is a simulation dataset for robotics, specifically focusing on pick-and-place tasks. It contains 500 training episodes, with a total of 222,451 frames sampled at 50 FPS. The data is stored in parquet format, with a total data file size of 100 MB and video file size of 200 MB. The dataset structure includes multiple feature fields: observation state (a 9-dimensional float array representing the positions of 7 joints and 2 finger joints), action (an 8-dimensional float array representing 7 joint targets and 1 gripper target), image observations from three perspectives (left wrist, right wrist, and base cameras, all as 480x480 RGB videos), timestamp, frame index, episode index, global index, and task index. It is suitable for language-conditioned, multi-object scenario robot control tasks, such as reinforcement learning training in simulated environments. Labels indicate its association with the MuJoCo simulator, open-source robotics platform (OpenArm), and reinforcement learning-generated data.
数据集概述:openarm-cube-distractor-lang-v3
数据集名称: openarm-cube-distractor-lang-v3
发布机构: qualia-robotics
许可证: Apache-2.0
任务类别: 机器人学 (robotics)
相关标签: LeRobot, OpenArm, MuJoCo, 仿真, 拾取与放置, 语言条件控制, 多物体, 强化学习生成
数据集结构
- 代码库版本: v3.0
- 机器人类型:
openarm_bimanual_sim(双臂OpenArm仿真机器人) - 总集数 (episodes): 500
- 总帧数 (frames): 222,451
- 任务数量: 5 种
- 帧率 (fps): 50
- 数据划分: 全部500个episodes用于训练 (
train: 0:500)
文件存储:
- 数据文件:
data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet(Parquet格式,共约100 MB) - 视频文件:
videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4(H.264编码,共约200 MB)
特征 (Features)
| 特征名 | 数据类型 | 形状 | 说明 |
|---|---|---|---|
observation.state |
float32 | (9,) | 包含7个关节目标 + 2个手指关节位置 |
action |
float32 | (8,) | 包含7个关节目标 + 1个夹爪目标 |
observation.images.wrist_left |
视频 (h264) | 480×480×3 | 左腕摄像头图像 (50 fps) |
observation.images.wrist_right |
视频 (h264) | 480×480×3 | 右腕摄像头图像 (50 fps) |
observation.images.base |
视频 (h264) | 480×480×3 | 基部摄像头图像 (50 fps) |
timestamp |
float32 | (1,) | 时间戳 |
frame_index |
int64 | (1,) | 帧索引 |
episode_index |
int64 | (1,) | 集索引 |
index |
int64 | (1,) | 全局索引 |
task_index |
int64 | (1,) | 任务索引 |
生成与使用
- 创建工具: 使用 LeRobot 框架生成
- 可视化: 可通过 Hugging Face 空间可视化 查看
- 用途: 适用于语言条件控制的拾取与放置任务仿真数据集,由强化学习生成,支持多物体场景




