eval_DT_pickplace_draw_complete
收藏Hugging Face2026-06-29 更新2026-06-29 收录
下载链接:
https://huggingface.co/datasets/edmos7/eval_DT_pickplace_draw_complete
下载链接
链接失效反馈官方服务:
资源简介:
该数据集使用LeRobot工具创建,专为机器人学任务设计。它包含20个episodes,总计13140帧数据,以parquet格式存储数据文件,视频则以MP4格式存储。数据集的特征包括:动作特征为6维浮点数组,表示机器人关节位置;观测状态为6维浮点数组,同样表示机器人关节位置;侧视图像观测为480x640x3分辨率的视频,采用AV1编码,帧率为30帧/秒;腕部图像观测也为480x640x3分辨率的视频,AV1编码,30帧/秒。此外,数据集还包含元数据,如时间戳、帧索引、episode索引、索引和任务索引。其结构支持训练分割(范围从0到20),适用于机器人控制和学习相关任务。
This dataset is created using LeRobot and is specifically designed for robotics tasks. It contains 20 episodes with a total of 13,140 frames, stored in parquet format for data and MP4 format for videos. The features include: actions as 6-dimensional floating-point arrays representing robot joint positions; observation states as 6-dimensional floating-point arrays also representing robot joint positions; side-view image observations as 480x640x3 videos with AV1 encoding at 30 frames per second; and wrist image observations as 480x640x3 videos with AV1 encoding at 30 frames per second. Additionally, the dataset includes metadata such as timestamps, frame indices, episode indices, indices, and task indices. Its structure supports a training split (0:20) and is suitable for robot control and learning tasks.
提供机构:
edmos7创建时间:
2026-06-29
原始信息汇总
数据集概述
- 数据集名称: eval_DT_pickplace_draw_complete
- 数据源链接: https://huggingface.co/datasets/edmos7/eval_DT_pickplace_draw_complete
- 许可证: Apache-2.0
- 任务类别: 机器人学 (robotics)
- 标签: LeRobot
数据集详情
- 创建方式: 使用 LeRobot 创建。
- 机器人类型: so_follower
- 总片段数: 20
- 总帧数: 13140
- 总任务数: 1
- 块大小: 1000
- 数据文件大小: 100 MB
- 视频文件大小: 200 MB
- 帧率: 30 FPS
- 数据集划分: 全部 20 个片段用于训练。
数据结构
- 代码库版本: v3.0
- 数据路径:
data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet - 视频路径:
videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4
特征说明
- 动作 / 观察状态:
- 数据类型: float32
- 维度: 6
- 名称: shoulder_pan.pos, shoulder_lift.pos, elbow_flex.pos, wrist_flex.pos, wrist_roll.pos, gripper.pos
- 观察图像 (侧视角):
- 尺寸: 480x640x3
- 编码: AV1
- 格式: yuv420p
- 帧率: 30 FPS
- 非深度图,有 3 个颜色通道,不含音频
- 观察图像 (腕部视角):
- 尺寸: 480x640x3
- 编码: AV1
- 格式: yuv420p
- 帧率: 30 FPS
- 非深度图,有 3 个颜色通道,不含音频
- 时间戳: float32,形状 [1]
- 帧索引: int64,形状 [1]
- 片段索引: int64,形状 [1]
- 全局索引: int64,形状 [1]
- 任务索引: int64,形状 [1]



