so101_act_dualcam_green_v1_20260723_175739
收藏资源简介:
该数据集是使用LeRobot工具创建的机器人学数据集,主要用于机器人控制和学习任务。数据集包含3个episodes,总计1134帧,数据以parquet文件和视频文件形式存储。特征包括动作(action)和观测状态(observation.state),其中动作和状态都涉及6个关节位置(如肩部平移、肩部提升、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置)。观测部分还包括来自手腕摄像头和前方摄像头的视频数据,视频分辨率为480x640,帧率为30fps,使用AV1编解码器。此外,数据集包含时间戳、帧索引、episode索引、索引和任务索引等元数据。机器人类型为so_follower,数据集总大小约为300MB(数据文件100MB,视频文件200MB),并按照训练分割(train: 0:3)组织。该数据集适用于机器人模拟、动作预测和视觉-动作对齐等研究。
This robotics dataset was developed using the LeRobot toolkit, and is primarily designed for robotic control and learning tasks. It comprises 3 episodes, with a total of 1134 frames, and is stored in both Parquet files and video files. The dataset features two core types of data: actions and observation states (observation.state), both of which encompass six joint positions, namely shoulder translation, shoulder elevation, elbow flexion, wrist flexion, wrist rotation, and gripper position. The observation modality additionally includes video data captured by the wrist-mounted camera and the front-facing camera; these videos have a resolution of 480×640, a frame rate of 30 fps, and use the AV1 codec. Furthermore, the dataset contains various metadata items, including timestamps, frame indices, episode indices, indices, and task indices. The target robot platform is so_follower. The total size of the dataset is approximately 300 MB, consisting of 100 MB of data files and 200 MB of video files, and it is organized following the training split (train: 0:3). This dataset is suitable for research in areas such as robotic simulation, action prediction, and vision-action alignment.
数据集概述
- 数据集名称:so101_act_dualcam_green_v1_20260723_175739
- 许可证:Apache-2.0
- 任务类别:机器人学(robotics)
- 标签:LeRobot
- 创建工具:使用 LeRobot 创建
数据集结构
-
基础信息:
- 代码库版本:v3.0
- 帧率:30 FPS
- 机器人类型:so_follower
- 总片段数(episodes):3
- 总帧数:1134
- 总任务数:1
- 数据块大小:1000
- 数据文件大小:100 MB
- 视频文件大小:200 MB
-
特征(Features):
- action:6维浮点数组,包含机器人关节位置(shoulder_pan、shoulder_lift、elbow_flex、wrist_flex、wrist_roll、gripper)
- observation.state:6维浮点数组,观测到的机器人关节位置,与 action 结构相同
- observation.images.wrist:手腕摄像头视频,分辨率 480×640,3通道(RGB),编码格式 av1,帧率 30 FPS
- observation.images.front:前方摄像头视频,分辨率 480×640,3通道(RGB),编码格式 av1,帧率 30 FPS
- timestamp:时间戳,float32,形状 [1]
- frame_index:帧索引,int64,形状 [1]
- episode_index:片段索引,int64,形状 [1]
- index:全局索引,int64,形状 [1]
- task_index:任务索引,int64,形状 [1]
-
数据划分:
- 训练集(train):片段索引 0 至 2(共3个片段)
-
文件存储路径:
- 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
- 数据文件:
其他信息
- 主页:暂缺
- 论文:暂缺
- 引用:暂缺(BibTeX 待补充)



