smolvla_blue_white_to_gray_part1
收藏资源简介:
该数据集为机器人操作数据集,使用LeRobot框架创建。数据来自Aloha Mini机器人,包含50个片段(episodes),共28860帧。每个片段记录了一次操作任务。数据包含18维的动作和观察状态,涉及左右手臂各关节位置、夹爪位置、移动速度和升降轴高度。同时提供五个摄像头视角(前、后、胸部、左手腕、右手腕)的视频流,分辨率为480x640,帧率30fps。任务类型为单一任务。
This dataset is a robot manipulation dataset created using the LeRobot framework. The data comes from the Aloha Mini robot, containing 50 episodes with a total of 28,860 frames. Each episode records a manipulation task. The data includes 18-dimensional action and observation states, involving joint positions of left and right arms, gripper positions, movement speed, and lift axis height. It also provides video streams from five camera perspectives (front, back, chest, left wrist, right wrist) with a resolution of 480x640 and a frame rate of 30fps. The task type is a single task.
数据集概述
基本信息
- 数据集地址:https://huggingface.co/datasets/yj678/smolvla_blue_white_to_gray_part1
- 许可证:apache-2.0
- 任务类别:robotics(机器人)
- 标签:LeRobot
- 创建工具:LeRobot(https://github.com/huggingface/lerobot)
- 机器人类型:alohamini_client
- 代码库版本:v3.0
数据集配置
- 配置名称:default
- 数据文件路径:data//.parquet
数据集结构
数据规模
- 总回合数(total_episodes):50
- 总帧数(total_frames):28860
- 总任务数(total_tasks):1
- 帧率(fps):30
- 数据划分:train(0:50)
- 分块大小(chunks_size):1000
- 数据文件大小上限:100 MB
- 视频文件大小上限:200 MB
- 数据路径格式:data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet
- 视频路径格式:videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4
特征(Features)
action(动作)
- 数据类型:float32
- 形状:[18]
- 名称列表:
- arm_left_shoulder_pan.pos
- arm_left_shoulder_lift.pos
- arm_left_elbow_flex.pos
- arm_left_wrist_flex.pos
- arm_left_wrist_yaw.pos
- arm_left_wrist_roll.pos
- arm_left_gripper.pos
- arm_right_shoulder_pan.pos
- arm_right_shoulder_lift.pos
- arm_right_elbow_flex.pos
- arm_right_wrist_flex.pos
- arm_right_wrist_yaw.pos
- arm_right_wrist_roll.pos
- arm_right_gripper.pos
- x.vel
- y.vel
- theta.vel
- lift_axis.height_mm
observation.state(观测状态)
- 数据类型:float32
- 形状:[18]
- 名称列表:与 action 相同(18 个维度)
observation.images.forward(前视图像)
- 数据类型:video
- 形状:[480, 640, 3]
- 维度名称:height, width, channels
- 视频信息:
- 分辨率:480 × 640
- 编码格式:av1
- 像素格式:yuv420p
- 是否为深度图:否
- 帧率:30
- 通道数:3
- 是否含音频:否
- video.g:2
- video.crf:30
- video.preset:12
- video.fast_decode:0
- 视频后端:pyav
- 额外选项:{}
observation.images.backward(后视图像)
- 数据类型:video
- 形状:[480, 640, 3]
- 维度名称:height, width, channels
- 视频信息:与前视图像相同(分辨率 480×640,av1 编码,yuv420p,30 fps,3 通道,无音频,g=2,crf=30,preset=12,fast_decode=0,pyav 后端)
observation.images.chest(胸部图像)
- 数据类型:video
- 形状:[480, 640, 3]
- 维度名称:height, width, channels
- 视频信息:与前视图像相同
observation.images.wrist_left(左腕图像)
- 数据类型:video
- 形状:[480, 640, 3]
- 维度名称:height, width, channels
- 视频信息:与前视图像相同
observation.images.wrist_right(右腕图像)
- 数据类型:video
- 形状:[480, 640, 3]
- 维度名称:height, width, channels
- 视频信息:与前视图像相同
其他特征
- timestamp:float32,形状 [1]
- frame_index:int64,形状 [1]
- episode_index:int64,形状 [1]
- index:int64,形状 [1]
- task_index:int64,形状 [1]
数据集描述
- 主页:[More Information Needed]
- 论文:[More Information Needed]
- 许可证:apache-2.0
引用
BibTeX:
[More Information Needed]



