stack_d1

该数据集由LeRobot创建，主要用于机器人技术领域。数据集包含1000个episodes和108233帧数据，数据以parquet格式存储。数据集包含多种观察和动作特征，如图像（包括agentview_image和robot0_eye_in_hand_image）、状态（包括电机位置和夹持器状态）、动作（包括电机位置和夹持器状态）、奖励、完成标志、时间戳、帧索引、episode索引、任务索引等。

This dataset was developed by LeRobot and is primarily intended for robotics research. It contains 1000 episodes and a total of 108233 frames, with all data stored in Parquet format. The dataset encompasses various observation and action-related features, including images (specifically agentview_image and robot0_eye_in_hand_image), states (such as motor positions and gripper states), actions (comprising motor positions and gripper states), rewards, done flags, timestamps, frame indices, episode indices, and task indices, among others.

原始信息汇总

数据集概述

基本信息

• 许可证: Apache-2.0
• 任务类别: 机器人学 (robotics)
• 标签: LeRobot
• 代码库版本: v2.1

数据集结构

• 总集数: 1000
• 总帧数: 108233
• 总任务数: 1
• 总视频数: 0
• 总块数: 1
• 块大小: 1000
• 帧率 (fps): 10
• 训练集分割: 0:1000

数据文件路径

• 数据路径: data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet
• 视频路径: videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4

特征描述

• observation.images.agentview_image
  • 数据类型: 图像
  • 形状: [84, 84, 3]
  • 维度名称: ["height", "width", "channels"]
• observation.images.robot0_eye_in_hand_image
  • 数据类型: 图像
  • 形状: [84, 84, 3]
  • 维度名称: ["height", "width", "channels"]
• observation.state
  • 数据类型: float32
  • 形状: [9]
  • 维度名称:
    • motors: ["x", "y", "z", "rx", "ry", "rz", "rw", "gripper_1, gripper_2"]
• action
  • 数据类型: float32
  • 形状: [7]
  • 维度名称:
    • motors: ["x", "y", "z", "roll", "pitch", "yaw", "gripper"]
• next.reward
  • 数据类型: float32
  • 形状: [1]
• next.done
  • 数据类型: bool
  • 形状: [1]
• timestamp
  • 数据类型: float32
  • 形状: [1]
• frame_index
  • 数据类型: int64
  • 形状: [1]
• episode_index
  • 数据类型: int64
  • 形状: [1]
• index
  • 数据类型: int64
  • 形状: [1]
• task_index
  • 数据类型: int64
  • 形状: [1]