saipuneethgottam/sweep_dataset_merged_normalized_v2

该数据集是使用LeRobot创建的机器人数据集，专门用于机器人控制任务。数据集包含200个episodes，总计137050帧，数据以parquet文件格式存储，视频文件以mp4格式存储，帧率为30fps。数据集特征包括：动作数据（6个关节位置：肩部平移、肩部提升、肘部弯曲、手腕弯曲、手腕旋转和夹爪位置）、状态观测数据（与动作相同的6个关节位置）、以及来自三个视角的图像观测：前视角（分辨率720x1280，3通道）、左视角（分辨率720x1280，3通道）和手腕视角（分辨率480x640，3通道）。所有视频均使用av1编解码器，像素格式为yuv420p，不包含音频。数据集还包含时间戳、帧索引、episode索引、索引和任务索引等元数据。数据集适用于训练和评估机器人模仿学习或强化学习模型。

This dataset is created using LeRobot and is designed for robotics control tasks. It contains 200 episodes with a total of 137,050 frames, stored in parquet format for data and mp4 format for videos, with a frame rate of 30fps. The dataset features include: action data (6 joint positions: shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper), state observation data (same 6 joint positions as actions), and image observations from three perspectives: front view (resolution 720x1280, 3 channels), left view (resolution 720x1280, 3 channels), and wrist view (resolution 480x640, 3 channels). All videos use the av1 codec with yuv420p pixel format and no audio. The dataset also includes metadata such as timestamp, frame index, episode index, index, and task index. It is suitable for training and evaluating robot imitation learning or reinforcement learning models.