eval_molmoact_cup_stacking_in-distribution
收藏资源简介:
该数据集由LeRobot项目创建,是一个机器人领域的数据集,采用Apache-2.0许可证。数据集包含35个episodes,总计32048帧,对应1个任务,数据以parquet文件格式组织,总数据文件大小约100MB,视频文件大小约200MB,帧率为30fps。数据集仅提供训练集(索引0-35)。数据特征包括:动作(action)和观测状态(observation.state)均为14维浮点数组,分别表示左右机械臂的6个关节位置和1个夹爪位置;观测图像包含三路视频(右、左、顶视角),每路视频分辨率为360x640,3通道,采用AV1编码,无音频;此外还包含时间戳、帧索引、episode索引、全局索引和任务索引等元数据字段。数据集适用于机器人控制、模仿学习或强化学习等任务,但具体任务定义和背景信息未在README中说明。
This dataset is created by the LeRobot project and is a dataset in the robotics domain, licensed under Apache-2.0. It contains 35 episodes, totaling 32048 frames, corresponding to 1 task. The data is organized in parquet file format, with a total data file size of approximately 100MB and a video file size of about 200MB, at a frame rate of 30fps. The dataset only provides a training set (indices 0-35). Data features include: action and observation.state, both being 14-dimensional floating-point arrays, representing the 6 joint positions and 1 gripper position for the left and right robotic arms; observation images consist of three video streams (right, left, and top views), each with a resolution of 360x640, 3 channels, encoded in AV1 without audio; additionally, metadata fields such as timestamp, frame index, episode index, global index, and task index are included. The dataset is suitable for tasks like robot control, imitation learning, or reinforcement learning, but specific task definitions and background information are not described in the README.




