zijian2022/eval_bgb1

这是一个用于机器人研究的的数据集，包含5个总的剧集，2917个总帧数，1个任务，10个视频和1个数据块。数据集以Parquet格式存储，并包含行动、状态、视频画面、时间戳、帧索引、剧集索引等特征。其中，行动和状态特征包括机器人的主要关节角度和夹爪开合程度。视频画面特征包括来自笔记本电脑和手机的画面，均为480p分辨率，30帧每秒，无音频。

This is a dataset for robotics research, containing 5 total episodes, 2917 total frames, 1 task, 10 videos, and 1 chunk. The dataset is stored in Parquet format and includes features such as actions, states, video frames, timestamps, frame indices, and episode indices. The action and state features include angles of the robots main joints and the degree of grip opening and closing. The video frame features include images from a laptop and a phone, both at 480p resolution, 30 frames per second, without audio.