tmeynier/so100_test_3

该数据集是一个机器人学习数据集，包含了50个视频文件，共计4995帧，分为25个剧集。每个视频都有对应的Parquet格式数据文件，包含动作、状态、图像等信息。数据集的结构详细定义了各种特征的数据类型、形状和名称，例如动作包括主肩部旋转、主肩部提升、主肘部弯曲等六个维度。所有视频的帧率为30fps，分辨率为480x640，采用av1编码。

This dataset is a robotics learning dataset containing 50 video files, totaling 4995 frames, divided into 25 episodes. Each video has corresponding Parquet data files, including information such as actions, states, and images. The dataset structure defines in detail the data type, shape, and name of various features, for example, actions include six dimensions such as main shoulder rotation, main shoulder lift, and main elbow flexion. All videos have a frame rate of 30fps, a resolution of 480x640, and are encoded using av1.