gsethia08/part3-paint_clean_brush

该数据集是一个用于机器人学习的多模态数据集，专注于手部机器人操作任务。数据集包含1894个训练片段，总计968485帧，覆盖18个不同任务。数据采集自手部机器人，包括以下主要特征：1）观测数据：来自ego相机的视频数据（1080x1920分辨率，30fps，RGB格式），以及手部状态观测（48维向量，详细记录左右手腕的位置、旋转信息以及各手指的坐标）。2）相机参数：包含相机的外参（16维）和内参（9维）。3）动作数据：48维的动作向量。4）语言指令：文本形式的任务指令。5）元数据：时间戳、帧索引、片段索引、任务索引等。数据集以Parquet格式存储，视频以MP4格式存储，适用于机器人控制、模仿学习等研究。This dataset is a multimodal dataset for robot learning, focusing on hand robot manipulation tasks. It contains 1894 training episodes, totaling 968485 frames, covering 18 different tasks. The data is collected from a hand robot and includes the following main features: 1) Observation data: video data from an ego camera (1080x1920 resolution, 30fps, RGB format), and hand state observations (48-dimensional vector detailing left and right wrist positions, rotations, and finger coordinates). 2) Camera parameters: extrinsic (16-dimensional) and intrinsic (9-dimensional) parameters. 3) Action data: 48-dimensional action vectors. 4) Language instructions: text-based task instructions. 5) Metadata: timestamps, frame indices, episode indices, task indices, etc. The dataset is stored in Parquet format, with videos in MP4 format, suitable for research in robot control, imitation learning, and related fields.