gsethia08/part1-color

该数据集是一个机器人学数据集，专注于手部机器人操作任务。它包含1998个训练episodes，总计824,084帧数据，帧率为30fps。数据集结构丰富，包括以下特征：观察部分有第一人称视角的摄像头视频（分辨率1080x1920，RGB格式）和机器人状态（48维浮点数，涵盖左右手腕、手指关节的位置和旋转信息）；动作数据为48维浮点数；此外还提供摄像头外参（16维）、内参（9维）、语言指令（字符串）、时间戳、帧索引、episode索引、索引和任务索引。数据集以Apache 2.0许可证发布，使用LeRobot工具创建，适用于机器人控制与学习任务。This is a robotics dataset focused on dexterous robotic manipulation tasks. It includes 1998 training episodes, with a total of 824,084 frames at 30 frames per second (fps). The dataset has a rich structure with the following features: the observation section contains first-person camera videos (resolution 1080x1920, RGB format) and robot state data (48-dimensional floating-point vectors covering the position and rotation information of the left and right wrists and finger joints); the action data is 48-dimensional floating-point vectors. Additionally, it provides camera extrinsic parameters (16-dimensional), camera intrinsic parameters (9-dimensional), language instructions (strings), timestamps, frame indices, episode indices, indices, and task indices. The dataset is released under the Apache 2.0 license, developed using the LeRobot toolkit, and is suitable for robotic control and machine learning tasks.