taccap-g1-stapler-using-0724

该数据集是一个机器人技术领域的数据集，由LeRobot创建，专门用于双TacCap夹爪机器人（bi_taccap_gripper）的操作任务。数据集包含72个训练集片段，总计378,251帧，帧率为30fps。数据集中包含动作和观测特征：动作特征是一个20维浮点向量，表示左右夹爪的工具中心点（TCP）的x、y、z位置、六个旋转参数以及夹爪位置；状态观测特征也是一个20维浮点向量，结构与动作特征相同。观测部分还包括多个视觉数据：左右腕部摄像头视频（分辨率为480x640，3通道），以及左右触觉传感器的视频（分辨率为400x700，3通道），所有视频均以h264编码，yuv420p像素格式，无音频。此外，数据集包含时间戳、帧索引、片段索引、任务索引等元数据，用于追踪数据序列。该数据集旨在支持机器人学习任务，特别是涉及触觉感知和视觉反馈的操作。This is a robotics dataset created by LeRobot, specifically tailored for manipulation tasks with the dual TacCap gripper (bi_taccap_gripper). The dataset consists of 72 training episodes, totaling 378,251 frames with a frame rate of 30 fps. It includes two types of features: action features and observation features. The action features are 20-dimensional floating-point vectors, representing the x, y, z positions, six rotation parameters, and gripper position of the Tool Center Point (TCP) for both the left and right grippers. The state observation features are also 20-dimensional floating-point vectors with an identical structure to the action features. The observation part additionally contains multiple visual data modalities: videos from the left and right wrist cameras (resolution 480×640, 3-channel), and videos from the left and right tactile sensors (resolution 400×700, 3-channel). All videos adopt h264 encoding, use the yuv420p pixel format, and contain no audio. Furthermore, the dataset provides metadata including timestamps, frame indices, episode indices, and task indices to facilitate tracking of data sequences. This dataset aims to support robotics learning tasks, particularly manipulation tasks involving tactile perception and visual feedback.