gsethia08/part1-clean_surface

该数据集是一个用于机器人手部操作任务的大规模数据集，由LeRobot创建。它包含4323个训练集片段，总计超过132万帧数据，覆盖17种不同的任务。数据集中包含丰富的特征：观测数据包括来自ego视角的相机视频（分辨率为1080x1920，30fps）和手部状态信息（如左右手腕和手指的坐标及旋转参数）；动作数据为48维浮点向量；此外还有相机内外参数、语言指令、时间戳、帧索引、片段索引等元数据。数据集以Apache 2.0许可证发布，适用于机器人学习、模仿学习等研究领域。This dataset is a large-scale dataset for robotic hand manipulation tasks, created using LeRobot. It contains 4323 training episodes with over 1.32 million frames in total, covering 17 different tasks. The dataset includes rich features: observations consist of ego-view camera videos (resolution 1080x1920, 30fps) and hand state information (e.g., coordinates and rotation parameters for left and right wrists and fingers); actions are represented as 48-dimensional float vectors; additionally, it includes camera intrinsics and extrinsics, language instructions, timestamps, frame indices, episode indices, and other metadata. Released under the Apache 2.0 license, it is suitable for research in robotics, imitation learning, and related fields.