physical-intelligence/libero

该数据集是由四个单独的Libero数据集（Libero-Spatial、Libero-Object、Libero-Goal和Libero-10）组合而成，这些数据集来源于HuggingFace的openvla/modified_libero_rlds并转换为LeRobot格式。数据集包含了1693个剧集，共有273465帧，40个任务。数据集使用的是Panda型机器人，并以10帧每秒的帧率进行记录。数据集的结构包括图像、手腕图像、状态、动作、时间戳、帧索引、剧集索引、索引和任务索引等特征。数据集的授权协议为CC-BY 4.0。

This dataset is a combination of four individual Libero datasets: Libero-Spatial, Libero-Object, Libero-Goal, and Libero-10, sourced from HuggingFaces openvla/modified_libero_rlds and converted into the LeRobot format. The dataset contains 1693 episodes, totaling 273465 frames and 40 tasks. It uses the Panda robot type and is recorded at a frame rate of 10 frames per second. The structure of the dataset includes features such as images, wrist images, states, actions, timestamps, frame indices, episode indices, indices, and task indices. The dataset is licensed under CC-BY 4.0.

搜集汇总

背景与挑战

背景概述

The 'physical-intelligence/libero' dataset is a robotics-focused collection combining four Libero datasets into the LeRobot format, featuring image and time-series data for robotics tasks. It includes 273,465 rows of data with detailed metadata and is licensed under CC-BY 4.0, supporting research in lifelong robot learning.