CoRL2026-CSI/SO101-MA-MultiTask-700epi_10fps

该私有数据集将七个100个episode的SO101 10 fps数据集合并为一个LeRobot v3.0多任务数据集。具体信息包括：机器人类型为so_follower，格式为LeRobot v3.0，总episodes为700，总frames为358,006，帧率为10 FPS，包含7个任务标签，每个任务100个episodes。必需的数字字段是observation.state [6]和action [6]，必需的图像字段是observation.images.top和observation.images.left_wrist。数据集涵盖多个机器人任务，例如堆叠RGB块在蓝色盘子上、分类RGB块到匹配的盘子、分发巧克力派到每个盘子、打开顶部抽屉、关闭锅、拾取放置基准测试以及将勺子放在面包和麦片之间。数据集经过模式归一化处理，保留了关键字段，并验证了数据完整性和一致性。This private-first dataset merges seven 100-episode SO101 10 fps datasets into a single LeRobot v3.0 multitask dataset. Details include: Robot is so_follower, format is LeRobot v3.0, episodes are 700, frames are 358,006, FPS is 10, task labels are 7 dataset-level task labels with 100 episodes each. Required numeric fields are observation.state [6] and action [6], required image fields are observation.images.top and observation.images.left_wrist. The dataset covers multiple robotics tasks such as stacking RGB blocks on a blue dish, sorting RGB blocks to matching plates, distributing chocolate pie to each plate, opening a top drawer, closing a pot, pick-and-place baseline, and placing a spoon between bread and cereal. It uses schema normalization to retain essential fields and has been validated for data integrity and consistency.