Pick-and-Place and Bin Pushing Datasets

该数据集包含了针对抓取和放置任务以及推桶任务的原始演示，旨在通过视频演示来促进机器人任务的训练和元学习。数据集中包含了多种纹理和物体位置随机化，以创造多样化的训练场景。规模上，该数据集拥有1800个抓取和放置的原语以及56个推桶演示。任务目标是，从单一视频演示中学习复合视觉运动任务。

This dataset contains raw demonstrations for grasping-and-placing tasks and bucket-pushing tasks, aiming to facilitate robot task training and meta-learning via video demonstrations. It incorporates randomization of multiple textures and object positions to create diverse training scenarios. In terms of scale, this dataset includes 1800 grasping-and-placing primitives and 56 bucket-pushing demonstrations. The task objective is to learn complex visuomotor tasks from a single video demonstration.