Instruction and Scene Dataset

该数据集包含语言指令与相应演示任务的数据，这些任务在模拟环境中通过机械臂完成。数据集中的场景涉及3到5个不同类别和颜色的积木块，每个数据点包括初始场景、最终场景以及一条语言指令，且这些指令下没有任何子目标的监督。该数据集共包含5000个数据点，旨在支持语言引导的机器人操作任务。

This dataset contains data pairing language instructions with their corresponding demonstration tasks, which are completed by a robotic arm in a simulated environment. The scenarios in the dataset involve 3 to 5 building blocks of different categories and colors. Each data point includes the initial scene, the final scene, and a language instruction, with no sub-goal supervision provided for these instructions. This dataset consists of 5000 total data points and is intended to support language-guided robotic manipulation tasks.