Robotic Vision-Language Planning (ViLa) Tasks

该数据集包含了16个长期视角的操作任务，旨在评估ViLa在理解常识知识、目标指定上的灵活性和利用视觉反馈方面的表现。此外，该数据集还包括了评估对空间布局和对象属性理解的任务，并提供了与基准方法成功率比较的数据。数据集在多种场景和光照条件下具有多样性，其任务是机器人操作任务规划。

This dataset comprises 16 long-horizon manipulation tasks intended to evaluate ViLa's flexibility in commonsense knowledge understanding, goal specification, and utilization of visual feedback. Additionally, it includes tasks that assess understanding of spatial layouts and object attributes, and provides data comparing success rates against baseline methods. The dataset is diverse across multiple scenarios and varying lighting conditions, with its tasks focused on robotic manipulation task planning.