A Basic Functions

Each question in PTR is associated with a functional program built from a set of basic functions. A.1 Data Types Our basic functional building blocks operate on values of the following types: Object: A single object in the scene. Part-level functions are listed in Table 4. B have certain spatial relationships. For NS-VQA, we first use Mask-RCNN to propose segmentations for objects and parts. If an object is unstable, possible changes (to_left, to_right, to_front, to_behind) are predicted.