USCOCO

A test set of grammatically correct sentences and layouts (visual “imagined” situations), called Unexpected Situations of Common Objects in Context (USCOCO) describing compositions of entities and relations that are unlikely to be found in MS COCO.

Contains 2.4K (caption, layout) pairs, where a layout is a sequence of bounding boxes in COCO format (normalized to 1.0 image width and height).

Can be used to test whether layout generation models generalize compositionally.