nayana-c1.867c992ae9100904cc61c84591adf0f8f863e0387e30f0d0dbc86640e9f50a1a.ar.train.42657

The task

Detect the annotated document layout regions. Return only a JSON array of {"label":"text","bbox":[xmin,ymin,xmax,ymax]} objects. Use absolute pixel coordinates in the supplied image, with origin at its top left. Allowed labels: text, title, caption, table, image, formula. Include every region once; no extra keys or prose. These labels describe document regions, not arbitrary objects.