ltguo19/VSUA-Captioning

coco_pred_sg_rela.npy contains 3 dicts, it confused me a lot

kangkang59812 opened this issue · 0 comments

coco_pred_sg_rela.npy contains the vocabularies for the object, attribute and relation labels.
While it has keys dict_keys(['i2w', 'rela_dict', 'w2i']),
i2w-->object
w2i-->attribute
rela_dict-->relation
Is it right? And the three dicts all include noun,verb,and adj. It make me confused which one is right . The object and location seems in a same dict,but which one?