Visual Relationship Detection with Language Priors

VRD is a dataset for visual relationship detection, containing 100 annotated relationships.

BibTex: