RelCLIP: Adapting Language-Image Pretraining for Visual Relationship Detection via Relational Contrastive Learning

30891

Related Entities