Deeply Supervised Multimodal Attentional Translation Embeddings for Visual Relationship Detection | IEEE Conference Publication | IEEE Xplore