Aligning Images and Text with Semantic Role Labels for Fine-Grained Cross-Modal Understanding

14769

Knowledge Triples