Referring Expression Comprehension as Scene Graph Grounding

Efficient object recognition with linguistic references using Graph Neural Networks and CLIP.