Multi modal and few shot learning, Visual Scene Understanding, Scene Graphs