Text this: A Novel Hybrid Deep Learning Framework for Image Captioning Using Combined Semantic and Object-Level Features