Towards Generating Coherent Stories from Image Sequences: A Computational Approach Using Suspension of Disbelief