Scene understanding, which blends computer vision with natural language processing skills, includes image caption, which automatically generates natural language descriptions based on the content observed in a picture.