Simple Baselines

AbstractConventional multimodal tasks, such as caption generation and visual question answering, have allowed machines to understand an image by describing or being asked about it in natural language, often via …...