Single-Modal Entropy based Active Learning for Visual Question Answering