Learning Visual Question Answering by Bootstrapping Hard Attention | Read Paper on Bytez