A large, realistic multimodal dataset consisting of real personal photos and crowd-sourced questions/answers. Source: MemexQA: Visual Memex Question Answering