ABC-CNN: An Attention Based Convolutional Neural Network for Visual Question Answering

  VQA的相关应用及其挑战：   VQA is of great importance to many applications, including image retrieval, early education, and navigation for blind people as it provides user-specific information through the understanding of both the natural language questions and image content.VQA is a highly challenging problem as it requires the machine to understand natural language queries, extract semantic contents from images, and relate them in a unified framework。   本文提出question-guided attention map(QAM)，把QAM当作是潜在的信息，这些map并不需要为各种可能的搜索提供明确的标签。QAM是通过在空间图像特征图中搜索与问题语义相关的视觉特征生成的。这种搜索是通过configurable convolution neural network实现的，这个网路是利用feature map 与configurable convolutional kernel卷积而成。configurable convolutional kernel是一个特殊的卷积核，在把问题向量映射到视觉特征空间中过程中生成configurable convolutional kernel   框架图：      主要分为4个模块：图像特征提取、问题特征提取、Attention提取、Answer生成。