Hierarchical Multimodal LSTM for Dense Visual-Semantic Embedding | IEEE Conference Publication | IEEE Xplore