Attention mechanisms are now widely used in image captioning models. However, most attention models only focus on visual features. When generating syntax related words, little visual information is needed. In this case, these attention models could mislead the word generation. In this paper, we prop...
No comments yet
Be the first to share your thoughts!