January 2018
Intermediate to advanced
310 pages
7h 48m
English
Xu et al., in the paper, https://arxiv.org/pdf/1502.03044.pdf, proposed a method for image captioning using an attention mechanism. The attention mechanism gives more weight to certain regions of the image than others. Attention also enables visualization, showing us where the model is focusing when it generates the next word. The proposed model is shown here:

First, CNN features are extracted from the image. Then, RNN with attention is applied to the image from which the words are generated.
Read now
Unlock full access