Deep Learning for Computer Vision

Dense captioning 

Johnson et al., in the paper https://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/Johnson_DenseCap_Fully_Convolutional_CVPR_2016_paper.pdf, proposed a method for dense captioning. First, let's see some results, to understand the task:

Reproduced from Johnson et al.

As you can see, separate captions are generated for objects and actions in the image; hence the name; dense captioning. Here is the architecture proposed by Johnson et al.:

Reproduced from Johnson et al.

The architecture is essentially a combination ...