Recurrent models of visual attention can be used to answer some of the challenges we covered in the earlier section. These models use the hard attention method, as covered in an earlier (Types of attention) section. Here we use one of the popular variants of recurrent models of visual attention, the Recurrent Attention Model (RAM).
As covered earlier, hard attention problems are non-differentiable and have to use RL for the control problem. The RAM thus uses RL for this optimization.
A recurrent model of visual attention does not process the entire image, or even a sliding-window-based bounding box, at once. It mimics the human eye and works on the concept of Fixation of Gaze at different locations of ...