Integrate Saliency model

[atar13]

Subtask of #15.

Depends on #16.

At a high level, this component should be able to take in a full resolution image (after image processing), pass it through the saliency model, and gather model predictions (such as the bounding boxes around a target, or a target is emergent or not).

Here is a reference of how this used to work in the python version of the OBC.

Testing

TODO

Documentation

TODO