Image Segmentation

How to find objects as subimages in an image:

Basically, use classifying networks on suggested subboxes. Then there are some tricks layered on top of that idea, like using a netowkr to suggest possible subboxes. There exist implementations of these things in tensorflow, caffe and others.

One shot learning

Differentiate whether two pictures are of the same object using only the one image.

One-Shot Imitation learning