I wonder if further research on this has been done. For example, what if some ML tagging mechanism is used to find the silhouette of interesting objects in the

I wonder if further research on this has been done. For example, what if some ML tagging mechanism is used to find the silhouette of interesting objects in the image (people, animals, traffics signs, etc), and then "freezing" them to prevent the energy function from operating on those areas, thus preserving those objects intact, while resizing the rest of the image.