While I think that this is a somewhat useful rough overview, it should have had some more details about the actual differences and in particular how to accomplish the same thing with some code examples in each library.

On a side note: I know that I can also link tensorflow directly to a C++ application instead of using it with python. Can the same be done with pytorch?