Trained models for ocr tensor flow

Question

I am starting a tensor flow course in udacity, and at the same time I am browsing this topic on the Internet.

I believe that typical use cases are well resolved already, in the best way that I can achieve on my own. In other words, in some place there are ready-made models for ordinary cases, ready to use. I found zooModels that if I understood correctly, this is what I am looking for. but I cannot understand that there is no published ocr model that can recognize a number in an image:

image example

Do I need to train my own model? Is there a repository that I don't know?