Awesome example! I was able to make it work on my own "solution", but there is something missing... How do you apply this to a specific use case? Example: I need to recognize text from images I already have as cv2 objects. - How would I be able to make the "prediction_model.predict" method work with a single OpenCV image I have in PIL or cv2 object? This is maybe a dump question, but I'm pretty new in this AI world, and of course don't fully understand everything... Hope you are able to answer. :)