is it possible to add new persons to group afterwards, without re-submitting and re-training everything every time a new person is added?
The sample application (https://github.com/Microsoft/Cognitive-Face-Python) requires images per person in separate directories, to be submitted as a group.
You can add persons to a group anytime you want as long as you don't exceed the limits. However in order for the new persons (or the faces you added for the existing persons) to be used in the "Identify" process you need to call "train". Otherwise newly added persons (or faces) won't be found.