Is augmented reality a training system for computer vision? As in, Augmented systems use their data to help train computer vision algorithms, or is augmented reality computer vision itself?
Let me re-frame your question before answering:
Can augmented reality be a training system for computer vision?
Yes, why not! Also, I'm pretty sure the big companies who have their aug. reality products out there are secretly optimizing their CV algorithms(although we'll have no proof).
A very similar(but a different) example would be how Comma.ai's app pays users for recording their driving, and then uses it to train their own self-driving cars.
So, if that can be done, I don't see why augmented reality cannot be used to train CV algos. In fact, it's a much more richer, localized dataset.