I don't think OpenCV really solved computer vision to be fair. There's definitely no model out there that can do image-based question & answering as well as a human can, or interpret the contents of an image (parse it, if you will) in an accurate way, with the exception of very few special cases.