[mwcampbell]

I agree that it's good to make our own accessibility where possible. However, computer vision seems like gross overkill for making computer software itself accessible. After all, the information you need is already in the computer; it's just not yet being exposed to the screen reader in a standard way. So processing a screenshot or image from a camera would be very wasteful, though it would work as a last resort. But other hacks to make software accessible? Absolutely! For example, one can imagine writing bits of JavaScript code to fill in gaps in the accessibility of specific websites and applications.