Incidentally this thread leaves me thinking: It may be appealing to blind "merephone" users if the screenreader image viewer & camera had computer vision? Listing what's seen?

Theoretically it probably could be done, using plenty of energy & SD card storage.
In practice merephone's a big enough project without this, so if desired I'd want someone else to tackle it!

On the otherhand I'm keen to do the equivalent for maps! No complex algorithms there, just fancy datastructures!

4/3 Truly fin!