Sure, don't have a reference on hand but this looks like it...

Sure, don't have a reference on hand but this looks like it covers the basics
https://courses.cs.washington.edu/courses/cse455/09wi/Lects/lect16.pdf

The wiki article for epipolar geometry also gives the general intuition for how it works
https://en.wikipedia.org/wiki/Epipolar_geometry

It is essentially the same method that was used by those ANU researchers to combine a DVS with a regular camera feed, with the added step of applying a sobel filter to the regular camera feed so that it could be matched with the DVS feed.

Also, just in case it wasn't clear, I wasn't arguing in favour of Elon's ML approach to inferring depth. Just the intuition behind reverse engineering how regular old humans drive. I think he will eventually end up using stereo vision at the end of the day.