wanted to model the rounded corners of my desk so i could design a mount for a small pendant light that sits in the corner. my first thought was a top down photo and a sketched curve over it, but that is not precise and i hate eyeballing a radius. the whole job turns out to come down to two pieces of math and a straightedge, and neither needs a fancy instrument.
the corner of a desk is just an arc that sits tangent to two straight edges. so you can extend the two edges out past where the corner is until they meet at an imaginary point, the corner if it had no radius at all. the arc's center sits along the 45 degree line from that point, and the closest distance from the point to the arc is about 0.4142 times the radius. that means the radius is 2.4142 times that measured gap, which is a dead simple relation to run backwards.
the backup method uses a chord and its sagitta. bridge a straightedge across the arc, measure how far the straight edge spans and the biggest gap underneath it, and the radius drops out of a formula. a longer chord means a smaller relative error, so you want the straightedge covering as much of the curve as it will touch. between the two methods you can cross check around a corner to the fraction of a millimeter, then model the exact fillet in cad instead of guessing.
the actual point of all of it is a part that will sit flush against the corner so a socket and a shade can hang over the desk. the geometry was the only real unknown, and it was never a photo problem at all, just a couple of clean measurements and the right relation between them.