Can the scale Struct2depth be recovered given the real focal length?

Question

Recently I read a great paper, struct2depth.

But as I noticed that the scale is normalized in this paper, I wonder:

• if one can recover the a