Hacker News new | ask | show | jobs
by rsp1984 3271 days ago
My guess is they estimate one or several dominant scene planes from the sparse triangulated feature points and get scale through incorporating accelerometer measurements.