{"id":"arxiv/2610.31797","url":"https://arxiv.org/abs/2610.31797","tags":["cs.CV","cs.RO","math.OC"],"score":0,"title":"PnP*: A certifiable solver for Perspective-n-Point","domain":"cs.CV","source":"arxiv","authors":["Bang-Shien Chen","Caitrin E. Eaton","David M. Rosen"],"reviews":0,"subject":"cs.CV","summary":"Estimating a camera pose from 3D-2D correspondences - Perspective-n-Point - is one of the most fundamental problems in computer vision. Existing methods either solve a minimal problem or replace the objective with an easier one. Surprisingly, no scalable certifiable solver for the actual Maximum Likelihood Estimation (MLE) exists. We propose PnP*, the first practical solver that guarantees efficient computation and returns a certifiable globally optimal solution. We prove that the underlying convex relaxation is exact under moderate noise, which to our knowledge is the first stability…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-10-22","reactions":{}}