AI Paper Atlas

Research map / 3D vision / Human pose estimation

Polarimetric Relative Pose Estimation

Zhaopeng Cui, Viktor Larsson, Marc Pollefeys

ICCV 2019 · 3D vision › Human pose estimation · approach: Learned camera pose estimation · 28 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add polarimetric normal constraints (2-point solver) in place of the 5-point solver, initial estimaterelative pose estimation · Synthetic two-view (1000 trials)rotation angular error (deg): 6.10 -> 2.30Improved
use the 2-point polarimetric solver instead of 5-point, initial translationrelative pose estimation · Synthetic two-view (1000 trials)translation angular error (deg): 9.30 -> 3.25Improved
add the proposed polarimetric joint refinement instead of Sampson-only refinement (ablation)2-point solver · Synthetic two-view (1000 trials)rotation angular error (deg): 3.59 -> 1.80Improved
add the proposed polarimetric joint refinement instead of Sampson-only refinement, translation (ablation)2-point solver · Synthetic two-view (1000 trials)translation angular error (deg): 4.08 -> 2.52Improved
apply Sampson refinement after the 2-point solver2-point solver · Synthetic two-view (1000 trials)rotation angular error (deg): 2.30 -> 3.59Hurt
use 2-point solver instead of 5-point on real data TALLVASErelative pose estimation · Real polarization TALLVASErotation angular error (deg): 22.91 -> 12.74Improved
use 2-point solver instead of 5-point on real data TALLVASE, translationrelative pose estimation · Real polarization TALLVASEtranslation angular error (deg): 17.36 -> 13.03Improved
use 2-point solver instead of 5-point on real data VASErelative pose estimation · Real polarization VASErotation angular error (deg): 14.70 -> 14.44No clear effect
use 2-point solver instead of 5-point on real data BALLOONrelative pose estimation · Real polarization BALLOONrotation angular error (deg): 7.44 -> 6.87No clear effect
use 2-point solver, fewer samples needed per RANSAC hypothesisRANSAC · Synthetic two-view (1000 trials)RANSAC iterations: 17.8 -> 7.8Improved
use 2-point solver (SVD alignment) instead of 5-pointrelative pose estimation · Synthetic two-view (1000 trials)runtime (ms): 7.1 ms -> 14.3 msHurt
use per-point refractive indices instead of one shared index (ablation)2-point Optimized · Synthetic, per-point refractive indexrotation angular error (deg): 1.80 -> 3.66Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy