Research map / 3D vision / Human pose estimation
Gravity-aware partially calibrated absolute pose estimation from affine- or rotation-covariant features
Marcus Valtonen Örnhag, Alberto Jaenal Gálvez, Stefan Adalbjörnsson
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use proposed UP1PfAC instead of P4Pf for unknown-focal localization | GC-RANSAC pipeline · Cambridge GreatCourt (SuperPoint+LightGlue) | median position error (cm): 61.5 -> 57.7 | Improved |
| use proposed UP1PfAC instead of P4Pf for unknown-focal localization | GC-RANSAC pipeline · Cambridge OldHospital (SIFT+NN) | median position error (cm): 51.1 -> 48.4 | Improved |
| use semi-calibrated UP1PfAC instead of calibrated UP2P (known f) | GC-RANSAC pipeline · Cambridge GreatCourt (SuperPoint+LightGlue) | median position error (cm): 31.6 -> 57.7 | Hurt |
| use UP1PfAC instead of P4Pf for rotation | GC-RANSAC pipeline · Cambridge OldHospital (SuperPoint+LightGlue) | median rotation error (deg): 0.42 -> 0.38 | Improved |
| use UP1PfAC instead of P4Pf on Aachen night images | GC-RANSAC pipeline · Aachen Night (SuperPoint+LightGlue, 0.25m/2deg) | recall (%): 54.5 -> 60.2 | Improved |
| use UP2PfORI instead of P4Pf on Aachen night images | GC-RANSAC pipeline · Aachen Night (SuperPoint+LightGlue, 0.25m/2deg) | recall (%): 54.5 -> 62.3 | Improved |
| use UP1PfAC instead of P4Pf on Aachen day images | GC-RANSAC pipeline · Aachen Day (SuperPoint+LightGlue, 0.25m/2deg) | recall (%): 46.1 -> 48.5 | Improved |
| use semi-calibrated UP1PfAC instead of calibrated UP2P | GC-RANSAC pipeline · Aachen Day (SuperPoint+LightGlue, 0.25m/2deg) | recall (%): 71.8 -> 48.5 | Hurt |
| use UP1PfAC instead of P4Pf for focal length estimation | GC-RANSAC pipeline · Aachen (SIFT+NN) | focal length error: 0.019 -> 0.017 | Improved |
| use 2-point UP2PfORI instead of 1-point UP1PfAC | proposed solvers · Cambridge and Aachen | accuracy: 2149 ns vs 2586 ns | Mixed |
| use UP1PfAC instead of calibrated UP2P in runtime | solver · Solver runtime | median time (ns): 484 -> 2586 | Hurt |
| use UP1PfAC instead of P3.5Pf in runtime | solver · Solver runtime | median time (ns): 19118 -> 2586 | Improved |
Closest papers by meaning
- Absolute Pose from One or Two Scaled and Oriented Features (CVPR 2024 · similarity 0.90)
- Globally Optimal Relative Pose Estimation With Gravity Prior (CVPR 2021 · similarity 0.89)
- P1AC: Revisiting Absolute Pose From a Single Affine Correspondence (ICCV 2023 · similarity 0.88)
- A Complete Solution to Generalized Relative Pose Estimation From Affine Correspondences (TPAMI 2026 · similarity 0.87)
- Vanishing Point Estimation in Uncalibrated Images with Prior Gravity Direction (ICCV 2023 · similarity 0.86)
- Relative Pose Estimation through Affine Corrections of Monocular Depth Priors (CVPR 2025 · similarity 0.86)
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.
