Research map / 3D vision / Human pose estimation
Combining Absolute and Semi-Generalized Relative Poses for Visual Localization
Vojtech Panek, Torsten Sattler, Zuzana Kúkelová
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 multiplied MSAC scores instead of multiplied inlier counts for pose scoring (ablation) | Adaptive · Aachen Day-Night v1.1 night, SfM, 1/20 database images | recall @ (0.25m, 2deg): 23.0 -> 40.8 | Improved |
| use summed MSAC scores instead of summed inlier counts (ablation) | Adaptive · Aachen Day-Night v1.1 night, SfM, 1/20 database images | recall @ (0.25m, 2deg): 8.4 -> 39.3 | Improved |
| use hybrid instead of split local optimization (ablation) | Adaptive (mult. msac) · Aachen Day-Night v1.1 night, SfM, 1/20 database images | recall @ (0.25m, 2deg): 36.1 -> 40.8 | Improved |
| use hybrid instead of split local optimization (ablation) | Adaptive (sum. msac) · Aachen Day-Night v1.1 night, SfM, 1/20 database images | recall @ (0.25m, 2deg): 35.1 -> 39.3 | Improved |
| use hybrid instead of split local optimization (ablation) | Adaptive (mult. msac) · Aachen Day-Night v1.1 night, SfM, 1/50 database images | recall @ (0.25m, 2deg): 13.6 -> 15.7 | Improved |
| use adaptive selection of 2D-2D and 2D-3D poses instead of P3P alone on sparse scenes | Adaptive vs P3P · Aachen Day-Night v1.1 night, SfM, 1/50 database images | recall @ (0.25m, 2deg): 7.3 -> 15.7 | Improved |
| use adaptive selection instead of E5+1 alone on dense scenes | Adaptive vs E5+1 · Aachen Day-Night v1.1 night, SfM, 1/1 database images | recall @ (0.25m, 2deg): 69.1 -> 78.5 | Improved |
| use Adaptive instead of P3P on sparse CMU Seasons | Adaptive vs P3P · Extended CMU Seasons urban, 1/11 database images | recall @ (0.25m, 2deg): 63.6 -> 70.5 | Improved |
| use Adaptive instead of P3P with full database | Adaptive vs P3P · Extended CMU Seasons park, 1/1 database images | recall @ (0.25m, 2deg): 90.7 -> 89.5 | No clear effect |
| use Adaptive instead of P3P on Gangnam Station | Adaptive vs P3P · Gangnam Station B1 | recall @ (0.1m, 1deg): 40.4 -> 41.4 | Improved |
| select pose with alpha 0.8 instead of alpha 1.0 (ablation) | Select · Aachen Day-Night v1.1 night, SfM, 1/1 database images | recall @ (0.25m, 2deg): 57.6 -> 73.3 | Improved |
| add 3D point filtering for noisy Metric3D depths (ablation) | Adaptive (sum. msac, hybrid) · Aachen day, Metric3D depth | recall @ (0.25m, 2deg): 51.6 -> 60.7 | Improved |
Closest papers by meaning
- A Guide to Structureless Visual Localization (IJCV 2026 · similarity 0.94)
- Efficient & Effective Prioritized Matching for Large-Scale Image-Based Localization (TPAMI 2016 · similarity 0.90)
- Learning to Produce Semi-dense Correspondences for Visual Localization (CVPR 2024 · similarity 0.88)
- Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization (CVPR 2025 · similarity 0.88)
- 3D-Aware Hypothesis & Verification for Generalizable Relative Object Pose Estimation (ICLR 2024 · similarity 0.88)
- Enhancing Visual Localization with Cross-Domain Image Generation (ICML 2025 · similarity 0.87)
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.
