Research map / 3D vision / Human pose estimation
Back to the Feature: Learning Robust Camera Localization From Pixels To Pose
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| learn multiscale features and unroll the optimizer instead of using photometric alignment (ablation) | PixLoc · CMU validation | AUC at 25cm: 1.2 -> 49.2 | Helped |
| add confidence weights to the alignment (ablation) | PixLoc · CMU validation | AUC at 25cm: 49.2 -> 53.3 | Helped |
| learn per-parameter damping (ablation) | PixLoc · CMU validation | AUC at 25cm: 53.3 -> 59.8 | Helped |
| reduce feature dimension from 128 to 16 (ablation) | PixLoc · CMU validation | AUC at 25cm: 59.8 -> 50.4 | Hurt |
| use RGB-D depth points instead of SfM points (ablation) | PixLoc · 7Scenes | recall at (5cm, 5°): 69.5 -> 75.7 | Helped |
| average poses from the top retrieved images rather than select top-1 (ablation) | PixLoc · Aachen Day-Night Day | recall at (25cm, 2°): 61.7 -> 64.3 | Helped |
| average poses from the top retrieved images rather than select top-1 (ablation) | PixLoc · Aachen Day-Night Night | recall at (25cm, 2°): 46.9 -> 51.0 | Helped |
| train on CMU rather than MegaDepth (ablation) | PixLoc · Extended CMU Seasons — Urban | recall at (25cm, 2°): 78.3 -> 91.9 | Helped |
| train on CMU rather than MegaDepth (ablation) | PixLoc · Aachen Day-Night — Day | recall at (25cm, 2°): 68.0 -> 54.4 | Hurt |
| refine hloc poses with PixLoc | hloc · Extended CMU Seasons — Urban | recall at (25cm, 2°): 95.5 -> 96.9 | Helped |
| refine hloc poses with PixLoc | hloc · RobotCar Seasons — Day | localization recall | No clear effect |
| refine hloc poses with PixLoc | hloc · Aachen Day-Night | localization recall at 0.25m | Hurt |
Closest papers by meaning
- Learning To Detect Scene Landmarks for Camera Localization (CVPR 2022 · similarity 0.91)
- Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization (CVPR 2025 · similarity 0.90)
- Learning Less Is More - 6D Camera Localization via 3D Surface Regression (CVPR 2018 · similarity 0.89)
- AtLoc: Attention Guided Camera Localization (AAAI 2020 · similarity 0.88)
- Understanding the Limitations of CNN-Based Absolute Camera Pose Regression (CVPR 2019 · similarity 0.88)
- PIDLoc: Cross-View Pose Optimization Network Inspired by PID Controllers (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.