Research map / 3D vision / Human pose estimation
Camera Pose Auto-Encoders for Improving Pose Regression
Yoli Shavit, Yosi Keller
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 |
|---|---|---|---|
| add Fourier Features to the 4-layer MLP PAE (ablation) | PAE (PoseNet+MobileNet teacher) · KingsCollege | position error (m): 1.26 -> 1.15 | Improved |
| add Fourier Features to the 4-layer MLP PAE (ablation) | PAE (PoseNet+MobileNet teacher) · KingsCollege | orientation error (deg): 3.54 -> 3.58 | Hurt |
| use 4 layers instead of 2 in the MLP PAE (ablation) | PAE without Fourier Features · KingsCollege | position error (m): 1.27 -> 1.26 | No clear effect |
| use 4 layers instead of 2 in the MLP PAE (ablation) | PAE without Fourier Features · KingsCollege | orientation error (deg): 3.41 -> 3.54 | Hurt |
| add test-time position refinement with PAE-encoded train poses | MS-Transformer · Cambridge Landmarks (average) | position error (m): 1.28 -> 0.96 | Improved |
| add test-time position refinement with PAE-encoded train poses | MS-Transformer · 7Scenes (average) | position error (m): 0.18 -> 0.15 | Improved |
| add virtual relative pose regression refinement | MS-Transformer · Shop Facade | position error (m): 0.86 -> 0.62 | Improved |
| add virtual relative pose regression refinement | MS-Transformer · Heads | position error (m): 0.14 -> 0.10 | Improved |
| decode pose from student PAE instead of teacher APR image encoder | MS-Transformer · Cambridge KingsCollege | position error (m): 0.83 -> 0.90 | Hurt |
| decode pose from student PAE instead of teacher APR image encoder | MS-Transformer · 7Scenes Fire | orientation error (deg): 9.60 -> 9.31 | Improved |
| swap teacher APR backbone to EfficientNet (ablation) | PAE · KingsCollege | position error (m): 1.24 -> 0.88 | Improved |
| use the PAE student in place of the teacher with other backbones (ablation) | PAE · KingsCollege (PoseNet+ResNet50) | position error (m): 1.56 -> 1.50 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MapNet — MapNet: An Allocentric Spatial Memory for Mapping Environments (CVPR 2018) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning Multi-Scene Absolute Pose Regression With Transformers (ICCV 2021 · similarity 0.92)
- Neural Refinement for Absolute Pose Regression with Feature Synthesis (CVPR 2024 · similarity 0.90)
- Map-Relative Pose Regression for Visual Re-Localization (CVPR 2024 · similarity 0.89)
- Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization (CVPR 2025 · similarity 0.88)
- ConDo: Continual Domain Expansion for Absolute Pose Regression (AAAI 2025 · similarity 0.87)
- FAR: Flexible Accurate and Robust 6DoF Relative Camera Pose Estimation (CVPR 2024 · 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.
