Research map / 3D vision / Human pose estimation
Learning Neural Volumetric Pose Features for Camera Localization
Jingyu Lin, Jiaqi Gu, Bojian Wu, Lubin Fan, Renjie Chen, Ligang Liu, Jieping Ye
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 |
|---|---|---|---|
| supervise with PoseMap features (L_posemap) instead of no pose feature (ablation) | PMNet (DFNet baseline) · Kings (Cambridge) | median translation error (m): 1.03 -> 0.68 | Improved |
| use raw NeRF feature map (NeRFMap) supervision instead of PoseMap (ablation) | PMNet · Kings (Cambridge) | median rotation error (deg): 2.94 -> 5.76 | Hurt |
| add L_image feature adapter loss (ablation) | PMNet · Kings (Cambridge) | median translation error (m): 0.75 -> 0.68 | Improved |
| add RVS loss for novel view augmentation (ablation) | PMNet · Kings (Cambridge) | median translation error (m): 0.94 -> 0.68 | Improved |
| add L_posemap during online refinement with unlabelled data (ablation) | PMNet_ud · Kings (Cambridge) | median translation error (m): 0.39 -> 0.31 | Improved |
| PMNet_ud versus DFNet_dm using unlabelled data | PMNet_ud · Cambridge Landmarks (average) | median translation error (m): 0.39 -> 0.31 | Improved |
| PMNet_ud versus DFNet_dm on outlier test views | PMNet_ud · Cambridge Kings top-30 outliers | mean translation error (m): 1.94 -> 0.85 | Improved |
| reduce training data to 25 and 50 percent (ablation) | PMNet versus DFNet · Kings (Cambridge) | accuracy | Improved |
| add PoseMap loss versus DFNet, cost in APR training time | PMNet · Kings training time | APR training hours: 6.0 -> 7.5 | Hurt |
| PMNet without post-processing versus iterative refinement methods | PMNet · Cambridge Landmarks (average) | median translation error (m): 0.90 vs 0.35 | Hurt |
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.
- DFNet — DFNet: Enhance Absolute Pose Regression with Direct Feature Matching (ECCV 2022) · better on 3 of 3 shared results
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
- Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization (CVPR 2025 · similarity 0.89)
- Map-Relative Pose Regression for Visual Re-Localization (CVPR 2024 · similarity 0.89)
- 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 0.87)
- Neural Refinement for Absolute Pose Regression with Feature Synthesis (CVPR 2024 · similarity 0.87)
- HyperPose: Hypernetwork-Infused Camera Pose Localization and an Extended Cambridge Landmarks Dataset (CVPR 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.
