Research map / 3D vision / Depth and stereo
On the Limits of Pseudo Ground Truth in Visual Camera Re-Localisation
Eric Brachmann, Martin Humenberger, Carsten Rother, Torsten Sattler
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 |
|---|---|---|---|
| switch reference pGT from original SLAM to SfM | Active Search · 7Scenes average | images localised within 5cm,5deg (%): 68.7 -> 98.5 | Improved |
| switch reference pGT from original SLAM to SfM | Active Search · 7Scenes Pumpkin | images localised within 5cm,5deg (%): 34.1 -> 99.6 | Improved |
| switch reference pGT from original SLAM to SfM | Active Search · 7Scenes Red Kitchen | images localised within 5cm,5deg (%): 45.1 -> 99.8 | Improved |
| switch reference pGT from original SLAM to SfM | DSAC* (+D) · 7Scenes average | images localised within 5cm,5deg (%): 94.7 -> 95.3 (rank first to last) | Hurt |
| switch reference pGT from original SLAM to SfM | DenseVLAD+R2D2 (+D) · 7Scenes average | images localised within 5cm,5deg (%): 80.2 -> 92.9 | Mixed |
| switch reference pGT from original SLAM to SfM | hLoc · 7Scenes average | images localised within 5cm,5deg (%): 76.8 -> 95.7 | Improved |
| switch reference pGT from original SLAM to SfM | DSAC* · 7Scenes average | images localised within 5cm,5deg (%): 82.9 -> 97.8 | Improved |
| use +BA refined test poses instead of original poses | Active Search · 7Scenes average | images localised within 5cm,5deg (%): 68.7 -> 79.5 | Improved |
| use +BA refined test poses instead of original poses | DSAC* (+D) · 7Scenes average | images localised within 5cm,5deg (%): 94.7 -> 93.2 | Hurt |
| use +BA refined test poses instead of original poses | DenseVLAD+R2D2 (+D) · 7Scenes average | images localised within 5cm,5deg (%): 80.2 -> 85.0 | Improved |
| add depth input (+D) to DSAC* | DSAC* · 7Scenes (original SLAM pGT) | images localised within 5cm,5deg (%): 82.9 -> 94.7 | Improved |
| add depth input (+D) to DSAC* | DSAC* · 7Scenes (SfM pGT) | images localised within 5cm,5deg (%): 97.8 -> 95.3 | Hurt |
Closest papers by meaning
- Real-Time RGB-D Camera Pose Estimation in Novel Scenes Using a Relocalisation Cascade (TPAMI 2019 · similarity 0.89)
- Beyond Controlled Environments: 3D Camera Re-Localization in Changing Indoor Scenes (ECCV 2020 · similarity 0.89)
- Investigating the Role of Image Retrieval for Visual Localization (IJCV 2022 · similarity 0.88)
- Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization (CVPR 2025 · similarity 0.87)
- Map-Relative Pose Regression for Visual Re-Localization (CVPR 2024 · similarity 0.86)
- Depth2Pose: A Pose-Based Benchmark for Monocular Depth Estimation without Ground-Truth Depth (NeurIPS 2026 · 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.
