Research map / 3D vision / Depth and stereo
Learning Multi-Scene Absolute Pose Regression With Transformers
Yoli Shavit, Ron Ferens, 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 |
|---|---|---|---|
| use Transformer encoders/decoders for multi-scene regression instead of per-scene FC layers (MSPN) | absolute pose regressor · Cambridge Landmarks (4 scenes, average) | position/orientation error: 2.47/5.34 -> 1.28/2.73 | Improved |
| use Transformer encoders/decoders for multi-scene regression instead of MSPN | absolute pose regressor · 7Scenes (average) | position/orientation error: 0.20/8.41 -> 0.18/7.28 | Improved |
| train one multi-scene model instead of single-scene models | MS-Transformer · Cambridge Landmarks | position/orientation error: 1.43/2.85 -> 1.28/2.73 | Improved |
| train one multi-scene model instead of single-scene models | MS-Transformer · 7Scenes | position/orientation error: 0.24/7.87 -> 0.18/7.28 | Improved |
| train jointly on both datasets (multi-dataset) instead of one dataset | MS-Transformer · Cambridge Landmarks | position/orientation error: 1.28/2.73 -> 1.50/2.57 | Mixed |
| train jointly on both datasets (multi-dataset) instead of one dataset | MS-Transformer · 7Scenes | position/orientation error: 0.18/7.28 -> 0.22/6.78 | Mixed |
| use EfficientNetB0 backbone instead of ResNet50 (ablation) | MS-Transformer · 7Scenes | position/orientation error: 0.19/8.6 -> 0.18/7.28 | Improved |
| use EfficientNetB1 instead of EfficientNetB0 backbone (ablation) | MS-Transformer · 7Scenes | position/orientation error: 0.18/7.28 -> 0.17/7.26 | No clear effect |
| sample position map at coarse 14x14x112 and orientation map at fine 28x28x40 instead of swapped (ablation) | MS-Transformer · 7Scenes | position/orientation error: 0.22/7.47 -> 0.18/7.28 | Improved |
| use different resolutions for position and orientation maps instead of the same 14x14x112 map (ablation) | MS-Transformer · 7Scenes | position/orientation error: 0.19/7.78 -> 0.18/7.28 | Improved |
| use 6 encoder/decoder layers instead of 2 (ablation) | MS-Transformer · 7Scenes | position/orientation error: 0.19/7.48 -> 0.18/7.28 | Improved |
| use 4 encoder/decoder layers instead of 6 (ablation) | MS-Transformer · 7Scenes | orientation error: 7.28 -> 6.94 | 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 2 of 2 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
- Map-Relative Pose Regression for Visual Re-Localization (CVPR 2024 · similarity 0.89)
- ConDo: Continual Domain Expansion for Absolute Pose Regression (AAAI 2025 · similarity 0.88)
- Neural Refinement for Absolute Pose Regression with Feature Synthesis (CVPR 2024 · similarity 0.88)
- Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization (CVPR 2025 · similarity 0.86)
- Scene-agnostic Pose Regression for Visual Localization (CVPR 2025 · similarity 0.86)
- HyperPose: Hypernetwork-Infused Camera Pose Localization and an Extended Cambridge Landmarks Dataset (CVPR 2025 · similarity 0.85)
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.
