Research map / 3D vision / Depth and stereo
Boosting 3D Foundation Models with Featureless Pose Optimization
Mattia D'Urso, Christian Sormann, Mattia Rossi, Friedrich Fraundorfer
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 EPO refinement on VGGT instead of best BA method | VGGT · ScanNet++ | AUC@5deg: 70.9 -> 77.1 | Improved |
| add EPO refinement instead of best BA method | VGGT · TerraSky3D | AUC@5deg: 75.5 -> 79.2 | Improved |
| add EPO refinement instead of best BA method | VGGT · Mip-NeRF 360 | AUC@5deg: 87.8 -> 90.5 | Improved |
| add EPO instead of Ref+BA for runtime | VGGT · ScanNet++ | total inference time (s): 303.5 -> 52.1 | Improved |
| add track refinement before BA | VGGT+BA · ScanNet++ | AUC@5deg: 70.0 -> 70.9 | No clear effect |
| add track refinement before BA | VGGT+BA · TerraSky3D | AUC@5deg: 71.1 -> 75.5 | Improved |
| add EPO for novel view synthesis instead of Ref+BA | 3DGS on VGGT poses · Mip-NeRF 360 | LPIPS: 0.320 -> 0.284 | Improved |
| add pose MLP refiner (ablation) | EPO free-variable baseline · Mip-NeRF 360 | AUC@5deg: 77.0 -> 79.0 | Improved |
| add parametric K and Z (ablation) | EPO · Mip-NeRF 360 | AUC@5deg: 79.0 -> 80.2 | Improved |
| add translation offset delta (ablation) | EPO · Mip-NeRF 360 | AUC@5deg: 80.2 -> 82.0 | Improved |
| increase iterations from 100 to 1000 (ablation) | EPO · Mip-NeRF 360 | LPIPS: 0.308 -> 0.269 | Improved |
| increase iterations from 1000 to 2000 (ablation) | EPO · Mip-NeRF 360 | AUC@5deg: 91.0 -> 92.1 | 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.
- VGGT — VGGT: Visual Geometry Grounded Transformer (CVPR 2025) · better on 4 of 7 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
- Learning Scene Coordinate Reconstruction from Unposed Images via Pose Graph Optimization (CVPR 2026 · similarity 0.86)
- SDFit: 3D Object Pose and Shape by Fitting a Morphable SDF to a Single Image (ICCV 2025 · similarity 0.85)
- VGGSfM: Visual Geometry Grounded Deep Structure From Motion (CVPR 2024 · similarity 0.85)
- Leveraging Camera Triplets for Efficient and Accurate Structure-from-Motion (CVPR 2024 · similarity 0.84)
- Light3R-SfM: Towards Feed-forward Structure-from-Motion (CVPR 2025 · similarity 0.84)
- LoRA3D: Low-Rank Self-Calibration of 3D Geometric Foundation models (ICLR 2025 · similarity 0.84)
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.
