Research map / 3D vision / Depth and stereo
Visual Localization Using Imperfect 3D Models From the Internet
Vojtech Panek, Zuzana Kukelova, 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 |
|---|---|---|---|
| Use a realistic, detailed Internet 3D model as the scene representation | MeshLoc localization · Notre Dame and Pantheon | mean DCRE (% of image diagonal): 10% or less | Improved |
| Use higher appearance fidelity to compensate for lower geometric fidelity | MeshLoc localization · Pantheon and Reichstag | localization performance | Improved |
| Use substantially distorted or low-detail models | MeshLoc localization · St. Vitus Cathedral and St. Peter's Square | localization performance | Hurt |
| Use locally refined rather than globally aligned poses as DCRE ground truth | MeshLoc evaluation · Internet-model scenes | mean DCRE performance | Hurt |
| Increase non-uniform scale inconsistency in model width or height (ablation) | Notre Dame A with LoFTR · Notre Dame | localization success | Hurt |
| Increase non-uniform scale inconsistency in model depth (ablation) | Notre Dame A with LoFTR · Notre Dame | localization success | No clear effect |
| Replace rendered model appearance with real images while retaining Internet-model geometry for depth maps (ablation) | MeshLoc localization · IMC 2021 scenes | pose accuracy | Improved |
| Simplify geometry and/or texture of Notre Dame A, with both simplified at ratio 1/16 (ablation) | MeshLoc with LoFTR · Notre Dame | localization accuracy | Hurt |
| Reduce model size through simplification to ratio 1/16 (ablation) | Notre Dame A · Notre Dame | model size: 18.4 MB original vs. 0.7 MB at 1/16 ratio | Improved |
Closest papers by meaning
- A Guide to Structureless Visual Localization (IJCV 2026 · similarity 0.89)
- LaMAR: Benchmarking Localization and Mapping for Augmented Reality (ECCV 2022 · similarity 0.89)
- Are Large-Scale 3D Models Really Necessary for Accurate Visual Localization? (CVPR 2017 · similarity 0.89)
- Long-Term Visual Localization Revisited (TPAMI 2020 · similarity 0.88)
- 360Loc: A Dataset and Benchmark for Omnidirectional Visual Localization with Cross-device Queries (CVPR 2024 · similarity 0.87)
- Rooms from Motion: Un-posed Indoor 3D Object Detection as Localization and Mapping (NeurIPS 2025 · 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.
