Research map / 3D vision / Depth and stereo
Pixel-Perfect Structure-From-Motion With Featuremetric Refinement
Philipp Lindenberger, Paul-Edouard Sarlin, Viktor Larsson, Marc Pollefeys
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 featuremetric refinement to SIFT | SIFT · ETH3D triangulation (indoor) | accuracy @1cm (%): 75.62 -> 82.82 | Improved |
| add featuremetric refinement to SuperPoint | SuperPoint · ETH3D triangulation (indoor) | accuracy @1cm (%): 75.76 -> 89.33 | Improved |
| add featuremetric refinement to D2-Net | D2-Net · ETH3D triangulation (indoor) | accuracy @1cm (%): 47.18 -> 82.49 | Improved |
| replace Patch Flow with featuremetric keypoint refinement | SIFT · ETH3D camera pose estimation | AUC @1mm (%): 14.62 -> 25.38 | Improved |
| add featuremetric refinement to SuperPoint localization | SuperPoint · ETH3D camera pose estimation | AUC @1mm (%): 15.38 -> 40.00 | Improved |
| add featuremetric refinement to SuperPoint+SuperGlue | SuperPoint+SuperGlue · IMC 2020 multiview | AUC @5 deg@N=5 (%): 63.02 -> 68.87 | Improved |
| add featuremetric refinement to D2-Net | D2-Net · IMC 2020 stereo | AUC @5 deg (%): 16.83 -> 25.89 | Improved |
| add featuremetric bundle adjustment after keypoint adjustment (ablation) | SuperPoint F-KA · ETH3D Facade | triangulation accuracy @1cm (%): 36.85 -> 46.46 | Improved |
| add featuremetric keypoint adjustment before bundle adjustment (ablation) | SuperPoint F-BA · ETH3D Facade | triangulation accuracy @1cm (%): 43.65 -> 46.46 | Improved |
| use S2DNet features instead of photometric NCC (ablation) | SuperPoint refinement · ETH3D Facade | triangulation accuracy @1cm (%): 28.43 -> 46.46 | Improved |
| let BA reference features drift (ablation) | SuperPoint full · ETH3D Facade | localization AUC @1cm (%): 71.97 -> 64.51 | Mixed |
| add query keypoint adjustment (qKA) before pose estimation (ablation) | SuperPoint KA+BA · ETH3D camera pose | AUC @1mm (%): 16.92 -> 38.46 | Improved |
Closest papers by meaning
- Detector-Free Structure from Motion (CVPR 2024 · similarity 0.89)
- VGGSfM: Visual Geometry Grounded Deep Structure From Motion (CVPR 2024 · similarity 0.88)
- Reconstructing People, Places, and Cameras (CVPR 2025 · similarity 0.87)
- World Tracing: Pixel-Aligned Geometry Beyond the Visible (NeurIPS 2026 · similarity 0.87)
- Light3R-SfM: Towards Feed-forward Structure-from-Motion (CVPR 2025 · similarity 0.87)
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 2024 · 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.
