Research map / 3D vision / Depth and stereo
ShapeFit and ShapeKick for Robust, Scalable Structure from Motion
Thomas Goldstein, Paul Hand, Choongbum Lee, Vladislav Voroninski, Stefano Soatto
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 1dSfM outlier filtering before location recovery (ablation) | ShapeKick · Roman Forum (monopartite k=6) | median error (m): 39 -> 4.3 | Improved |
| add 1dSfM outlier filtering before location recovery (ablation) | LUD · Roman Forum (monopartite k=6) | median error (m): 12 -> 6.6 | Improved |
| use kicking (increase tau by 10x on stagnation) instead of plain ADMM ShapeFit | ShapeFit · Piccadilly (monopartite k=6) | translation solve time (s): 240 -> 26 | Improved |
| use kicking (increase tau by 10x on stagnation) instead of plain ADMM ShapeFit | ShapeFit · Roman Forum (monopartite k=6) | translation solve time (s): 42 -> 6.8 | Improved |
| use kicking instead of plain ADMM ShapeFit | ShapeFit · Roman Forum (monopartite k=6) | median error (m): 12 -> 39 | Hurt |
| use kicking instead of plain ADMM ShapeFit | ShapeFit · Ellis Island (monopartite k=6) | median error (m): 5.7 -> 2.7 | Improved |
| replace least-squares (NLS) solver with Huber loss after 1dSfM | 1dSfM · Ellis Island (monopartite k=6) | median error (m): 3.4 -> 1.7 | Improved |
| replace monopartite k=6 with Robust PCA direction estimates | ShapeKick · Ellis Island | median error (m): 2.7 -> 30 | Hurt |
| use ShapeFit instead of LUD on noiseless synthetic corrupted directions | LUD · synthetic noiseless (RFE) | exact-recovery corruption level: LUD around 20% vs ShapeFit 10% to 50% | Improved |
Closest papers by meaning
- Estimation of Camera Locations in Highly Corrupted Scenarios: All About That Base, No Shape Trouble (CVPR 2018 · similarity 0.89)
- Certifiably Optimal Outlier-Robust Geometric Perception: Semidefinite Relaxations and Scalable Global Optimization (TPAMI 2022 · similarity 0.85)
- Convex Relaxation for Robust Vanishing Point Estimation in Manhattan World (CVPR 2025 · similarity 0.84)
- Depth2Pose: A Pose-Based Benchmark for Monocular Depth Estimation without Ground-Truth Depth (NeurIPS 2026 · similarity 0.84)
- TokenHMR: Advancing Human Mesh Recovery with a Tokenized Pose Representation (CVPR 2024 · similarity 0.84)
- Rectified Point Flow: Generic Point Cloud Pose Estimation (NeurIPS 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.
