Research map / 3D vision / Depth and stereo
Deep Two-View Structure-From-Motion Revisited
Jianyuan Wang, Yiran Zhong, Yuchao Dai, Stan Birchfield, Kaihao Zhang, Nikolai Smolyanskiy, Hongdong Li
Check 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 the classic correspondence-to-pose pipeline with scale-invariant depth matching | Deep two-view SfM · KITTI Depth Eigen | RMSE: 2.946 -> 2.273 | Improved |
| use the proposed two-view SfM pipeline rather than DeepV2D | KITTI Depth Eigen SfM · KITTI Depth Eigen SfM | D1-all: 9.1% vs 3.1% | Improved |
| use the proposed deep correspondence and geometric pose pipeline | Two-view SfM · KITTI VO Seq. 09 | t_err (%): 3.49 -> 1.70 | Improved |
| use the proposed deep correspondence and geometric pose pipeline | Two-view SfM · KITTI VO Seq. 10 | t_err (%): 5.81 -> 1.49 | Improved |
| add rigid-flow supervision to fine-tune optical flow (ablation) | Our optical flow estimator · MVS | Tran: 10.984 -> 3.878 | Improved |
| add rigid-flow supervision to fine-tune optical flow (ablation) | Our optical flow estimator · Scenes11 | Tran: 6.617 -> 2.041 | Improved |
| add rigid-flow supervision to fine-tune optical flow (ablation) | Our optical flow estimator · SUN3D | Tran: 12.905 -> 10.757 | Improved |
| mask optical-flow matches using SIFT feature locations before five-point RANSAC (ablation) | Five-point pose estimation · MVS | Tran: 37.292 -> 3.878 | Improved |
| use scale-invariant plane-sweep matching (ablation) | Depth estimation pipeline · KITTI Depth Eigen | RMSE: 3.120 -> 2.273 | Improved |
| replace scale-invariant matching with a scale-invariant loss (ablation) | Depth estimation pipeline · KITTI Depth Eigen | RMSE: 2.273 -> 2.981 | Hurt |
| train with ground-truth pose despite the scale mismatch (ablation) | Depth estimation pipeline · KITTI Depth Eigen | RMSE: 3.120 -> 3.421 | Hurt |
| apply scale-invariant matching to DeepV2D (ablation) | DeepV2D · KITTI Depth Eigen | Abs Rel: 0.064 -> 0.063 | 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.
- DeepSFM — DeepSFM: Structure From Motion Via Deep Bundle Adjustment (ECCV 2020) · better on 6 of 6 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
- Multi-body Depth and Camera Pose Estimation from Multiple Views (ICCV 2023 · similarity 0.91)
- VGGSfM: Visual Geometry Grounded Deep Structure From Motion (CVPR 2024 · similarity 0.90)
- Depth2Pose: A Pose-Based Benchmark for Monocular Depth Estimation without Ground-Truth Depth (NeurIPS 2026 · similarity 0.90)
- MP-SfM: Monocular Surface Priors for Robust Structure-from-Motion (CVPR 2025 · similarity 0.89)
- Multi-view Reconstruction via SfM-guided Monocular Depth Estimation (CVPR 2025 · similarity 0.88)
- Light3R-SfM: Towards Feed-forward Structure-from-Motion (CVPR 2025 · 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.
