Research map / 3D vision / Depth and stereo
Dense Dynamic Scene Reconstruction and Camera Pose Estimation from Multi-View Videos
Shuo Sun, Unal Artan, Malcolm Mielle, Achim Lilienthal, Martin Magnusson
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 wide-baseline initialization with VGGT (ablation) | Ours · MultiCamRobolab RoboDog overlap | ATE: +0.284 | Improved |
| add wide-baseline initialization (ablation) | Ours · MultiCamRobolab RoboDog non-overlap | ATE: +0.496 | Improved |
| add wide-baseline initialization (ablation) | Ours · MultiCamRobolab DynamicHuman | ATE: +0.130 | Improved |
| add spatio-temporal graph (ablation) | Ours · MultiCamRobolab RoboDog overlap | ATE: +0.158 | Improved |
| add spatio-temporal graph (ablation) | Ours · MultiCamRobolab DynamicHuman | ATE: +0.141 | Improved |
| add spatio-temporal graph with little camera overlap (ablation) | Ours · MultiCamRobolab RoboDog non-overlap | ATE: +0.006 | No clear effect |
| add Phase 1 per-frame scale alignment (ablation) | Ours · MultiCamRobolab RoboArm | Md: +0.226 -> +0.087 | Improved |
| add Phase 2 pixel depth and pose refinement after Phase 1 (ablation) | Ours · MultiCamRobolab RoboDog overlap | Md: +0.029 -> 0.091 | Improved |
| add refinement Phase 1 on RoboArm delta accuracy (ablation) | Ours · MultiCamRobolab RoboArm | delta 1.25: -0.123 -> -0.083 | Improved |
| add 0.10 m translation noise to initialization (ablation) | Ours · MultiCamRobolab | ATE: 0.013 -> 0.044 | Hurt |
| add 0.02 m translation noise to initialization (ablation) | Ours · MultiCamRobolab | ATE: 0.013 -> 0.013 | No clear effect |
| mask dynamic objects in COLMAP/GLOMAP | COLMAP · MultiCamRobolab DynamicHuman | ATE | Hurt |
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.
- Fast3R — Fast3R: Towards 3D Reconstruction of 1000+ Images in One Forward Pass (CVPR 2025) · better on 2 of 2 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
- MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos (CVPR 2025 · similarity 0.91)
- LivePose: Online 3D Reconstruction from Monocular Video with Dynamic Camera Poses (ICCV 2023 · similarity 0.91)
- R3D3: Dense 3D Reconstruction of Dynamic Scenes from Multiple Cameras (ICCV 2023 · similarity 0.91)
- Back on Track: Bundle Adjustment for Dynamic Scene Reconstruction (ICCV 2025 · similarity 0.90)
- Dynamic Visual SLAM using a General 3D Prior (CVPR 2026 · similarity 0.89)
- Depth2Pose: A Pose-Based Benchmark for Monocular Depth Estimation without Ground-Truth Depth (NeurIPS 2026 · similarity 0.88)
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.
