Research map / 3D vision / Depth and stereo
Structure and Motion from Casual Videos
Zhoutong Zhang, Forrester Cole, Zhengqi Li, Noah Snavely, Michael Rubinstein, William T. Freeman
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 |
|---|---|---|---|
| remove uncertainty map (movement map), use plain L1 loss; i.e. add uncertainty map (ablation) | CasualSAM · Sintel camera pose | ATE: 0.134 -> 0.089 | Improved |
| add movement/uncertainty map (ablation) | CasualSAM · Sintel dynamic regions depth | Rel. L1 error: 2.295 -> 1.267 | Improved |
| add movement/uncertainty map (ablation) | CasualSAM · Sintel depth | avg Rel. L1 error: 0.779 -> 0.484 | Improved |
| add movement/uncertainty map (ablation) | CasualSAM · Sintel static regions depth | Rel. L1 error: 0.284 -> 0.227 | Improved |
| add full joint optimization after initialization stage (ablation) | CasualSAM · Sintel camera pose | ATE: 0.122 -> 0.089 | Improved |
| add full joint optimization after initialization stage (ablation) | CasualSAM · Sintel depth | avg Rel. L1 error: 0.697 -> 0.484 | Improved |
| provide ground truth focal length instead of optimizing focal (ablation) | CasualSAM · Sintel camera pose | ATE: 0.089 -> 0.036 | Improved |
| CasualSAM vs DROID-SLAM with GT intrinsics | DROID-SLAM · Sintel camera pose | ATE: 0.077 -> 0.036 | Improved |
| CasualSAM vs RCVD with optimized focal length | RCVD · Sintel camera pose | ATE: 0.164 -> 0.089 | Improved |
| CasualSAM vs RCVD on DAVIS | RCVD · DAVIS | Residual Flow Error: 53% | Improved |
| CasualSAM with GT focal vs DROID-SLAM on TUM dynamic | DROID-SLAM · TUM RGB-D dynamic | ATE: w_halfsphere 0.023 -> 0.080 | Mixed |
| CasualSAM vs RCVD on TUM dynamic with unknown intrinsics | RCVD · TUM RGB-D dynamic | ATE: s_halfsphere 0.254 -> 0.096 | Improved |
Closest papers by meaning
- MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos (CVPR 2025 · similarity 0.90)
- VGGSfM: Visual Geometry Grounded Deep Structure From Motion (CVPR 2024 · similarity 0.88)
- AnyMap: Learning a General Camera Model for Structure-from-Motion with Unknown Distortion in Dynamic Scenes (CVPR 2025 · similarity 0.87)
- MP-SfM: Monocular Surface Priors for Robust Structure-from-Motion (CVPR 2025 · similarity 0.86)
- AnyCam: Learning to Recover Camera Poses and Intrinsics from Casual Videos (CVPR 2025 · similarity 0.86)
- Fast Encoder-Based 3D from Casual Videos via Point Track Processing (NeurIPS 2024 · 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.
