Research map / 3D vision / Depth and stereo
Learning Rigidity in Dynamic Scenes with a Moving Camera for 3D Motion Field Estimation
Zhaoyang Lv, Kihwan Kim, Alejandro Troccoli, Deqing Sun, James M. Rehg, Jan Kautz
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 RTN-learned rigidity to pose refinement (ablation) | Refine only · SINTEL test split clean, no finetuning | projected scene flow EPE: 8.47 -> 5.10 | Improved |
| use RTN rigidity instead of semantic rigidity (ablation) | Refine pipeline · SINTEL test split clean, no finetuning | projected scene flow EPE: 6.42 -> 5.10 | Improved |
| use RTN rigidity instead of semantic rigidity (ablation) | Refine pipeline · SINTEL test split final, no finetuning | projected scene flow EPE: 12.81 -> 6.29 | Improved |
| add joint pose learning to RTN (ablation) | RTN + Refine · SINTEL test split clean, no finetuning | projected scene flow EPE: 7.09 -> 5.10 | Improved |
| add joint pose learning to RTN (ablation) | RTN + Refine · SINTEL test split final, no finetuning | projected scene flow EPE: 7.14 -> 6.29 | Improved |
| train RTN on REFRESH instead of Things3D (ablation) | RTN + Refine · SINTEL test split final, no finetuning | projected scene flow EPE: 11.88 -> 6.29 | Improved |
| train RTN on REFRESH instead of Things3D (ablation) | RTN · SINTEL clean val | rigidity mean IOU: 0.283 -> 0.542 | Improved |
| use RTN rigidity instead of semantic rigidity (ablation) | RTN trained on REFRESH · SINTEL clean val | rigidity mean IOU: 0.392 -> 0.542 | Improved |
| use RTN rigidity instead of semantic rigidity (ablation) | RTN trained on REFRESH · SINTEL final val | rigidity mean IOU: 0.446 -> 0.627 | Improved |
| use RTN rigidity instead of semantic rigidity (ablation) | RTN trained on REFRESH · REFRESH val | rigidity mean IOU: 0.934 -> 0.956 | Improved |
| finetune RTN on SINTEL (ablation) | RTN + Refine · SINTEL test split clean, no finetuning | projected scene flow EPE: 5.10 -> 5.03 | No clear effect |
| finetune RTN on SINTEL (ablation) | RTN + Refine · SINTEL test split final, no finetuning | projected scene flow EPE: 6.29 -> 6.12 | No clear effect |
Closest papers by meaning
- EffiScene: Efficient Per-Pixel Rigidity Inference for Unsupervised Joint Learning of Optical Flow, Depth, Camera Pose and Motion Segmentation (CVPR 2021 · similarity 0.90)
- Self-Supervised 3D Scene Flow Estimation and Motion Prediction Using Local Rigidity Prior (TPAMI 2024 · similarity 0.89)
- Optical Flow in Mostly Rigid Scenes (CVPR 2017 · similarity 0.89)
- RigidFlow: Self-Supervised Scene Flow Learning on Point Clouds by Local Rigidity Prior (CVPR 2022 · similarity 0.89)
- VoteFlow: Enforcing Local Rigidity in Self-Supervised Scene Flow (CVPR 2025 · similarity 0.87)
- Degrees of Freedom Matter: Inferring Dynamics from Point Trajectories (CVPR 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.
