Research map / 3D vision / Depth and stereo
Learning To Segment Rigid Motions From Two Frames
Gengshan Yang, Deva Ramanan
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 |
|---|---|---|---|
| Add rigidity cost-map inputs (ablation) | Rigid motion segmentation network · Sintel | background IoU (%): 76.81 to 84.63 | Improved |
| Add uncertainty inputs (ablation) | Rigid motion segmentation network · KITTI-SF | object IoU (%): 85.09 to 89.53 | Improved |
| Add monocular depth inputs (ablation) | Rigid motion segmentation network · KITTI-SF | object IoU (%): 84.46 to 89.53 | Improved |
| Add optical expansion to upgrade 2D flow to 3D (ablation) | Rigid motion segmentation network · KITTI-SF | object IoU (%): 81.28 to 89.53 | Improved |
| Use a learned segmentation network instead of manually thresholded rigidity cost maps (ablation) | Rigid motion segmentation · KITTI-SF | object IoU (%): 25.83 to 89.53 | Improved |
| Use rigid motion masks instead of Mask R-CNN masks for rigid-body scene-flow fitting | Scene-flow estimation · KITTI benchmark | scene flow error (%): 5.61 to 4.89 | Improved |
| Apply rigid-body refinement to optical-expansion scene flow (ablation) | Scene-flow estimation · KITTI-SF | scene flow error reduction (%): 24.6% | Improved |
| Use rigid motion masks rather than appearance-based instance masks | Monocular scene-flow estimation · KITTI | scene flow error (%): 22.47 to 22.19 | Improved |
| Use rigid motion masks rather than appearance-based instance masks | Monocular scene-flow estimation · Sintel | scene flow error (%): 47.40 to 46.05 | Improved |
| Use rigid motion masks rather than appearance-based instance masks (ablation) | Stereo scene-flow estimation · KITTI benchmark | scene flow error (%): 5.61 to 4.89 | Improved |
| Use rigid motion segmentation instead of optical expansion without rigid-body refinement | Monocular scene-flow estimation · KITTI | scene flow error (%): 44.87 to 22.19 | Improved |
| Use rigid motion segmentation instead of optical expansion without rigid-body refinement | Monocular scene-flow estimation · Sintel | scene flow error (%): 55.43 to 46.05 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Mono-SF — Mono-SF: Multi-View Geometry Meets Single-View Depth for Monocular Scene Flow Estimation of Dynamic Traffic Scenes (ICCV 2019) · better on 0 of 1 shared result
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
- Learning Rigidity in Dynamic Scenes with a Moving Camera for 3D Motion Field Estimation (ECCV 2018 · similarity 0.89)
- The Best of Both Worlds: Combining CNNs and Geometric Constraints for Hierarchical Motion Segmentation (CVPR 2018 · similarity 0.89)
- Multi-frame Motion Segmentation by Combining Two-Frame Results (IJCV 2022 · similarity 0.88)
- Appearance-based Refinement for Object-Centric Motion Segmentation (ECCV 2024 · similarity 0.88)
- Layered Motion Fusion: Lifting Motion Segmentation to 3D in Egocentric Videos (CVPR 2025 · similarity 0.88)
- Learning Segmentation from Point Trajectories (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.
