Research map / Video / Video segmentation and tracking
RoMo: Robust Motion Segmentation Improves Structure from Motion
Lily Goli, Sara Sabour, Mark Matthews, Marcus A. Brubaker, Dmitry Lagun, Alec Jacobson, David J. Fleet, Saurabh Saxena, Andrea Tagliasacchi
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 |
|---|---|---|---|
| use RoMo zero-shot masks instead of synthetic-supervised OCLR-adap | motion segmentation · FBMS59 | IoU: 69.8 -> 75.5 | Improved |
| RoMo versus OCLR-adap on single-primary-object VOS benchmarks | RoMo · DAVIS16 | IoU: 80.9 -> 77.3 | Hurt |
| add RoMo masks to bundle adjustment on dense tracks | TAPIR+TheiaSFM · MPI Sintel | RPE-R: 1.175 -> 0.217 | Improved |
| RoMo masks vs LEAP-VO on Sintel ATE | RoMo+TheiaSFM · MPI Sintel | ATE: 0.093 vs 0.089 | No clear effect |
| add RoMo masks to COLMAP | COLMAP · Casual Motion (8 scenes) | ATE: 8.00 -> 2.79 | Improved |
| replace SAMv2 features with Stable Diffusion features (ablation) | RoMo classifier · SegTrackV2 (13 scenes) | IoU: 73.08 -> 71.53 | No clear effect |
| add SAMv2 post-processing refinement (ablation) | RoMo · SegTrackV2 (13 scenes) | IoU: 68.68 -> 73.08 | Improved |
| add learned MLP classifier on epipolar labels before SAMv2 (ablation) | RoMo · DAVIS2016 | IoU: 70.9 -> 79.0 | Improved |
| increase refinement iterations from 1 to 2 (ablation) | RoMo · DAVIS2016 | IoU: 78.1 -> 79.0 | Improved |
| increase refinement iterations from 2 to 3 (ablation) | RoMo · DAVIS2016 | IoU: 79.0 -> 79.0 | No clear effect |
| increase MLP hidden size from 8 to 32 (ablation) | RoMo classifier · DAVIS2016 | IoU: 79.0 -> 75.8 | Hurt |
| use looser thresholds (theta_u 1.0, theta_l 0.10) (ablation) | RoMo · DAVIS2016 | IoU: 79.0 -> 77.8 | Hurt |
Closest papers by meaning
- Segmenting the Motion Components of a Video: A Long-Term Unsupervised Model (TPAMI 2025 · similarity 0.87)
- The Dynamic Prior: Understanding 3D Structures for Casual Dynamic Videos (ECCV 2026 · similarity 0.87)
- MOVE: Motion-Guided Few-Shot Video Object Segmentation (ICCV 2025 · similarity 0.87)
- GMOS: Grounding Moving Object Segmentation in 3D Space and Time (NeurIPS 2026 · similarity 0.86)
- VGGSfM: Visual Geometry Grounded Deep Structure From Motion (CVPR 2024 · similarity 0.86)
- Segment Any Motion in Videos (CVPR 2025 · similarity 0.85)
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.
