Research map / Detection and segmentation / Open-set and few-shot detection
The Best of Both Worlds: Combining CNNs and Geometric Constraints for Hierarchical Motion Segmentation
Pia Bideau, Aruni RoyChowdhury, Rakesh R. Menon, Erik Learned-Miller
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 |
|---|---|---|---|
| combine perspective-geometry rigid-motion segmentation with CNN object proposals and object-level motion grouping | motion segmentation · FBMS-motion, complex background, camouflaged animals; multi-label | F: 64.97 vs 63.59; 65.60 vs 45.83; 72.15 vs 61.52 | Improved |
| combine perspective-geometry rigid-motion segmentation with CNN object proposals and object-level motion grouping | motion segmentation · FBMS-motion; binary | F: 81.85 vs 86.96 | Hurt |
| combine perspective-geometry rigid-motion segmentation with CNN object proposals and object-level motion grouping | motion segmentation · camouflaged animals; binary | F: 75.95 vs 76.31 | Hurt |
| combine perspective-geometry rigid-motion segmentation with CNN object proposals and object-level motion grouping | motion segmentation · complex background; binary | F: 90.11 vs 86.56 | Improved |
| model flow-error variance as a function of flow magnitude instead of using constant variance (ablation) | motion segmentation · all (44 sequences), multi-label | F: 64.86 -> 66.45 | Improved |
| model flow-error variance as a function of flow magnitude instead of using constant variance (ablation) | motion segmentation · all (44 sequences), binary | F: 81.15 -> 81.27 | Improved |
| add grouping of rigid motion components into composed object-motion models (ablation) | motion segmentation · all (44 sequences), multi-label | F: 52.69 -> 66.45 | Improved |
| add composed motion models to refine object boundaries rather than use object proposals alone (ablation) | motion segmentation · all (44 sequences), binary | F: 78.78 -> 81.27 | Improved |
| add a fully-connected CRF refinement (ablation) | motion segmentation · all (44 sequences), multi-label | F: 66.45 -> 66.51 | No clear effect |
| add a fully-connected CRF refinement (ablation) | motion segmentation · all (44 sequences), binary | F: 81.27 -> 81.49 | No clear effect |
Closest papers by meaning
- Learning To Segment Rigid Motions From Two Frames (CVPR 2021 · similarity 0.89)
- Learning Segmentation from Point Trajectories (NeurIPS 2024 · similarity 0.88)
- Semi-Weakly Supervised Object Kinematic Motion Prediction (CVPR 2023 · similarity 0.88)
- Segment Any Motion in Videos (CVPR 2025 · similarity 0.87)
- Discovering the Physical Parts of an Articulated Object Class From Multiple Videos (CVPR 2016 · similarity 0.87)
- Articulation in Prime: Primitive-Based Articulated Object Understanding from a Single Casual Video (NeurIPS 2026 · 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.
