Research map / Detection and segmentation / Open-set and few-shot detection
The Right Spin: Learning Object Motion from Rotation-Compensated Flow Fields
Pia Bideau, Erik G. Miller, Cordelia Schmid, Karteek Alahari
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 rotation compensation to ground-truth flow input (ablation) | motion segmentation network · FT3D test | IoU: 90.68 -> 93.23 | Improved |
| add rotation compensation to estimated PWC flow input (ablation) | motion segmentation network · FT3D test | IoU: 77.18 -> 78.69 | Improved |
| add flow magnitude to angle input (GT t-flow) (ablation) | motion segmentation network · FT3D test | IoU: 77.47 -> 93.23 | Improved |
| add flow magnitude to angle input (PWC t-flow) (ablation) | motion segmentation network · FT3D test | IoU: 77.79 -> 78.69 | Improved |
| test GT-trained network on estimated PWC t-flow instead of GT t-flow (ablation) | motion segmentation network · FT3D test | IoU: 93.23 -> 24.44 | Hurt |
| use likelihood maximization instead of TMM for rotation estimation (GT flow) (ablation) | camera rotation estimator · Sintel | yaw/pitch/roll error (deg): 0.08 / 0.22 / 0.02 -> 0.03 / 0.06 / 0.01 | Improved |
| use likelihood maximization instead of TMM for rotation estimation (PWC flow) (ablation) | camera rotation estimator · Sintel | yaw/pitch/roll error (deg): 0.13 / 0.34 / 0.04 -> 0.05 / 0.11 / 0.03 | Improved |
| add 3D Gaussian filtering and dense CRF post-processing (ablation) | Ours · MoCA locomotion | J mean: 58.3 -> 65.8 | Improved |
| add appearance cues to flow-only model | Ours · DAVIS2016 | J mean: 59.7 -> 73.5 | Improved |
| mask moving objects with GT segments for camera motion estimation (ablation) | Ours · DAVIS2016 (flow only) | J mean: 59.7 -> 62.5 | Improved |
| add rotation compensation and geometry vs flow-only LMP network | Ours vs LMP · DAVIS2016 (flow only) | J mean: 58.4 -> 59.7 | Improved |
Closest papers by meaning
- Learning Motion Patterns in Videos (CVPR 2017 · similarity 0.88)
- It’s Moving! A Probabilistic Model for Causal Motion Segmentation in Moving Camera Videos (ECCV 2016 · similarity 0.88)
- Flow Snapshot Neurons in Action: Deep Neural Networks Generalize to Biological Motion Perception (NeurIPS 2024 · similarity 0.87)
- Motion4D: Learning 3D-Consistent Motion and Semantics for 4D Scene Understanding (NeurIPS 2025 · similarity 0.87)
- Layered Motion Fusion: Lifting Motion Segmentation to 3D in Egocentric Videos (CVPR 2025 · similarity 0.87)
- Object segmentation from common fate: Motion energy processing enables human-like zero-shot generalization to random dot stimuli (NeurIPS 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.
