Research map / Architectures and efficiency / RNNs and normalization
ScopeFlow: Dynamic Scene Scoping for Optical Flow
Aviram Bar-Haim, Lior Wolf
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 |
|---|---|---|---|
| Replace fixed partial crops with randomized crop sizes in [0.95,1] (ablation) | IRR-PWC · KITTI validation | MEPE: 1.651 -> 1.402 | Improved |
| Use random-range crops [0.95,1] instead of fixed full crops (ablation) | IRR-PWC · KITTI validation | MEPE: 1.594 -> 1.402 | Improved |
| Re-finetune from the [0.9,1] crop-range checkpoint rather than the [0.95,1] checkpoint (ablation) | IRR-PWC on KITTI · KITTI validation | MEPE: 1.393 -> 1.377 | Improved |
| Remove weight decay and random noise during early FChairsOcc pretraining (ablation) | IRR-PWC · Sintel test | MEPE: 3.138 -> 3.264 | Hurt |
| Reduce zoom-in during early FChairsOcc pretraining (ablation) | IRR-PWC · Sintel test | MEPE: 3.138 -> 3.321 | Hurt |
| Increase scope with maximal crops during FThings training, starting from C1 rather than C* (ablation) | IRR-PWC · Sintel test | MEPE: 2.613 -> 2.544 | Improved |
| Remove weight decay and random noise during additional FThings training (ablation) | IRR-PWC · Sintel test | MEPE: 2.544 -> 2.545 | No clear effect |
| Remove weight decay and random noise during Sintel finetuning (ablation) | IRR-PWC · Sintel validation | MEPE: 2.119 -> 2.108 | Improved |
| Remove weight decay and random noise during Sintel finetuning (ablation) | IRR-PWC · Sintel validation | occlusion F1: 0.700 -> 0.703 | Improved |
| Train IRR-PWC with the ScopeFlow protocol rather than the baseline protocol | IRR-PWC · MPI Sintel final pass | all EPE: 4.579 -> 4.098 | Improved |
| Train IRR-PWC with the ScopeFlow protocol rather than the baseline protocol | IRR-PWC · KITTI 2012 | Avg-All: 1.6 -> 1.3 | Improved |
| Train IRR-PWC with the ScopeFlow protocol rather than the baseline protocol | IRR-PWC · KITTI 2015 | Fl-all: 7.65 -> 6.82 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PWC-Net — PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volume (CVPR 2018) · better on 1 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
- Improving Optical Flow on a Pyramid Level (ECCV 2020 · similarity 0.87)
- MEMFOF: High-Resolution Training for Memory-Efficient Multi-Frame Optical Flow Estimation (ICCV 2025 · similarity 0.87)
- StreamFlow: Streamlined Multi-Frame Optical Flow Estimation for Video Sequences (NeurIPS 2024 · similarity 0.87)
- DPFlow: Adaptive Optical Flow Estimation with a Dual-Pyramid Framework (CVPR 2025 · similarity 0.87)
- SynFlow: Scaling Up LiDAR Scene Flow Estimation with Synthetic Data (ECCV 2026 · similarity 0.87)
- MemFlow: Optical Flow Estimation and Prediction with Memory (CVPR 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.
