Research map / 3D vision / Depth and stereo
MirrorFlow: Exploiting Symmetries in Joint Optical Flow and Occlusion Estimation
Junhwa Hur, Stefan Roth
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 occlusion-disocclusion symmetry term and forward-backward consistency to the asymmetric model (ablation) | MirrorFlow · KITTI 2015 train | Fl-all (all pixels): 12.68 -> 9.98 | Improved |
| add occlusion-disocclusion symmetry term alone to the asymmetric model (ablation) | MirrorFlow · KITTI 2015 train | Fl-all (all pixels): 12.68 -> 10.19 | Improved |
| add forward-backward consistency term alone to the asymmetric model (ablation) | MirrorFlow · KITTI 2015 train | Fl-all (all pixels): 12.68 -> 14.74 | Hurt |
| add forward-backward consistency term on top of the symmetry term (ablation) | MirrorFlow · KITTI 2015 train | Fl-all (all pixels): 10.19 -> 9.98 | Improved |
| add forward-backward consistency term on top of the symmetry term (ablation) | MirrorFlow · KITTI 2015 train | Fl-all (non-occluded): 7.62 -> 7.41 | Improved |
| add occlusion-disocclusion symmetry term to the consistency-only model (ablation) | MirrorFlow · KITTI 2015 train | Fl-all (all pixels): 14.74 -> 9.98 | Improved |
| add occlusion-disocclusion symmetry term to the consistency-only model (ablation) | MirrorFlow · KITTI 2015 train | Fl-bg (non-occluded): 10.41 -> 6.52 | Improved |
| joint symmetric formulation versus its initialization DiscreteFlow | MirrorFlow · KITTI 2015 test | Fl-all (all pixels): 21.57 -> 10.29 | Improved |
| joint symmetric formulation versus DiscreteFlow | MirrorFlow · MPI Sintel Clean | EPE all: 3.567 -> 3.316 | Improved |
| joint symmetric formulation versus DiscreteFlow | MirrorFlow · MPI Sintel Final | EPE all: 6.077 -> 6.071 | No clear effect |
| joint symmetric formulation versus DiscreteFlow on occluded pixels | MirrorFlow · MPI Sintel Clean | EPE occluded: 23.626 -> 19.470 | Improved |
Closest papers by meaning
- Occlusions, Motion and Depth Boundaries with a Generic Network for Disparity, Optical Flow or Scene Flow Estimation (ECCV 2018 · similarity 0.91)
- Learning To Estimate Hidden Motions With Global Motion Aggregation (ICCV 2021 · similarity 0.91)
- OCAI: Improving Optical Flow Estimation by Occlusion and Consistency Aware Interpolation (CVPR 2024 · similarity 0.89)
- MaskFlownet: Asymmetric Feature Matching With Learnable Occlusion Mask (CVPR 2020 · similarity 0.89)
- Imposing Consistency for Optical Flow Estimation (CVPR 2022 · similarity 0.89)
- Iterative Residual Refinement for Joint Optical Flow and Occlusion Estimation (CVPR 2019 · similarity 0.89)
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.
