Research map / 3D vision / Depth and stereo
Occlusions, Motion and Depth Boundaries with a Generic Network for Disparity, Optical Flow or Scene Flow Estimation
Eddy Ilg, Tonmoy Saikia, Margret Keuper, Thomas Brox
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 |
|---|---|---|---|
| provide ground-truth forward flow as input to occlusion network (ablation) | FlowNetS occlusion estimation · FlyingThings3D | F-measure: 0.790 -> 0.932 | Improved |
| provide ground-truth forward flow as input to occlusion network (ablation) | FlowNetS occlusion estimation · Sintel clean | F-measure: 0.545 -> 0.653 | Improved |
| add warped and flipped backward flow input (ablation) | FlowNetS occlusion estimation · FlyingThings3D | F-measure: 0.932 -> 0.943 | Improved |
| add raw backward flow input (ablation) | FlowNetS occlusion estimation · FlyingThings3D | F-measure: 0.932 -> 0.930 | No clear effect |
| estimate occlusions jointly with flow (ablation) | FlowNetC · Sintel train clean | EPE: 3.21 -> 3.20 | No clear effect |
| add forward/backward flow estimation (FlowNetC-Bi) (ablation) | FlowNetC joint flow and occlusion · Sintel train clean | EPE: 3.20 -> 3.26 | Hurt |
| add occlusion estimation to refinement stack (ablation) | FlowNet2-CS · Sintel train clean | EPE: 2.28 -> 2.25 | No clear effect |
| add backward direction estimation to stack (ablation) | FlowNet2-CS with occlusions · Sintel train clean | EPE: 2.25 -> 2.77 | Hurt |
| add mutual warping of fwd/bwd flows (ablation) | FlowNet2-CS with occlusions · Sintel train clean | EPE: 2.25 -> 2.25 | No clear effect |
| use own occlusions in motion segmentation with FlowNet2 flow | Keuper motion segmentation · FBMS-59 test (8px) | F-measure: 76.71% -> 77.14% | Improved |
| replace FlowNet2 flow and occlusions with ours | Keuper motion segmentation · FBMS-59 test (8px) | F-measure: 76.71% -> 80.45% | Improved |
| move correlation layer from level 2 to level 3 (ablation) | DispNetC · Sintel train clean disparity | EPE: 5.62 -> 3.19 | Improved |
Closest papers by meaning
- MirrorFlow: Exploiting Symmetries in Joint Optical Flow and Occlusion Estimation (ICCV 2017 · similarity 0.91)
- Learning To Estimate Hidden Motions With Global Motion Aggregation (ICCV 2021 · similarity 0.90)
- Feature-Level Collaboration: Joint Unsupervised Learning of Optical Flow, Stereo Depth and Camera Motion (CVPR 2021 · similarity 0.89)
- Occlusion Aware Unsupervised Learning of Optical Flow (CVPR 2018 · similarity 0.89)
- OCAI: Improving Optical Flow Estimation by Occlusion and Consistency Aware Interpolation (CVPR 2024 · similarity 0.88)
- OAMaskFlow: Occlusion-Aware Motion Mask for Scene Flow (AAAI 2025 · 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.
