Research map / Label-efficient and robust learning / Self-supervised and contrastive
Flow2Stereo: Effective Self-Supervised Learning of Optical Flow and Stereo Matching
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Jointly train flow and stereo with geometric constraints and challenging self-supervision | PWC-Net-based Flow2Stereo · KITTI 2012 and KITTI 2015 optical-flow test benchmarks | Fl-all: KITTI 2012: 7.68% (SelFlow) -> 7.63%; KITTI 2015: 14.19% (SelFlow) -> 11.10% | Helped |
| Use the optical-flow model to estimate stereo disparity | Flow2Stereo · KITTI stereo test benchmarks | D1-all: KITTI 2012: 5.93% -> 5.11%; KITTI 2015: 6.67% -> 6.61% | Helped |
| Add both quadrilateral and triangle geometric constraints (ablation) | Photometric-only teacher model · KITTI 2012 and KITTI 2015 train non-occluded pixels | EPE-noc: KITTI 2012: 1.06 -> 0.84; KITTI 2015: 2.85 -> 2.24 | Helped |
| Add quadrilateral constraint (ablation) | Photometric-only teacher model · KITTI 2012 train | EPE-noc: 1.06 -> 0.84 | Helped |
| Add triangle constraint (ablation) | Photometric-only teacher model · KITTI 2012 train | EPE-noc: 1.06 -> 0.86 | Helped |
| Add self-supervision (ablation) | Flow2Stereo with photometric and geometric constraints · KITTI 2012 train | EPE-all: 4.91 -> 1.45 | Helped |
| Use v2 rather than v1 self-supervision, without distinguishing occluded and non-occluded pixels (ablation) | Flow2Stereo · KITTI 2012 train | EPE-all: 1.92 -> 1.89 | No clear effect |
| Add additional challenging conditions in self-supervision (ablation) | Flow2Stereo self-supervision v2 · KITTI 2012 and KITTI 2015 train | EPE-all: KITTI 2012: 1.89 -> 1.62; KITTI 2015: 5.48 -> 4.12 | Helped |
| Add geometric constraints during self-supervision (ablation) | Flow2Stereo self-supervision v3 · KITTI 2012 and KITTI 2015 train | EPE-all: KITTI 2012: 1.62 -> 1.56; KITTI 2015: 4.12 -> 3.66 | Hurt |
Closest papers by meaning
- Bridging Stereo Matching and Optical Flow via Spatiotemporal Correspondence (CVPR 2019 · similarity 0.92)
- SelFlow: Self-Supervised Learning of Optical Flow (CVPR 2019 · similarity 0.90)
- Feature-Level Collaboration: Joint Unsupervised Learning of Optical Flow, Stereo Depth and Camera Motion (CVPR 2021 · similarity 0.90)
- DualNet: Robust Self-Supervised Stereo Matching with Pseudo-Label Supervision (AAAI 2025 · similarity 0.89)
- Self-Point-Flow: Self-Supervised Scene Flow Estimation From Point Clouds With Optimal Transport and Random Walk (CVPR 2021 · similarity 0.89)
- Self-Supervised Monocular Scene Flow Estimation (CVPR 2020 · 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.