Research map / 3D vision / Depth and stereo
U^2Flow: Uncertainty-Aware Unsupervised Optical Flow Estimation
Xunpei Sun, Wenwei Lin, Yi Chang, Gang Chen
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 uncertainty head without decoupling flow and uncertainty losses (ablation) | U2Flow · KITTI 2015 train | Fl-all (%): 6.82 -> 7.72 | Hurt |
| decouple uncertainty loss from flow loss (ablation) | U2Flow · KITTI 2015 train | Fl-all (%): 7.72 -> 6.87 | Improved |
| add uncertainty-aware flow refinement (ablation) | U2Flow · Sintel Final train | EPE: 2.41 -> 2.32 | Improved |
| add uncertainty-based homography smoothness on KITTI (ablation) | U2Flow · KITTI 2015 train | Fl-all (%): 6.85 -> 6.59 | Improved |
| add uncertainty-enhanced homography loss on Sintel (ablation) | U2Flow · Sintel Final train | EPE: 2.32 -> 2.52 | Hurt |
| add uncertainty-guided bidirectional flow fusion (ablation) | U2Flow · KITTI 2015 train | Fl-all (%): 6.59 -> 6.30 | Improved |
| use occlusion mask instead of uncertainty mask for fusion (ablation) | U2Flow · KITTI 2015 train | Fl-all (%): 6.59 -> 8.17 | Hurt |
| refinement without uncertainty versus with uncertainty (ablation) | U2Flow · KITTI 2015 train | Fl-all (%): 6.81 -> 6.59 | Improved |
| scale flow feature by uncertainty weight instead of plain concatenation (ablation) | U2Flow · KITTI 2015 train | Fl-all (%): 7.24 -> 6.59 | Improved |
| use spatial plus appearance augmentation rather than appearance only for uncertainty (ablation) | U2Flow · Sintel uncertainty | AUSE: 0.46 -> 0.11 | Improved |
| add appearance augmentation on top of spatial only (ablation) | U2Flow · Sintel uncertainty | AUSE: 0.12 -> 0.11 | No clear effect |
| add flow fusion for cross-dataset testing (ablation) | U2Flow · KITTI->Sintel Final | EPE: 5.80 -> 5.49 | Improved |
Closest papers by meaning
- ProbFlow: Joint Optical Flow and Uncertainty Estimation (ICCV 2017 · similarity 0.91)
- UnFlow: Unsupervised Learning of Optical Flow With a Bidirectional Census Loss (AAAI 2018 · similarity 0.91)
- Uncertainty Estimates and Multi-Hypotheses Networks for Optical Flow (ECCV 2018 · similarity 0.90)
- UnSAMFlow: Unsupervised Optical Flow Guided by Segment Anything Model (CVPR 2024 · similarity 0.88)
- FlowDiffuser: Advancing Optical Flow Estimation with Diffusion Models (CVPR 2024 · similarity 0.86)
- DPFlow: Adaptive Optical Flow Estimation with a Dual-Pyramid Framework (CVPR 2025 · 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.
