Research map / 3D vision / Depth and stereo
Unifying Flow, Stereo and Depth Estimation
Haofei Xu, Jing Zhang, Jianfei Cai, Hamid Rezatofighi, Fisher Yu, Dacheng Tao, Andreas Geiger
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 cross-attention (ablation) | GMFlow · Things val (clean) | EPE: 10.84 -> 6.67 | Improved |
| add position encoding (ablation) | GMFlow · Things val (clean) | EPE: 8.38 -> 6.67 | Improved |
| add FFN (ablation) | GMFlow · Things val (clean) | EPE: 8.71 -> 6.67 | Improved |
| add self-attention (ablation) | GMFlow · Things val (clean) | EPE: 7.04 -> 6.67 | Improved |
| use global matching instead of local 9x9 matching (ablation) | GMFlow · Things val (clean) | EPE: 19.88 -> 6.67 | Improved |
| use 2x2 window splits instead of 1x1 (ablation) | GMFlow · Things val (clean) | EPE and time: 6.34 -> 6.67 EPE, 105 -> 53 ms | Mixed |
| add flow propagation (ablation) | GMFlow · Sintel (clean) | EPE: 2.28 -> 1.89 | Improved |
| add flow propagation (unmatched pixels) (ablation) | GMFlow · Sintel (clean) | EPE unmatched: 15.54 -> 10.39 | Improved |
| add one hierarchical refinement (ablation) | GMFlow · Things val (clean) | EPE: 3.48 -> 2.80 | Improved |
| use 1D cross-attention instead of 2D (ablation) | GMStereo · Things (stereo) | EPE: 1.25 -> 1.22 | Improved |
| add cross-attention (ablation) | GMStereo · Things (stereo) | EPE: 1.96 -> 1.22 | Improved |
| initialize stereo model from pretrained flow model | GMStereo · Scene Flow (Things) | EPE: 1.22 -> 1.10 | Improved |
Closest papers by meaning
- Feature-Level Collaboration: Joint Unsupervised Learning of Optical Flow, Stereo Depth and Camera Motion (CVPR 2021 · similarity 0.92)
- Bridging Stereo Matching and Optical Flow via Spatiotemporal Correspondence (CVPR 2019 · similarity 0.91)
- CroCo v2: Improved Cross-view Completion Pre-training for Stereo Matching and Optical Flow (ICCV 2023 · similarity 0.90)
- UFM: A Simple Path towards Unified Dense Correspondence with Flow (NeurIPS 2025 · similarity 0.89)
- DepthFM: Fast Generative Monocular Depth Estimation with Flow Matching (AAAI 2025 · similarity 0.88)
- Flow-NeRF: Joint Learning of Geometry, Poses, and Dense Flow within Unified Neural Representations (CVPR 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.
