Research map / 3D vision / Depth and stereo
Upgrading Optical Flow to 3D Scene Flow Through Optical Expansion
Gengshan Yang, Deva Ramanan
Check 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 |
|---|---|---|---|
| use optical expansion with optical flow and camera intrinsics to estimate normalized 3D motion | monocular scene flow · KITTI-15 validation | MiD: 623 -> 75 | Improved |
| use optical expansion with optical flow and camera intrinsics to estimate normalized 3D motion | monocular scene flow · KITTI-15 benchmark foreground | SF: 39.57 -> 36.67 | Improved |
| use optical expansion to estimate depth change from monocular images | stereo scene flow · KITTI-15 benchmark foreground | SF: 15.94 -> 13.44 | Improved |
| use monocular optical flow and expansion with LiDAR depth | LiDAR scene flow · KITTI-15 LiDAR validation | EPE (m): 0.116 -> 0.119 | Mixed |
| use optical expansion-derived motion-in-depth for collision prediction | time-to-collision estimation · KITTI | Err-1s: 5.91 -> 4.21 | Improved |
| add scale-aware expansion supervision to census loss (ablation) | self-supervised optical flow · KITTI-15 validation | Fl EPE: 7.000 -> 6.564 | Improved |
| add residual prediction to the optical expansion network (ablation) | optical expansion estimation · KITTI | log-L1: 255 -> 245 | Improved |
| add residual prediction to the optical expansion network (ablation) | optical expansion estimation · Sintel | log-L1: 83 -> 78 | Improved |
| use initial expansion from a local affine transform instead of optical flow as the network input (ablation) | optical expansion estimation · KITTI | log-L1: 383 -> 245 | Improved |
| use initial expansion from a local affine transform instead of optical flow as the network input (ablation) | optical expansion estimation · Sintel | log-L1: 116 -> 78 | Improved |
| use initial local affine expansion rather than warped image features as the network input (ablation) | optical expansion estimation · KITTI | log-L1: 450 -> 245 | Improved |
| use initial local affine expansion rather than warped image features as the network input (ablation) | optical expansion estimation · Sintel | log-L1: 174 -> 78 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Mono-SF — Mono-SF: Multi-View Geometry Meets Single-View Depth for Monocular Scene Flow Estimation of Dynamic Traffic Scenes (ICCV 2019) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Digging Into Monocular Texture-Scale Feature Fusion for Optical Flow and Optical Expansion Estimation (TPAMI 2026 · similarity 0.90)
- Learning Optical Expansion From Scale Matching (CVPR 2023 · similarity 0.88)
- SynFlow: Scaling Up LiDAR Scene Flow Estimation with Synthetic Data (ECCV 2026 · similarity 0.88)
- DifFlow3D: Hierarchical Diffusion Models for Uncertainty-Aware 3D Scene Flow Estimation (TPAMI 2025 · similarity 0.87)
- Dense Flow from Adaptive Correspondence (NeurIPS 2026 · similarity 0.87)
- 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.
