Research map / Video / Video segmentation and tracking
RealFlow: EM-Based Realistic Optical Flow Dataset Generation from Videos
Yunhui Han, Kunming Luo, Ao Luo, Jiangyu Liu, Haoqiang Fan, Guiming Luo, Shuaicheng Liu
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 realistic image pair rendering (RIPR) (ablation) | RealFlow with RAFT · KITTI15 train (RF-Ktrain, 1 EM iteration) | EPE: 3.93 -> 2.79 | Improved |
| render with RIPR instead of QVI interpolation (ablation) | RealFlow with RAFT · KITTI15 train (RF-Ktrain, 1 EM iteration) | EPE: 7.27 -> 2.79 | Improved |
| use softmax splatting instead of max splatting (ablation) | RealFlow with RAFT · KITTI15 train (RF-Ktrain, 1 EM iteration) | EPE: 3.03 -> 2.79 | Improved |
| use DPT depth instead of MiDaS (ablation) | RealFlow with RAFT · KITTI15 train (RF-Ktrain, 1 EM iteration) | EPE: 2.90 -> 2.79 | Improved |
| use DPT depth instead of bi-directional occlusion map (ablation) | RealFlow with RAFT · KITTI15 train (RF-Ktrain, 1 EM iteration) | EPE: 3.01 -> 2.79 | Improved |
| add bi-directional hole filling (BHF) vs no filling (ablation) | RealFlow with RAFT · KITTI15 train (RF-Ktrain, 1 EM iteration) | EPE: 2.95 -> 2.79 | Improved |
| use BHF instead of RFR inpainting (ablation) | RealFlow with RAFT · KITTI15 train (RF-Ktrain, 1 EM iteration) | EPE: 2.95 -> 2.79 | Improved |
| sample alpha from [0,2] instead of fixed 1 (ablation) | RealFlow with RAFT · KITTI15 train (RF-Ktrain, 1 EM iteration) | EPE: 3.38 -> 2.79 | Improved |
| run 4 EM iterations instead of 1 (ablation) | RealFlow with RAFT · KITTI15 train | EPE: 2.79 -> 2.16 | Improved |
| train 4x longer on Iter.1 data instead of EM iterations (ablation) | RealFlow with RAFT · KITTI15 train | EPE: 2.79 -> 2.86 | No clear effect |
| train RAFT on RF-AB instead of C+T | RAFT · KITTI15 train | EPE: 5.04 -> 3.48 | Improved |
| train RAFT on RF-AB instead of C+T | RAFT · Sintel Final train | EPE: 2.71 -> 3.28 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- SMURF — SMURF: Self-Teaching Multi-Frame Unsupervised RAFT With Full-Image Warping (CVPR 2021) · better on 0 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
- Learning Optical Flow From Still Images (CVPR 2021 · similarity 0.90)
- Taming generative video models for zero-shot optical flow extraction (NeurIPS 2025 · similarity 0.89)
- StreamFlow: Streamlined Multi-Frame Optical Flow Estimation for Video Sequences (NeurIPS 2024 · similarity 0.88)
- MemFlow: Optical Flow Estimation and Prediction with Memory (CVPR 2024 · similarity 0.87)
- ADFactory: An Effective Framework for Generalizing Optical Flow with NeRF (CVPR 2024 · similarity 0.87)
- FlowVid: Taming Imperfect Optical Flows for Consistent Video-to-Video Synthesis (CVPR 2024 · 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.
