Research map / Video / Video segmentation and tracking
SplatFlow: Learning Multi-frame Optical Flow via Splatting
Bo Wang, Yifan Zhang, Jian Li, Yang Yu, Zhenping Sun, Li Liu, Dewen Hu
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 forward-warp splatting alignment instead of backward warp (ablation) | SplatFlow-GMA · Things val Clean | EPE: 3.23 -> 3.02 | Improved |
| add forward-warp splatting alignment instead of backward flow (ablation) | SplatFlow-GMA · KITTI train | EPE: 4.30 -> 3.70 | Improved |
| make splatting differentiable instead of gradient-disabled (ablation) | SplatFlow-GMA · Things val Clean | EPE: 3.25 -> 3.02 | Improved |
| use sub-pixel splatting instead of nearest filling (ablation) | SplatFlow-GMA · Things val Clean | EPE: 4.11 -> 3.25 | Improved |
| average splatting instead of softmax splatting (ablation) | SplatFlow-GMA · KITTI train | EPE: 3.92 -> 3.70 | Improved |
| average splatting instead of softmax splatting (ablation) | SplatFlow-GMA · Things val Clean | EPE: 2.97 -> 3.02 | Hurt |
| Final-to-All embedding instead of One-to-One (ablation) | SplatFlow-GMA · Things val Final | EPE: 3.43 -> 3.23 | Improved |
| Final-to-All embedding instead of Final-to-Final (ablation) | SplatFlow-GMA · Things val Final | EPE: 4.09 -> 3.23 | Improved |
| add multi-frame SplatFlow to GMA | GMA · Sintel test Final | EPE: 2.47 -> 2.07 | Improved |
| add multi-frame SplatFlow to RAFT | RAFT · Things val Clean | EPE: 5.24 -> 3.43 | Improved |
| add multi-frame SplatFlow to RAFT | RAFT · Sintel train Final | EPE: 2.70 -> 2.84 | Hurt |
| add WarmStart to GMA | GMA · Things val Clean | EPE: 4.98 -> 5.16 | Hurt |
Later papers that compared with this method
1 later accepted paper reports SplatFlow as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- GeoVideo: Introducing Geometric Regularization into Video Generation Model (NeurIPS 2025 · 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
- SplatFlow: Multi-View Rectified Flow Model for 3D Gaussian Splatting Synthesis (CVPR 2025 · similarity 0.89)
- StreamFlow: Streamlined Multi-Frame Optical Flow Estimation for Video Sequences (NeurIPS 2024 · similarity 0.88)
- DPFlow: Adaptive Optical Flow Estimation with a Dual-Pyramid Framework (CVPR 2025 · similarity 0.87)
- OCAI: Improving Optical Flow Estimation by Occlusion and Consistency Aware Interpolation (CVPR 2024 · similarity 0.87)
- VoxelSplat: Dynamic Gaussian Splatting as an Effective Loss for Occupancy and Flow Prediction (CVPR 2025 · similarity 0.87)
- SplatFlow: Self-Supervised Dynamic Gaussian Splatting in Neural Motion Flow Field for Autonomous Driving (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.
