Research map / Architectures and efficiency / RNNs and normalization
Fast Optical Flow Using Dense Inverse Search
Till Kroeger, Radu Timofte, Dengxin Dai, Luc Van Gool
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 densification (patch averaging weighted by photometric error) between scales (ablation) | DIS inverse search · Sintel training sparse correspondences | EPE all (px): 7.76 -> 4.16 | Improved |
| add densification (ablation) | DIS inverse search · Sintel training, displacement 40+ px | EPE (px): 37.94 -> 23.09 | Improved |
| add densification (ablation) | DIS inverse search · Sintel training, displacement 0-10 px | EPE (px): 2.16 -> 0.84 | Improved |
| use inverse search with densification instead of exhaustive NCC nearest-neighbor | DIS · Sintel training sparse correspondences | EPE all (px): 32.06 -> 4.16 | Improved |
| use DeepMatching instead of DIS | sparse matching · Sintel training sparse correspondences | EPE all (px): 4.16 -> 3.60 | Improved |
| add variational refinement (ablation) | DIS · Sintel training | average EPE | Improved |
| increase finest-scale resolution, iterations or patch overlap (ablation) | DIS · Sintel training | average EPE | Improved |
| vary patch size theta_ps (ablation) | DIS · Sintel training | average EPE | Mixed |
| add OpenMP parallelization on 4 cores (ablation) | DIS · Sintel run-time | speed-up (x): 1.29 -> 3.70 | Improved |
| run DIS at higher frame rate through all intermediate frames instead of DeepFlow on large displacements | DIS operating point 3 · Sintel with synthesized low frame-rate | EPE | Improved |
| remove variational refinement (operating point 1 vs 2) (ablation) | DIS · Sintel | run-time: 301 Hz -> 606 Hz | Improved |
Closest papers by meaning
- Flow Fields: Dense Correspondence Fields for Highly Accurate Large Displacement Optical Flow Estimation (TPAMI 2019 · similarity 0.90)
- Dense Flow from Adaptive Correspondence (NeurIPS 2026 · similarity 0.89)
- DIP: Deep Inverse Patchmatch for High-Resolution Optical Flow (CVPR 2022 · similarity 0.88)
- Explicit Motion Disentangling for Efficient Optical Flow Estimation (ICCV 2023 · similarity 0.87)
- UFM: A Simple Path towards Unified Dense Correspondence with Flow (NeurIPS 2025 · similarity 0.86)
- Balancing Frequencies and Pixels in Flow Matching (NeurIPS 2026 · 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.
