Research map / Probabilistic and scientific modeling / Molecules and proteins
Accurate Optical Flow via Direct Cost Volume Processing
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 DC Flow instead of SPM-BPv2 | Optical flow estimation · Sintel final test set | AEPE all: 5.812 -> 5.119 | Helped |
| Use DC Flow instead of PatchBatch | Optical flow estimation · KITTI 2015 test set, all pixels | Fl-all (%): 21.07 -> 14.86 | Helped |
| Replace classical NCC matching cost with the learned feature embedding (ablation) | Full Flow · Sintel validation set | AEPE: 6.37 -> 6.01 | Helped |
| Replace classical NCC matching cost with the learned feature embedding (ablation) | Full Flow · KITTI 2015 validation set | non-occluded pixels (%): 14.33 -> 11.10 | Helped |
| Add Flow-SGM cost-volume processing instead of using winner-takes-all (ablation) | DC Flow · Sintel validation set | AEPE: 7.22 -> 5.51 | Helped |
| Add Flow-SGM cost-volume processing instead of using winner-takes-all (ablation) | DC Flow · KITTI 2015 validation set | occluded pixels (%): 27.37 -> 20.47 | Helped |
| Add homography-based inpainting after Flow-SGM (ablation) | DC Flow · KITTI 2015 validation set | occluded pixels (%): 20.47 -> 15.09 | Helped |
| Add homography-based inpainting after Flow-SGM (ablation) | DC Flow · Sintel validation set | AEPE: 5.51 -> 5.44 | Helped |
| Reduce feature dimensionality from 64 to 10 (ablation) | DC Flow · Sintel validation set | AEPE: 5.51 -> 5.71 | Hurt |
| Reduce feature dimensionality from 64 to 10 (ablation) | DC Flow · KITTI 2015 validation set | occluded pixels (%): 20.47 -> 21.42 | Hurt |
Closest papers by meaning
- Displacement-Invariant Matching Cost Learning for Accurate Optical Flow Estimation (NeurIPS 2020 · similarity 0.90)
- Full Flow: Optical Flow Estimation By Global Optimization Over Regular Grids (CVPR 2016 · similarity 0.88)
- Efficient All-Pairs Correlation Volume Sampling for Optical Flow Estimation (CVPR 2026 · similarity 0.88)
- DPFlow: Adaptive Optical Flow Estimation with a Dual-Pyramid Framework (CVPR 2025 · similarity 0.86)
- UFM: A Simple Path towards Unified Dense Correspondence with Flow (NeurIPS 2025 · similarity 0.85)
- Stereo Anywhere: Robust Zero-Shot Deep Stereo Matching Even Where Either Stereo or Mono Fail (CVPR 2025 · similarity 0.85)
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.