Research map / 3D vision / Depth and stereo
Learning Optical Flow From Still Images
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 |
|---|---|---|---|
| Train RAFT on depthstilled COCO images instead of Chairs→Things | RAFT · KITTI 2012 | EPE: 2.40 -> 1.82 | Helped |
| Train RAFT on depthstilled COCO images instead of Chairs→Things | RAFT · KITTI 2015 | EPE: 5.62 -> 3.81 | Helped |
| Train RAFT on depthstilled COCO images instead of Chairs→Things | RAFT · Sintel Clean | EPE: 1.64 -> 2.63 | Hurt |
| Combine synthetic Chairs→Things training with dCOCO training | RAFT · KITTI 2012 | EPE: 1.82 -> 1.78 | Helped |
| Combine synthetic Chairs→Things training with dCOCO training | RAFT · Sintel Clean | EPE: 2.63 -> 1.88 | Helped |
| Add monocular depth estimation rather than use constant depth (ablation) | RAFT trained on dCOCO · Sintel Clean | EPE: 5.50 -> 2.52 | Helped |
| Add hole filling to depth-based virtual views (ablation) | RAFT trained on dCOCO · KITTI 2012 | EPE: 2.02 -> 1.82 | Helped |
| Add simulated independent object motions (ablation) | RAFT trained on dCOCO · Sintel Clean | EPE: 2.63 -> 2.35 | Helped |
| Add simulated independent object motions (ablation) | RAFT trained on dCOCO · KITTI 2012 | EPE: 1.82 -> 1.83 | Hurt |
| Use MiDaS rather than MegaDepth for depth estimation (ablation) | RAFT trained on dCOCO · Sintel Clean | EPE: 2.91 -> 2.63 | Helped |
| Generate five virtual motions per image rather than one (ablation) | RAFT trained on 4K COCO images · Sintel Clean | EPE: 2.73 -> 2.56 | Helped |
| Generate five virtual motions per image rather than one (ablation) | RAFT trained on 4K COCO images · KITTI 2015 | EPE: 3.93 -> 3.93 | No clear effect |
Closest papers by meaning
- A Large Dataset to Train Convolutional Networks for Disparity, Optical Flow, and Scene Flow Estimation (CVPR 2016 · similarity 0.91)
- Taming generative video models for zero-shot optical flow extraction (NeurIPS 2025 · similarity 0.89)
- DepthFM: Fast Generative Monocular Depth Estimation with Flow Matching (AAAI 2025 · similarity 0.89)
- Flow-NeRF: Joint Learning of Geometry, Poses, and Dense Flow within Unified Neural Representations (CVPR 2025 · similarity 0.89)
- Recurrent Partial Kernel Network for Efficient Optical Flow Estimation (AAAI 2024 · similarity 0.89)
- ADFactory: An Effective Framework for Generalizing Optical Flow with NeRF (CVPR 2024 · similarity 0.88)
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.