Research map / 3D vision / Depth and stereo
Displacement-Invariant Matching Cost Learning for Accurate Optical Flow Estimation
Read the paperCheck 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 |
|---|---|---|---|
| learn matching cost with DICL module instead of dot product (ablation) | flow network · FlyingChair val | EPE: 1.86 -> 1.33 | Helped |
| learn matching cost with DICL module instead of cosine similarity (ablation) | flow network · FlyingChair val | EPE: 1.84 -> 1.33 | Helped |
| learn matching cost with DICL module instead of three-layer MLP (ablation) | flow network · FlyingChair val | EPE: 1.76 -> 1.33 | Helped |
| add motion-aware projection layer (ablation) | DICL flow network · KITTI-15 train (fine-tuned) | EPE: (1.06) -> (0.95) | Helped |
| add motion-aware projection layer (ablation) | DICL flow network · KITTI-15 train (C+T) | EPE: 8.78 -> 8.70 | Helped |
| add motion-aware projection layer (ablation) | DICL flow network · Sintel train clean (C+T) | EPE: 2.11 -> 1.99 | Helped |
| add motion-aware projection layer (ablation) | DICL flow network · Sintel train clean (+S) | EPE: (1.32) -> (1.19) | Helped |
| add motion-aware projection layer (ablation) | DICL flow network · Sintel | median d_peak: 0.57 -> 0.63 | Helped |
| use DICL network instead of VCN, fine-tuned | optical flow · Sintel test final | EPE: 4.40 -> 3.60 | Helped |
| use DICL network instead of VCN, C+T only | optical flow · KITTI-15 train | EPE: 8.36 -> 8.70 | Hurt |
Closest papers by meaning
- Accurate Optical Flow via Direct Cost Volume Processing (CVPR 2017 · similarity 0.90)
- Rethinking Optical Flow From Geometric Matching Consistent Perspective (CVPR 2023 · similarity 0.89)
- Adaptive Unimodal Cost Volume Filtering for Deep Stereo Matching (AAAI 2020 · similarity 0.89)
- CFNet: Cascade and Fused Cost Volume for Robust Stereo Matching (CVPR 2021 · similarity 0.89)
- Deep Stereo Matching With Explicit Cost Aggregation Sub-Architecture (AAAI 2018 · similarity 0.88)
- Learning Optical Flow From a Few Matches (CVPR 2021 · 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.