Research map / Architectures and efficiency / CNN design
Learnable Cost Volume Using the Cayley Representation
Taihong Xiao, Jinwei Yuan, Deqing Sun, Qifei Wang Xin-Yu Zhang, Kehan Xu, Ming-Hsuan Yang
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 LCV to VCN | VCN · Sintel Final test | AEPE: 4.40 -> 4.20 | Improved |
| add LCV to VCN | VCN · KITTI 2015 train | Fl-all (%): 4.10 -> 3.80 | Improved |
| add LCV to RAFT | RAFT · Sintel Final test | AEPE: 3.61 -> 3.55 | Improved |
| add LCV to RAFT | RAFT · Sintel Clean test | AEPE: 2.77 -> 2.75 | No clear effect |
| add LCV to RAFT (warm start) | RAFT · Sintel Clean test | AEPE: 2.42 -> 2.49 | Hurt |
| add LCV to DDFlow | DDFlow · KITTI 2015 unsupervised train | AEPE: 5.72 -> 5.15 | Improved |
| add LCV to DDFlow | DDFlow · KITTI 2015 unsupervised test | Fl-all (%): 14.29 -> 14.12 | Improved |
| use Cayley representation (positive definite W) instead of free W (ablation) | VCN+LCV · KITTI 2015 val | Fl-all: 1.193 -> 1.132 | Improved |
| use Cayley representation instead of 1x1 conv R^T R (positive semi-definite) (ablation) | VCN+LCV · KITTI 2015 val | Fl-all: 1.163 -> 1.132 | Improved |
| continue training VCN with small learning rate (ablation) | VCN · KITTI 2015 val | Fl-all: 1.144 -> 1.204 | Hurt |
| make only P learnable (Lambda fixed) (ablation) | VCN · KITTI 2015 val | Fl-all: 1.144 -> 1.129 | Improved |
| make only Lambda learnable (P fixed) (ablation) | VCN · KITTI 2015 val | AEPE: 3.9 -> 3.8 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- RAFT — RAFT: Recurrent All-Pairs Field Transforms for Optical Flow (ECCV 2020) · better on 4 of 4 shared results
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
- Separable Flow: Learning Motion Cost Volumes for Optical Flow Estimation (ICCV 2021 · similarity 0.87)
- Removing Cost Volumes from Optical Flow Estimators (ICCV 2025 · similarity 0.86)
- Learning Optical Flow From a Few Matches (CVPR 2021 · similarity 0.86)
- Displacement-Invariant Matching Cost Learning for Accurate Optical Flow Estimation (NeurIPS 2020 · similarity 0.85)
- FlowFormer++: Masked Cost Volume Autoencoding for Pretraining Optical Flow Estimation (CVPR 2023 · similarity 0.84)
- Recurrent Partial Kernel Network for Efficient Optical Flow Estimation (AAAI 2024 · similarity 0.84)
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.
