Research map / Vision-language and multimodal / Cross-modal learning
Rethinking Unsupervised Cross-modal Flow Estimation: Learning from Decoupled Optimization and Consistency Constraint
Runmin Zhang, Jialiang Wang, Si-Yuan Cao, Zhu Yu, Junchen Yu, Guangyi Zhang, Hui-liang Shen
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 |
|---|---|---|---|
| decoupled optimization with synthetic flow supervision instead of appearance-based (ablation) | DCFlow · MS2 (RGB-T) | EPE: 21.23 -> 5.80 | Improved |
| add outlier-robust loss (ablation) | DCFlow · MS2 (RGB-T) | EPE: 5.80 -> 4.81 | Improved |
| add cross-modal consistency constraint (ablation) | DCFlow · MS2 (RGB-T) | EPE: 4.81 -> 3.46 | Improved |
| geometry-aware synthesis instead of 2D transformation (ablation) | DCFlow · MS2 (RGB-T) | EPE: 13.12 -> 3.46 | Improved |
| geometry-aware synthesis instead of 3D Gaussian Splatting (ablation) | DCFlow · MS2 (RGB-T) | EPE: 5.11 -> 3.46 | Improved |
| use RAFT instead of GMA/FlowFormer/SEA-RAFT as flow network (ablation) | DCFlow · MS2 (RGB-T) | EPE: GMA 4.13, FlowFormer 3.66, SEA-RAFT 3.57 vs RAFT 3.46 | Mixed |
| DCFlow vs large-scale pretrained MINIMA | DCFlow · five cross-modal datasets | EPE: 5.97 -> 3.46 (MS2 RGB-T) | Improved |
| DCFlow vs supervised RAFT with sparse GT | DCFlow · MS2 (RGB-T) | EPE: 1.70 (RAFT supervised) vs 3.46 | Hurt |
| DCFlow vs NeMAR on RNS | DCFlow · RNS (RGB-NIR) | EPE: 25.11 -> 1.90 | Improved |
Closest papers by meaning
- Flowing from Words to Pixels: A Noise-Free Framework for Cross-Modality Evolution (CVPR 2025 · similarity 0.87)
- Direct Product Flow Matching: Decoupling Radial and Angular Dynamics for Few-Shot Adaptation (NeurIPS 2026 · similarity 0.86)
- Cross Flow: One-Step Generation Across Latent and Pixel Spaces (NeurIPS 2026 · similarity 0.86)
- Variational Rectified Flow Matching (ICML 2025 · similarity 0.85)
- DeltaFlow: An Efficient Multi-frame Scene Flow Estimation Method (NeurIPS 2025 · similarity 0.85)
- Structured Coupling for Flow Matching (NeurIPS 2026 · 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.
