Research map / Vision-language and multimodal / Cross-modal learning
CRFT: Consistent-Recurrent Feature Flow Transformer for Cross-Modal Image Registration
Xuecong Liu, Mengzhu Ding, Zixuan Sun, Zhang Li, Xichao Teng
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 L1 dense flow supervision (ablation) | XoFTR · OSdataset ablation (30 deg rotation) | AEPE: 3.76 -> 0.51 | Improved |
| add coarse-scale flow estimation (FE) (ablation) | XoFTR+L1 · OSdataset ablation (30 deg rotation) | CMR@1px (%): 73.3 -> 93.4 | Improved |
| add coarse-scale flow estimation (FE) (ablation) | XoFTR+L1 · OSdataset ablation (30 deg rotation) | AEPE: 0.51 -> 0.63 | Hurt |
| add coarse-scale flow estimation (FE) (ablation) | XoFTR+L1 · OSdataset ablation (30 deg rotation) | CMR@0.7px (%): 66.7 -> 73.3 | Improved |
| add iterative discrepancy-guided optimization (IDGO) (ablation) | X+FE+L1 · OSdataset ablation (30 deg rotation) | CMR@0.7px (%): 73.3 -> 88.9 | Improved |
| add iterative discrepancy-guided optimization (IDGO) (ablation) | X+FE+L1 · OSdataset ablation (30 deg rotation) | AEPE: 0.63 -> 0.43 | Improved |
| replace L1 with geometrically weighted iterative loss (IL) (ablation) | X+FE+IDGO · OSdataset ablation (30 deg rotation) | CMR@0.7px (%): 88.9 -> 93.1 | Improved |
| replace L1 with geometrically weighted iterative loss (IL) (ablation) | X+FE+IDGO · OSdataset ablation (30 deg rotation) | AEPE: 0.43 -> 0.40 | Improved |
| replace L1 with geometrically weighted iterative loss (IL) (ablation) | X+FE+IDGO · OSdataset ablation (30 deg rotation) | CMR@3px (%): 99.7 -> 99.7 | No clear effect |
| increase rotation to 90 deg with scaling [0.5,1.5] | CRFT · OSdataset (optical-SAR) | AEPE: 0.65 -> 9.75 | Hurt |
Closest papers by meaning
- RBE-Flow:Recurrent Bayesian Estimation on Feature Manifolds for Cross-Modal Registration (ECCV 2026 · similarity 0.91)
- Direct Product Flow Matching: Decoupling Radial and Angular Dynamics for Few-Shot Adaptation (NeurIPS 2026 · similarity 0.85)
- UFM: A Simple Path towards Unified Dense Correspondence with Flow (NeurIPS 2025 · similarity 0.84)
- CARL: A Framework for Equivariant Image Registration (CVPR 2025 · similarity 0.84)
- Variational Rectified Flow Matching (ICML 2025 · similarity 0.84)
- Self-Supervised Spatial Correspondence Across Modalities (CVPR 2025 · 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.
