Research map / Architectures and efficiency / CNN design
SOMA: Feature Gradient Enhanced Affine-Flow Matching for SAR-Optical Registration
Haodong Wang, Tao Zhuo, Xiuwei Zhang, Hanlin Yin, Wencong Wu, Yanning Zhang
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 frozen DINOv2 to coarsest level (ablation) | baseline · SEN1-2 | CMR@1px / CMR@2px: 27.75 vs 28.85 at 1px; 81.05 vs 76.28 at 2px | Mixed |
| add Feature Gradient Enhancer (FGE) (ablation) | baseline · SEN1-2 | CMR@1px: 28.85 -> 49.39 | Improved |
| add GLAM affine-flow matcher alone (ablation) | baseline · SEN1-2 | CMR@1px / CMR@2px: 28.85 -> 42.79 at 1px; 76.28 -> 71.15 at 2px | Mixed |
| add GLAM to DINOv2+FGE (ablation) | DINO+FGE · SEN1-2 | Ravg: 1.77 -> 0.94 | Improved |
| add frozen DINOv2 to FGE+GLAM (ablation) | FGE+GLAM · SEN1-2 | CMR@1px: 75.43 -> 87.53 | Improved |
| add FGE to DINOv2+GLAM (ablation) | DINO+GLAM · SEN1-2 | CMR@1px: 60.88 -> 87.53 | Improved |
| SOMA versus strongest baseline at 1px | SOMA · SEN1-2 | CMR@1px: 74.38 -> 86.67 | Improved |
| SOMA versus strongest baseline at 1px | SOMA · GFGE SO | CMR@1px: 60.88 -> 79.38 | Improved |
| evaluate SOMA zero-shot on external datasets (no baseline comparison) | SOMA · WHU-SEN-City / OSdataset | CMR@5px: CMR@5px 96.75 (WHU-SEN-City), 92.96 (OSdataset) | No clear effect |
Closest papers by meaning
- Cross-Rejective Open-Set SAR Image Registration (CVPR 2025 · similarity 0.84)
- SARMAE: Masked Autoencoder for SAR Representation Learning (CVPR 2026 · similarity 0.83)
- SAR-DisentDM: A Semantic-Disentangled Diffusion Model for Limited-Data SAR Image Synthesis (AAAI 2026 · similarity 0.83)
- RSAR: Restricted State Angle Resolver and Rotated SAR Benchmark (CVPR 2025 · similarity 0.83)
- SARDet-100K: Towards Open-Source Benchmark and ToolKit for Large-Scale SAR Object Detection (NeurIPS 2024 · similarity 0.82)
- FUSAR-GPT: A Spatiotemporal Feature-Embedded and Two-Stage Decoupled Visual Language Model for SAR Imagery (CVPR 2026 · similarity 0.81)
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.
