Research map / Vision-language and multimodal / Cross-modal learning
RAMEN: Resolution-Adjustable Multimodal Encoder for Earth Observation
Nicolas Houdré, Diego Marcos, Hugo Riffaud de Turckheim, Dino Ienco, Laurent Wendling, Camille Kurtz, Sylvain Lobry
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 |
|---|---|---|---|
| refine encoded central wavelengths from RGB-only to true Sentinel-2 values (ablation) | RAMEN · Sen1Floods11 | mIoU: 79.14 -> 83.74 | Improved |
| use adjustable resampling with true interpolation ratio instead of naive bilinear (ablation) | RAMEN · Sen1Floods11 (GSD_target 160 m) | mIoU: 80.84 -> 83.74 | Improved |
| use adjustable resampling instead of naive at fine target GSD (ablation) | RAMEN · Sen1Floods11 (GSD_target 40 m) | mIoU: 88.30 -> 88.99 | No clear effect |
| add Sentinel-1 SAR to Sentinel-2 (ablation) | RAMEN · Sen1Floods11 | mIoU: 89.96 -> 91.20 | Improved |
| add Sentinel-1 SAR to Sentinel-2 (ablation) | RAMEN · PASTIS | mIoU: 40.99 -> 44.25 | Improved |
| add Sentinel-1 SAR to Sentinel-2 (ablation) | RAMEN · CTM-SS | mIoU: 57.20 -> 57.35 | No clear effect |
| use coarser target GSD (ablation) | RAMEN · HLS BurnScars | mIoU: 87.07 -> 88.44 | Improved |
| use finer target GSD (ablation) | RAMEN · MADOS | mIoU: 57.09 -> 78.07 | Improved |
| add late fusion with LTAE instead of native temporal attention (ablation) | RAMEN · PASTIS | mIoU: 28.07 -> 42.29 | Improved |
| add late fusion with LTAE instead of native temporal attention (ablation) | RAMEN · CTM-SS | mIoU: 53.01 -> 53.27 | No clear effect |
| use RAMEN ViT-Base instead of TerraMindv1-L | PANGAEA · 8-task average | mIoU: 59.10 -> 60.03 | Improved |
| use RAMEN instead of TerraMindv1-L | PANGAEA MADOS · MADOS | mIoU: 75.57 -> 69.72 | Hurt |
Closest papers by meaning
- GeoLLaVA-8K: Scaling Remote-Sensing Multimodal Large Language Models to 8K Resolution (NeurIPS 2025 · similarity 0.88)
- EarthDial: Turning Multi-sensory Earth Observations to Interactive Dialogues (CVPR 2025 · similarity 0.88)
- AnySat: One Earth Observation Model for Many Resolutions, Scales, and Modalities (CVPR 2025 · similarity 0.88)
- GeoCore-9B: Towards Geo-Aware Generative Foundation Models in Earth Observation (NeurIPS 2026 · similarity 0.88)
- No Location Left Behind: Measuring and Improving the Fairness of Implicit Representations for Earth Data (ICLR 2025 · similarity 0.87)
- Earth-Agent: Unlocking the Full Landscape of Earth Observation with Agents (ICLR 2026 · similarity 0.86)
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.
