Research map / Vision-language and multimodal / Cross-modal learning
MaRS: A Multi-modality Very-high-resolution Remote Sensing Foundation Model with Cross-Granularity Meta-Modality Learning
Ruoyu Yang, Yinhe Liu, Heng Yan, Yiheng Zhou, Yihan Fu, Han Luo, Yanfei Zhong
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 CGCL to the MIM + contrastive baseline (ablation) | MaRS baseline · EarthMiss modality-missing mapping | mIoU: 42.0 -> 49.1 | Improved |
| add CGCL to the MIM + contrastive baseline (ablation) | MaRS baseline · WHU-CD change detection | IoU: 67.1 -> 76.0 | Improved |
| add MMA to the MIM + contrastive baseline (ablation) | MaRS baseline · WHU-CD change detection | IoU: 67.1 -> 86.2 | Improved |
| add MMA to the MIM + contrastive baseline (ablation) | MaRS baseline · EarthMiss modality-missing mapping | mIoU: 42.0 -> 48.4 | Improved |
| add MMA to the baseline (registration) (ablation) | MaRS baseline · GUSO registration | RMSE: 2.835 -> 2.854 | Hurt |
| add CGCL to the baseline (registration) (ablation) | MaRS baseline · GUSO registration | RMSE: 2.835 -> 2.795 | Improved |
| combine CGCL and MMA (full MaRS) vs either alone (ablation) | MaRS · SARDet-100K | mAP: 53.5 -> 55.4 | Improved |
| combine CGCL and MMA vs CGCL alone (ablation) | MaRS · WHU-CD change detection | IoU: 76.0 -> 87.0 | Improved |
| use hierarchical SwinV2 backbone instead of ViT-L | VFM baselines · Multi-modal VHR tasks | various: Building height Delta 23.26 -> 23.38 | Improved |
| pretrain on medium-resolution data (RSFMs) rather than VHR data | RSFM baselines · Multi-modal VHR tasks | various: e.g. EarthMiss mIoU SatMAE 35.0 vs MaRS 49.9 | Hurt |
| pretrain MaRS on MaRS-16M vs generic SwinV2 | SwinV2-B · GUSO translation | PSNR: 16.71 -> 20.69 | 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.
- CrossMAE — CrossMAE: Cross-Modality Masked Autoencoders for Region-Aware Audio-Visual Pre-Training (CVPR 2024) · better on 1 of 1 shared result
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
- SkySense: A Multi-Modal Remote Sensing Foundation Model Towards Universal Interpretation for Earth Observation Imagery (CVPR 2024 · similarity 0.87)
- RingMoE: Mixture-of-Modality-Experts Multi-Modal Foundation Models for Universal Remote Sensing Image Interpretation (TPAMI 2025 · similarity 0.87)
- RobSense: A Robust Multi-modal Foundation Model for Remote Sensing with Static, Temporal, and Incomplete Data Adaptability (CVPR 2025 · similarity 0.85)
- Galileo: Learning Global & Local Features of Many Remote Sensing Modalities (ICML 2025 · similarity 0.84)
- LDS2AE: Local Diffusion Shared-Specific Autoencoder for Multimodal Remote Sensing Image Classification with Arbitrary Missing Modalities (AAAI 2024 · similarity 0.83)
- FUSAR-GPT: A Spatiotemporal Feature-Embedded and Two-Stage Decoupled Visual Language Model for SAR Imagery (CVPR 2026 · similarity 0.83)
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.
