Research map / Vision-language and multimodal / Cross-modal learning
SkySense V2: A Unified Foundation Model for Multi-modal Remote Sensing
Yingying Zhang, Lixiang Ru, Kang Wu, Lei Yu, Lei Liang, Yansheng Li, Jingdong Chen
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 |
|---|---|---|---|
| replace SkySense separate per-modality backbones with a unified backbone (SkySense V2) | SkySense V2 · Model size | parameters: 1.26 billion -> 665 million | Improved |
| replace SkySense with SkySense V2 | SkySense V2 · OSCD | F1: 60.06 -> 65.29 | Improved |
| replace SkySense with SkySense V2 | SkySense V2 · Change detection (3 datasets) | average score: 2.7% average gain | Improved |
| replace SkySense with SkySense V2 | SkySense V2 · Semantic segmentation (4 datasets) | average score: 1.5% higher on average | Improved |
| replace SkySense with SkySense V2 | SkySense V2 · Object detection (DIOR, DIOR-R, FAIR1M) | mAP: 1.1% average mAP gain | Improved |
| replace SkySense with SkySense V2 | SkySense V2 · PASTIS-MM | overall accuracy: 1.0% average gain | Improved |
| replace SkySense with SkySense V2 at low training ratio | SkySense V2 · RESISC-45 (TR=20%) | OA: 96.32 -> 97.24 (SatMAE++ 97.48) | Mixed |
| add temporal Sentinel-2 series (S2-Ts) instead of single S2 image | SkySense V2 · PASTIS-MM | overall accuracy: 75.0 -> 85.5 | Improved |
| add S2 and S1 modalities to single S1 input | SkySense V2 · BEN-MM | accuracy: 86.5 -> 93.8 | Improved |
| add Planet. + S2 + S1 to Planet. only | SkySense V2 · Dyna.-MM | mIoU: 47.6 -> 48.9 | Improved |
| add modality-specific prompt tokens (ablation) | SkySense V2 · AID / RESISC-45 (k-NN) | k-NN accuracy | Improved |
| add Mixture of Experts scaling (ablation) | SkySense V2 · AID / RESISC-45 (k-NN) | k-NN accuracy | 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.
- SkySense — SkySense: A Multi-Modal Remote Sensing Foundation Model Towards Universal Interpretation for Earth Observation Imagery (CVPR 2024) · better on 18 of 18 shared results
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 0.94)
- RobSense: A Robust Multi-modal Foundation Model for Remote Sensing with Static, Temporal, and Incomplete Data Adaptability (CVPR 2025 · similarity 0.88)
- Parameter Efficient Self-Supervised Geospatial Domain Adaptation (CVPR 2024 · similarity 0.88)
- SkySense-O: Towards Open-World Remote Sensing Interpretation with Vision-Centric Visual-Language Modeling (CVPR 2025 · similarity 0.88)
- Meta-Learning Hyperparameters for Parameter Efficient Fine-Tuning (CVPR 2025 · similarity 0.87)
- SkyScript: A Large and Semantically Diverse Vision-Language Dataset for Remote Sensing (AAAI 2024 · similarity 0.87)
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.
