Research map / Label-efficient and robust learning / Domain adaptation and generalization
CrossEarth: Geospatial Vision Foundation Model for Domain Generalizable Remote Sensing Semantic Segmentation
Ziyang Gong, Zhixiang Wei, Di Wang, Xiaoxing Hu, Xianzheng Ma, Hongruixuan Chen, Yuru Jia, Yupeng Deng, Zhenming Ji, Xiangwei Zhu, Xue Yang, Naoto Yokoya and 4 more
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 GSE module (ablation) | Rein · CASID ablation | avg mIoU: 59.2 -> 59.6 | Improved |
| add MIM decoder and loss (ablation) | Rein · CASID ablation | avg mIoU: 59.2 -> 60.4 | Improved |
| add Earth-Style Injection (ablation) | Rein · CASID ablation | avg mIoU: 59.2 -> 59.8 | Improved |
| combine GSE, MIM and Style (full CrossEarth) (ablation) | Rein · CASID ablation | avg mIoU: 59.2 -> 61.6 | Improved |
| use cross-attention injector instead of FFN (ablation) | CrossEarth · CASID ablation | avg mIoU: 61.2 -> 61.6 | Improved |
| use ASPP MIM decoder instead of linear (ablation) | CrossEarth · CASID ablation | avg mIoU: 59.1 -> 61.6 | Improved |
| lower style gate probability p from 0.9 to 0.1 (ablation) | CrossEarth · CASID ablation | avg mIoU: 58.7 -> 61.6 | Improved |
| use l1 instead of l2 for mask loss (ablation) | CrossEarth · CASID ablation | avg mIoU: 59.8 -> 61.6 | Improved |
| use l1 instead of l2 for delta loss (ablation) | CrossEarth · CASID ablation | avg mIoU: 60.5 -> 61.6 | Improved |
| CrossEarth instead of Rein on building extraction A2S | Rein · WHU Building A2S | mIoU: 30.8 -> 44.3 | Improved |
| CrossEarth instead of Rein on A2S-I | Rein · WHU Building A2S-I | mIoU: 48.0 -> 57.1 | Improved |
| CrossEarth instead of Rein on S2A | Rein · WHU Building S2A | mIoU: 66.8 -> 64.5 | Hurt |
Closest papers by meaning
- CrossEarth-Gate: Fisher-Guided Adaptive Tuning Engine for Efficient Adaptation of Cross-Domain Remote Sensing Semantic Segmentation (CVPR 2026 · similarity 0.92)
- Exploring Efficient Open-Vocabulary Segmentation in the Remote Sensing (AAAI 2026 · similarity 0.90)
- Towards Open-Vocabulary Remote Sensing Image Semantic Segmentation (AAAI 2025 · similarity 0.90)
- ROS-SAM: High-Quality Interactive Segmentation for Remote Sensing Moving Object (CVPR 2025 · similarity 0.88)
- SemStereo: Semantic-Constrained Stereo Matching Network for Remote Sensing (AAAI 2025 · similarity 0.88)
- GeoViS: Geospatially Rewarded Visual Search for Remote Sensing Visual Grounding (CVPR 2026 · 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.
