Research map / Image generation / Image restoration
MetaEarth: A Generative Foundation Model for Global-Scale Remote Sensing Image Generation
Zhiping Yu, Chenyang Liu, Liqin Liu, Zhenwei Shi, Zhengxia Zou
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 spatial resolution guidance (ablation) | MetaEarth · Generated 64m images, all regions | FID: 27.79 -> 23.97 | Improved |
| add spatial resolution guidance (ablation) | MetaEarth · Generated 16m images, all regions | FID: 16.00 -> 15.12 | Improved |
| add spatial resolution guidance (ablation) | MetaEarth · Generated 4m images, all regions | FID: 13.43 -> 13.32 | Improved |
| add spatial resolution guidance (ablation) | MetaEarth · Generated 4m* images, all regions | FID: 92.61 -> 84.75 | Improved |
| add spatial resolution guidance (ablation) | MetaEarth · Generated 64m images, Southern Hemisphere | FID: 42.11 -> 44.16 | Hurt |
| add spatial resolution guidance (ablation) | MetaEarth · Generated 16m images, Southern Hemisphere | FID: 25.50 -> 27.76 | Hurt |
| add overlapped sliding window (ablation) | MetaEarth · Seam gradient | Average directional gradient: 24.67 -> 17.88 | Improved |
| add noise sampling constraint on top of overlap (ablation) | MetaEarth · Seam gradient | Average directional gradient: 17.88 -> 12.58 | Improved |
| add high-order degradation in training (ablation) | MetaEarth · 16m to 4m generation | FID: 148.31 -> 78.72 | Improved |
| use self-cascaded generation instead of direct 256m to 4m generation (ablation) | MetaEarth · 4m 1024x1024 generation | visual quality | Improved |
| add MetaEarth-generated augmentation data (5x) | VGG19 · 7-class remote sensing classification | accuracy: 95.72 -> 97.28 | Improved |
| add MetaEarth-generated augmentation data (5x) | ResNet34 · 7-class remote sensing classification | accuracy: 96.11 -> 99.22 | Improved |
Closest papers by meaning
- DiffusionSat: A Generative Foundation Model for Satellite Imagery (ICLR 2024 · similarity 0.89)
- MMM-RS: A Multi-modal, Multi-GSD, Multi-scene Remote Sensing Dataset and Benchmark for Text-to-Image Generation (NeurIPS 2024 · similarity 0.89)
- GeoCore-9B: Towards Geo-Aware Generative Foundation Models in Earth Observation (NeurIPS 2026 · similarity 0.88)
- EcoMapper: Generative Modeling for Climate-Aware Satellite Imagery (ICML 2025 · similarity 0.87)
- DiffuseHigh: Training-Free Progressive High-Resolution Image Synthesis Through Structure Guidance (AAAI 2025 · similarity 0.86)
- Increasing the Utility of Synthetic Images through Chamfer Guidance (NeurIPS 2025 · similarity 0.85)
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.
