Research map / Vision-language and multimodal / Multimodal LLMs
SGP2: Coarse-to-Fine Controllable Multimodal Remote Sensing Image Generation
Pengyu Chen, Xi Yang, Nannan Wang
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 Grassmann Miner (ablation) | DiT baseline · Single-condition optical, Sketch | FID: 60.6 -> 59.3 | Improved |
| add Physical Prior Expert (ablation) | DiT baseline · Single-condition optical, Sketch | FID: 60.6 -> 56.8 | Improved |
| add spectral loss on top of Physical Prior Expert (ablation) | PPE · Single-condition optical, Sketch | FID: 56.8 -> 53.2 | Improved |
| add Grassmann Miner on top of PPE (ablation) | PPE · Single-condition optical, Sketch | FID: 56.8 -> 54.9 | Improved |
| add spectral loss on top of GM and PPE (ablation) | GM+PPE · Single-condition optical, Sketch | FID: 54.9 -> 41.8 | Improved |
| add all three components GM, PPE, spectral loss (ablation) | baseline [JoDI] · Single-condition optical, Depth | FID: 35.3% | Improved |
| replace ToRA text processing with Grassmann Miner | text conditioning · Grassmann Miner variants (mean FID) | avg FID: 45.6 -> 42.6 | Improved |
| add SVD in Grassmann Miner (ablation) | Grassmann Miner · Grassmann Miner variants (mean FID) | avg FID: 45.8 -> 42.6 | Improved |
| partition global and detail subspaces Ug and Ud (ablation) | Grassmann Miner · Grassmann Miner variants (mean FID) | avg FID: 44.3 -> 42.6 | Improved |
| add geodesic trajectory (ablation) | Grassmann Miner · Grassmann Miner variants (mean FID) | avg FID: 43.6 -> 42.6 | Improved |
| add dynamic spectral modulation Gamma(tau) (ablation) | Grassmann Miner · Grassmann Miner variants (mean FID) | avg FID: 43.1 -> 42.6 | Improved |
| fine-tune JoDI on MMEarth-1.5M | JoDI · RSICD test (text-to-optical) | FID: 181.36 -> 57.24 | Improved |
Closest papers by meaning
- MMM-RS: A Multi-modal, Multi-GSD, Multi-scene Remote Sensing Dataset and Benchmark for Text-to-Image Generation (NeurIPS 2024 · similarity 0.89)
- GeoLLaVA-8K: Scaling Remote-Sensing Multimodal Large Language Models to 8K Resolution (NeurIPS 2025 · similarity 0.86)
- DiffusionSat: A Generative Foundation Model for Satellite Imagery (ICLR 2024 · similarity 0.85)
- SSDiff: Spatial-spectral Integrated Diffusion Model for Remote Sensing Pansharpening (NeurIPS 2024 · similarity 0.85)
- EcoMapper: Generative Modeling for Climate-Aware Satellite Imagery (ICML 2025 · similarity 0.85)
- S2MAE: A Spatial-Spectral Pretraining Foundation Model for Spectral Remote Sensing Data (CVPR 2024 · 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.
