MVRL/VectorSynth

1---2license: apache-2.03tags:4  - controlnet5  - stable-diffusion6  - satellite-imagery7  - osm8  - image-to-image9  - diffusers10base_model: stabilityai/stable-diffusion-2-1-base11pipeline_tag: image-to-image12library_name: diffusers13---14 15# VectorSynth16 17**VectorSynth** is a ControlNet model that generates satellite imagery from OpenStreetMap (OSM) vector data embeddings. It conditions [Stable Diffusion 2.1 Base](https://huggingface.co/stabilityai/stable-diffusion-2-1-base) on rendered OSM text to synthesize realistic aerial imagery.18 19## Model Description20 21VectorSynth uses a two-stage pipeline:221. **RenderEncoder**: Projects 768-dim CLIP text embeddings of OSM text to 3-channel control images232. **ControlNet**: Conditions Stable Diffusion 2.1 on the rendered control images24 25This model uses standard CLIP embeddings. For the COSA embedding variant, see [VectorSynth-COSA](https://huggingface.co/MVRL/VectorSynth-COSA).26 27## Files28 29- `config.json` - ControlNet configuration30- `diffusion_pytorch_model.safetensors` - ControlNet weights31- `render_encoder/clip-render_encoder.pth` - RenderEncoder weights32- `render.py` - RenderEncoder class definition33 34## Citation35 36```bibtex37@inproceedings{cher2025vectorsynth,38  title={VectorSynth: Fine-Grained Satellite Image Synthesis with Structured Semantics},39  author={Cher, Daniel and Wei, Brian and Sastry, Srikumar and Jacobs, Nathan},40  year={2025},41  eprint={arXiv:2511.07744},42  note={arXiv preprint}43}44```45 46## Related Models47 48- [VectorSynth-COSA](https://huggingface.co/MVRL/VectorSynth-COSA) - COSA embedding variant49- [GeoSynth](https://huggingface.co/MVRL/GeoSynth) - Text-to-satellite image generation