UrbanFusion: Stochastic Multimodal Fusion for Contrastive Learning of Robust Spatial Representations

This repository provides the pretrained weights of the UrbanFusion model — a framework for learning robust spatial representations through stochastic multimodal fusion.

UrbanFusion can generate location encodings from any subset of the following modalities:

• 📍 Geographic coordinates
• 🏙️ Street-view imagery
• 🛰️ Remote sensing data
• 🗺️ OSM basemaps
• 🏬 Points of interest (POIs)

🔗 The full source code is available on GitHub, and further details are described in our paper.

---

📖 Citation

@inproceedings{muehlematter2026urbanfusion,
  title     = {UrbanFusion: Stochastic Multimodal Fusion for Contrastive Learning of Robust Spatial Representations},
  author    = {M{\"u}hlematter, Dominik J. and Che, Lin and Hong, Ye and Raubal, Martin and Wiedemann, Nina},
  booktitle = {International Conference on Machine Learning (ICML)},
  year      = {2026},
}

---