Abstract
We envision a new generation of urban foundation models that unify semantic, spatial, and topological relationships within and across point, line, and polygon features. Existing approaches typically encode different feature types separately and fuse them only at later stages, missing opportunities for truly integrated contextualization. To illustrate our vision, we present RegionContext, a framework that generates contextual region embeddings capturing both fine-grained spatial structures and higher-order semantics. Finally, we highlight key research directions for region contextualization in urban foundation models.
| Original language | English (US) |
|---|---|
| Title of host publication | URBANAI 2025 - Proceedings of the 3rd ACM SIGSPATIAL International Workshop on Advances in UrbanAI |
| Editors | Haoran Niu, Hao Xue, Liang Zhao, Femi Omitaomu |
| Publisher | Association for Computing Machinery, Inc |
| Pages | 94-95 |
| Number of pages | 2 |
| ISBN (Electronic) | 9798400721892 |
| DOIs | |
| State | Published - Dec 2 2025 |
| Event | 3rd ACM SIGSPATIAL International Workshop on Advances in Urban AI, UrbanAI 2025 - Minneapolis, United States Duration: Nov 3 2025 → Nov 6 2025 |
Publication series
| Name | URBANAI 2025 - Proceedings of the 3rd ACM SIGSPATIAL International Workshop on Advances in UrbanAI |
|---|
Conference
| Conference | 3rd ACM SIGSPATIAL International Workshop on Advances in Urban AI, UrbanAI 2025 |
|---|---|
| Country/Territory | United States |
| City | Minneapolis |
| Period | 11/3/25 → 11/6/25 |
Bibliographical note
Publisher Copyright:© 2025 Copyright held by the owner/author(s).
Keywords
- region contextualization
- spatial semantics
- urban foundation models
Fingerprint
Dive into the research topics of 'Region Context from Unifying Points, Lines, and Polygons'. Together they form a unique fingerprint.Cite this
- APA
- Standard
- Harvard
- Vancouver
- Author
- BIBTEX
- RIS