San Diego State University researchers are rolling out an AI-powered map that tracks where people living outdoors in San Diego County pitch their tents and how those encampments shift over time. The tool is meant to give officials and service providers a near real-time view of unsheltered migration that goes beyond the annual Point-in-Time count. Project lead Ming Tsou describes the homeless population as very dynamic, moving every single day, and says the system is designed to help agencies anticipate where outreach, hygiene stations and street medical teams will be needed next.
According to KPBS, the project, organized under SDSU’s San Diego Homeless and Health EquAlity Research Team (SDHEART), blends city, county and nonprofit data with machine-learning models that scan aerial and street-level imagery for tents and makeshift structures, then estimate occupancy based on tent size. Per SDSU’s SDHEART project page, the initiative is funded by a National Science Foundation Build & Broaden award and integrates multiple data sources to map spatiotemporal migration patterns and neighborhood vulnerabilities.
How the map works
Researchers train deep learning models to flag tents and similar structures in aerial photos and street imagery, then combine those detections with local datasets within interactive dashboards. ESRI reports that the outputs feed ArcGIS dashboards and Experience Builder displays that highlight density hotspots rather than exact addresses. A working prototype is publicly viewable on ArcGIS.
Field work and human checks
Technology is only part of the operation. Students drive through city neighborhoods to ground-truth what the AI spots and to log the encampments the models miss or the objects they misclassify, while outreach teams conduct brief surveys asking people why they stay in a particular location and where they were before, as reported by KPBS. The team stresses that this human work is central to the project, pairing algorithms with on-the-ground conversations so that each mapped point comes with context rather than being just another dot on a screen.
Privacy worries from advocates
Not everyone is thrilled about a public map that traces encampments. Some advocates warn that making this kind of data too visible could make unhoused people easier targets for policing, sweeps or harassment. As Axios reported, Joshua Bohannan of Father Joe’s Villages cautioned that the map "risks putting a target on homeless people," and critics argue that another map will not solve the county’s deep shortage of shelter beds and services.
Funding, partnerships and next steps
The SDHEART consortium shares dashboards and data visualizations with municipal partners and service providers to coordinate outreach and refine geo-privacy safeguards, according to ESRI. The work is supported through a National Science Foundation Build & Broaden award, detailed in NSF Award 2417568, and aims to help local governments and nonprofits position shelters, handwashing stations and street-medicine teams where the model predicts displacement is likely to occur.
What to watch
Data-driven mapping could make it easier for outreach workers to follow people to safer locations and focus limited resources where they might matter most, but much depends on how decision-makers use the information and whether they commit to humane responses rather than punitive ones. Researchers say they plan to maintain community engagement and field verification as they refine the models, while walking the line between useful visibility and protecting the privacy of people who already live much of their lives in public.
