The analysis result
- The probability mask computed by the model (where it thinks it sees a pool);
- The number of pools detected and the confidence score;
- The tile coordinates (public map indices, e.g.
z19 / x / y).
Privacy
When you run the worker, only minimal anonymous information is sent. Here it all is, without exception.
When the worker analyzes a tile, it sends back to the server only:
z19 / x / y).Device information also helps us diagnose inference problems related to specific hardware — for example if the model behaves differently on a particular GPU or browser.
This information is sent only when you run the worker, and solely for assigning and coordinating contributions.
The AI model runs in your browser: your device does the computing, not a remote server. The server sends you aerial imagery to process and receives the result back — see the Technology and The model pages.
Your ID is a simple UUID generated by your browser and sent with your contributions. It is not linked to any personal information: it’s just a label to associate the tiles you analyze with you.
You can change it at any time from the app — a new ID is generated and the old one is no longer used. However, if you keep the same ID, it’s easier to track your contribution history: the total number of tiles analyzed and pools found remains associated with you across sessions.
The data you produce — masks and validations — will serve to enrich OpenStreetMap, an open geographic database. So it will be public by design and contain nothing personal: only the position of objects visible from the sky (swimming pools).