Artificial Intelligence

OSM-ready data sets: Improving OpenStreetMap with Esri and Map With AI

By Jason Sundram
July 13, 2020

Contributing to OpenStreetMap (OSM) can be a daunting prospect. Even seasoned geographic information systems (GIS) analysts take time to learn OSM’s tagging and editing conventions, tooling, and jargon. Updating a map based on third-party data sets is even more challenging. To do that, a contributor might need to play four different roles: An engineer (to write software or extend JOSM or iD); a lawyer (to ensure compliance with an OSM-compatible license); an algorithm expert (to create a method for conflating new data with existing features); and an OSM expert (to tag new map features so they’re compatible with surrounding ones).

To simplify this process, we have partnered with Esri to release new OSM-ready data sets. Esri’s ArcGIS Hub, already a valuable source of geo data, and ArcGIS Online platform now includes authoritative, OSM-ready data sets. These data sets are OSM-tagged, compatibly licensed, and available to use for building the map. To work with these new data sets, we are updating our Map With AI tech stack, which previously supported only AI-derived data. Rather than placing the onus on OSM contributors to annotate, ingest, conflate, tag, and ensure license compatibility, this new combined approach provides:

Recommended Reading
  1. New releases of RapiD and the JOSM plugin that talk directly to Esri’s APIs.
  2. An Esri-curated data group specifically for OSM-ready data sets: Tagging and licensing are verified ahead of time. Each data set in this data group is homogeneous (roads only, buildings only, etc.), allowing mappers to pick and choose the data they wish to add.

This approach opens up new horizons to craft mappers and frees them from the onus of wearing four different hats, but we recognize that it places more responsibility on data publishers to establish proper tagging and licensing. To aid aspiring publishers of OSM-ready data sets, Esri has created an onboarding process. You can learn more about it in Esri’s blog post

The Revolutionary Government of Zanzibar, supported by the World Bank, was one of the first to go through this process — providing a data layer containing 500,000 buildings in Zanzibar produced by Zanzibari civil servants and students from the State University of Zanzibar through the Zanzibar Mapping Initiative and Open Cities Zanzibar projects.

“The Government of Zanzibar is excited to be one of the first to contribute OSM-ready data sets. It will be great to have this high-quality, open data available to all on OpenStreetMap, and we hope to see the development of new geospatial applications that will give Zanzibar the opportunity to better plan for a resilient future,” says Dr. Iddi Hassan, Executive Secretary, Zanzibar Commission for Lands. 

The OSM community can start exploring these new data sets with one of two beta builds of our Map With AI tools:

  1. RapiD
  2. Map With AI plugin for JOSM 

These tools have been updated to work with OSM-ready data sets. Currently, two types of data are available: building footprints and address points. As more data types (e.g., roads, points of interest, water features, etc.) become available in the Esri data group, they will also be accessible in the tools. 

Map With AI plugin showing Esri-provided buildings in Sarpy County, Nebraska.
Map With AI tools have a very user-focused design; nothing is added to the map without explicit user say-so. We’ve extended this design to allow users to pick and choose exactly which Esri data they wish to layer onto the map. Seen here: Esri-provided address points (orange) and Microsoft AI-detected buildings (pink).

We believe these OSM-ready data sets will make it much easier for mappers to add high-quality features, which will help us make the best map possible, together. This benefits not only the OSM mapping community but also data creators, as their data will easily be incorporated into a wide variety of applications. For those who are using our tools, the Map With AI team welcomes your feedback. Tweet us @MapWithAI or find us in the #mapwithai_feedback channel of the OSM US Slack. To submit an issue, you may do so here for RapiD or here for the MapWithAI plugin for JOSM. 

We look forward to mapping with you!

Thanks to Benjamin Clark and the World AI maps team at Facebook Boston for their work on this release, and the original Esri team, whose 2017 work on iD paved the way for this collaboration.

We're hiring AI scientists!

Help us drive scientific breakthroughs in core AI research

Artificial Intelligence

Meta creates breakthrough technologies and advances AI to connect people to what matters and to help keep communities safe.