Overture Maps Foundation releases beta dataset to power modern map-building

Maps have long helped us navigate, visualize and make sense of the world. 

And today, maps are no longer just a depiction of the world — they are an underlying infrastructure to it. They power routing, logistics, data visualization, local search and other emerging technologies such as autonomous driving and the metaverse.

However, the built elements of the world are constantly changing, meaning map data can often quickly fall out of date — making them unreliable, sometimes dangerously so. 

To help democratize map data and usher in the next era of map-making, the Linux Foundation in December 2022 launched the Overture Maps Foundation. The goal was to build a reliable, interoperable open dataset that could be used by anyone. 

Now, anyone can: Sixteen months after its launch, Overture is releasing the beta version of its map dataset. Any interested parties can begin testing with the data and schema for mapping applications and geospatial analysis. As they do, Overture will collect feedback and make incremental changes for future production releases. 

“Maps are incredibly hard to do: They cost hundreds of millions up into billions of dollars to build and maintain,” Marc Prioleau, Overture executive director, told VentureBeat. “The hard part is building all the data that supports the maps, creating the digital representation of the physical world.”

Five critical base layers for building modern maps

The Overture Maps Foundation was initially founded by Amazon Web Services (AWS), Meta, Microsoft and TomTom. Today, it comprises 28 member companies from some of the biggest names in the technology, geospatial and automotive fields. 

“The community’s really grown, which was part of the vision,” said Prioleau. “Today the best maps are the ones that are used the most. The best maps will be built by a big and diverse community.”

Overture allows companies to attach outside data to its base map via its Global Entity Reference System or GERS. In the beta release, unique identifiers known as GERS IDs will be attached to entities in the data. This will allow data to be added to the map “consistently, correctly and more quickly,” according to the foundation. 

Overture’s data comes from open map, community-built data sources including OpenStreetMap, as well as AI-derived data from satellite and aerial imagery and commercial and government sources. These are conflated into a single dataset and put through validation and quality checks to ensure its safe use. 

With the Overture Maps Data Schema, developers can use map data in a standard, documented and interoperable way. The beta release includes five unique base layers: 

• Base layer: This includes land and water data.
• Administrative boundaries: This provides a global open dataset of national and regional administrative boundaries. Regional names have been translated into more than 40 different languages.
• Transportation: This represents a worldwide network of roads, footpaths and other travel infrastructure.
• Buildings: This features 2.3 billion unique building footprints worldwide. Users can add data on top of this information, creating new use cases around property management, risk assessment or economic development.
• Places of interest: This layer includes data on nearly 54 million global places with associated confidence scores, enabling the building of map-based local discovery tools or the powering of local search engines. 

Prioleau pointed out that place data is one of the most difficult elements in maps. Fixtures such as roads change very little, but places change about 25% a year due to businesses closing or moving, so it’s difficult to know when data is reliable. “It’s a very volatile dataset,” said Prioleau. 

He noted that “if you’re in the mapping business, you know that no matter how good you are, something in your map is wrong because the world changed.”

The goal, however, is to eventually get to a point where those gaps are squeezed out. 

“The great thing is, it’s an open project; it’s free for anyone to take the data and use it in whatever their application is,” said Prioleau. “We operate with the ethos of release early, release often with full knowledge that maps are never done.”

Powering Bing Maps, helping underwrite insurance

In the previous alpha phase, Overture was building baselines of roads, places, buildings and boundaries and releasing data monthly for users to download and experiment with. Now, the goal is to promote use and create feedback loops in anticipation of a production version 1.0. 

“We’re saying to people, ‘Now’s the time to start using the data, trying it, evaluating it, giving feedback,’” said Prioleau. “We’re moving into the adoption and growth phase.” 

Overture members are already using its data for their products. Microsoft, for instance, is adding coverage for Bing Maps and TomTom is incorporating it into its Orbis Maps. GIS mapping software company Esri is using the data to publish new 2D and 3D in ArcGIS. Addresscloud — an 11-person, 9-year-old niche U.K. insurance provider — is leveraging it to help underwrite insurance for buildings based on datasets on flood and fire risk and tax rates. 

From the “reasonably frivolous” (say, Pokemon Go), to the highly serious (assessing climate change) and everything in between, many things rely on map data that is rich and trustworthy, said Prioleau. This includes the emerging metaverse and realms of augmented and mixed reality where “you’re in the map” and effects must line up for the experience to be believable. 

“Maps are an incredibly important part of infrastructure,” he said. To some extent everyone uses them, and we expect them to be accurate (and not drive us into a lake like Michael on the U.S. version of “The Office.” “Our expectations have grown so high, because honestly the data’s gotten better.”

The rapid evolution of map building

Maps have been around for millennia (the oldest known is the Babylonian Map of the World, dating to 500-700 B.C.E.). 

Of course, mapping has dramatically changed since then, more recently impacted by the internet, mobile, cloud compute and AI. 

“There’s been a huge change in mapping over the last 10 to 15 years,” said Prioleau. 

Whereas maps were previously actual physical traversal and surveys, today they are based on sensor data. For instance, Waze pushes data out about traffic and ingests data back that it incorporates into its platform. 

Prioleau pointed out that we often think of maps as something visual, but today’s machines use spatial data and don’t need the visual element. 

“Maps are hard, but there’s a lot of stuff happening in technology that’s really going to change the way maps are built,” said Prioleau.