Recapping State of the Map US 2023

Earlier this month, the Mapillary team attended State of the Map US 2023 to stay up-to-date on new developments in the open geospatial world and to give updates on what we’ve been working on. Various members of Mapillary and the Maps team at Meta gave talks detailing Meta’s approach to open mapping, tooling updates, and contributions to OpenStreetMap.
Navjyot Sandhu
Edoardo Neerhut
23 June 2023

Overview

Every year, OpenStreetMap U.S. hosts a conference centered around open mapping, bringing together mappers, software developers, users, humanitarian organizations, and other groups. At this year’s State of the Map US conference, members of Mapillary and the Maps team at Meta had the chance to give talks to shed some light on problem areas and projects we’ve been working on.

Talks

Mapillary Camera Grant Program

At State of the Map US, we had the pleasure of announcing that our Camera Grant Program was back after a long hiatus. In summary, the program is designed to support a common goal between Mapillary and OpenStreetMap for improving the availability of open geospatial data in the United States and fostering local open data contributor communities by granting 360º cameras to selected applicants. You can read more on our announcement blog post. If you meet the criteria below, you can apply by filling out this form.

Putting the AR Puzzle Together with OpenStreetMap

Augmented Reality (AR) presents an exciting frontier for anyone interested in maps. The smartphone era showed us that the way maps are consumed can have a big impact on the way they are created and vice-versa. Industries such as ridesharing were made possible because drivers and riders suddenly had accurate maps in their pocket.

With the potential to be a major computing platform, AR may spark some similarly exciting use cases, with map data at their core. By AR, we are talking about the overlay of digital content over the real world.

This talk walked through AR as we know it today, some of the potential use cases, and why open map data is needed to power the future. It also encouraged OSM to position itself as a map suited to this future by thinking about some of the puzzle pieces that are needed. Pedestrian mapping in particular will be needed as more people consume digital information while directly viewing the real world rather than their smartphone.

Getting to Know the New Rapid 2.0 Editor

In April of this year, Meta released Rapid 2.0, an updated version of the Rapid editor for OpenStreetMap. Rapid 2.0 builds on the strengths of its predecessor in providing a powerful and efficient browser-based map editing tool, driving a higher level of mapping detail, quality and accuracy. The new version of Rapid boasts a 10x performance increase and allows users to more efficiently make edits to OSM, even in areas that are dense with existing features. In addition, Rapid is adding support for sidewalk and footway data, pending validation of these datasets by Esri; other approved Esri datasets are continuously being added to Rapid via their Community Maps program. One of the features that has set Rapid apart since 2019 is still front and center: machine learning-generated roads and buildings from Meta and Microsoft data.

Also on the way for Rapid is the addition of Mapillary object detections, meaning that users will be able to view features extracted from Mapillary imagery and add them to OSM through Rapid. Another upcoming addition is the 3D viewer, which takes live building data from OSM and allows users to view 3D building data, which is useful for finding things like minimum and maximum building heights in an area.

The Rapid editor is now hosted at RapidEditor.org, where you can visit the new landing page for an introduction to the tool, and start editing at https://rapideditor.org/edit.

A review of Mapillary Traffic Sign Data Quality and OpenStreetMap Coverage

Traffic signs are a key feature for navigating and managing either vehicular traffic or pedestrian traffic safely, affecting all of us on a daily basis. However, traffic sign datasets are lacking on open government data portals as well as OpenStreetMap (OSM).

We performed an experimental project to evaluate the quality of Mapillary generated traffic sign data in a few US cities, meanwhile identify traffic sign data gaps on OSM to understand how Mapillary generated data could help fill this gap.

Mapillary generated map objects with more than 95% average recall rate with 2.16 meter of average horizontal positional accuracy for traffic signs. We also did further analysis on target traffic sign coverage on OpenStreetMap and we observed 89% data missing on OSM of stop signs in Bellevue, 70% in Redmond, and 84% in San Francisco.

These results show that Mapillary generated traffic signs can be used to enrich OSM data and complete missing traffic sign data on OSM.

Other Takeaways

State of the Map US was the first OpenStreetMap conference since the announcement of Overture. It was positive to see the widespread interest and improved understanding of the Overture Foundation. Marc Prioleau hosted a talk outlining Overture’s mission and the relationship between the Overture project and how data from OpenStreetMap (along with data from the other members of the Overture Foundation) will flow through Overture, with the end goal of creating reliable, interoperable open map data. Another important point that Marc Prioleau stressed was that the path forward for Overture is still a work-in-progress; the details of the project are coming out as they’re being finalized. We will include a link to a recording of this talk (and all others above) when available on the OSMUS YouTube channel.

Another key takeaway from the conference was the general attitude of the government towards open data and non-”authoritative” data sources. Dr. Lee Schwartz, the official Geographer of the Department of State, opened the conference with a message highlighting the idea that the usefulness and accuracy of map data is ultimately more important than where it comes from or who commissions it. In a similar vein, the role of AI-generated map data for improving maps is one that is still prime for advancement. As mentioned previously, the Rapid editor for OSM uses machine learning to suggest road and building features, but there is still much room for exploration.

This year is shaping up to be an exciting one in the world of open mapping, and conferences such as these serve to reinforce that the open mapping community is capable of accomplishing wonderful things. We hope to see many more people, new and old, at mapping conferences in the future. If you’d like to get in touch with the Mapillary team, please don’t hesitate to reach out! Our email is support@mapillary.zendesk.com.

/Nav & Ed