1 of 14

Improving OpenStreetMap �with Open Data and Open Source

Bryan Housel�OpenThePaths 2025�Future of Non-Motorized Mobility

2 of 14

About me

Background in Open Source tooling for editing maps
Former maintainer of OpenStreetMap’s iD editor, also other OSM tools
Recently, worked at Meta to develop the Rapid editor, a successor to iD
Rapid can edit, but also merge open data into OpenStreetMap�
Try it!�https://rapideditor.org/�https://github.com/facebook/Rapid

First a bit about me.. The issue of -

how our tools can influence the data that we collect

and play a role in the kinds of maps that we can make -

is something that I’ve been working closely with for the past 10 years or so.

For about 6 years I was the maintainer of OpenStreetMap’s iD editor,

and during my time with that project

we saw it as our mission to make map editing fun and easy.

We really wanted to grow the OpenStreetMap community as much as possible,

and that meant building a tool that a casual mapper could pick up and learn in 15 minutes or so,

and start contributing useful data without a lot of training.

More recently I’ve been working with Meta on the Rapid map editor,

which started out as a fork of iD but has since been mostly rewritten to improve performance

and make it more than just a map editor for OpenStreetMap, but also a tool for merging in open data..

3 of 14

Why Rapid?

OpenStreetMap already has a lot of data, but it is inconsistent
There exist public datasets that can be used to fill in the gaps

Publicly-available authoritative data from GIS Departments
AI-generated data detected from aerial imagery or streetlevel imagery

Needed a simple but performant map editor that could combine OpenStreetMap with external data layers
Usable by anyone, not necessarily trained GIS experts

I’ll start off with a quick intro to the Rapid map editor,

Meta launched this tool in 2019 because they wanted to more efficiently fill in the gaps of missing data in OpenStreetMap.

Some places in OpenStreetMap are very well mapped, others not so much.

And - because Meta based their maps on OpenStreetMap

they had business reasons for wanting a more complete map,

so in 2019 they used AI to detect roads from aerial imagery and they made that data openly available

And they needed a simple tool that could show both OpenStreetMap and the additional AI features -

and allow humans to verify and accept or ignore those suggestions.

Around this same time Microsoft released their global AI-detected building footprints and these made it into Rapid also.

And Esri was also interested in giving their clients an easy way to comply with open data laws and contribute their GIS data. �

So Rapid grew out of this need for several businesses to improve OpenStreetMap with open data.

4 of 14

Rapid Demo! Adding Microsoft Buildings to OpenStreetMap

It’s probably easiest to just show what Rapid does with a short demo -

Here I’m going to add some AI detected Microsoft Buildings to OpenStreetMap

in a place in Pennsylvania that doesn’t have any buildings mapped yet.

I’ll open the Rapid menu and enable the Microsoft Buildings, they’re drawn in this beautiful magenta color.

We’ll zoom in and take a closer look at the building shapes.

When I select a building, you’ll see some information about it in the sidebar.

And I can choose to Accept or Ignore it,

Now, these are suggestions by AI, so in some places this works better than others,

but around here they look pretty good.

For each one, I’ll click accept, then the building is added as a new OSM feature.

Then I can make whatever adjustments to it that I want, change the tags,

or move it around with the arrow keys.

And you don’t have to click the button each time, we do have keyboard shortcuts,

So our users can be very efficient in reviewing these suggestions.

Still, It’s important to us that there is a human in the loop - these aren’t like Automated Edits or Imports,

which the OpenStreetMap community discourages.

In Rapid the user does need to click on each suggestion and make a decision about it.

5 of 14

Open Data added to Rapid data catalog

2019: Facebook AI detected roads
2020: Microsoft AI detected buildings
2021: Esri community data sharing (buildings, addresses, and more)
2022: Rapid v2 - WebGL rewrite 🚀
2023: Google AI detected buildings
2024: Overture Places
Now: Open Map Footways: Sidewalk and Crossing data

Austin Communications and Technology Department Footway Centerlines
Boston BDPA Footway Centerlines
Delaware Valley Regional Planning Commission Philadelphia Footway Centerlines
Denver DRCOG Planimetrics Footway Centerlines
Salt Lake City Open Data Footway Centerlines

Over time we’ve grown the open data catalog in Rapid that our users can access

Like I said before Rapid started with Facebook’s AI roads, Microsoft’s AI buildings,

and added to it a growing catalog of data contributed by Esri’s GIS clients.

We knew that the performance of the tool struggled to keep up with even the base OpenStreetMap data,

so we spent most of 2022 improving the rendering code, releasing Rapid v2 in early 2023.

We rewrote the rendering code to switch from a SVG based map to a WebGL based map. More on that on the next few slides..

But the list of open datasets has continued to grow, both through the Esri portal

also adding Google AI detected buildings in 2023,

and then Overture’s open places dataset in 2024.

Overture is a really interesting project and I’m looking forward to seeing more open data releases from them.

Finally the most recent addition to Rapid’s data catalog is a dataset called Open Map Footways,

just added last December, which combines publicly available footway data from 5 cities around the United States.

Austin, Boston, Philadelphia, Denver, and Salt Lake City.

6 of 14

Performance comparison - Rapid v1, Cape Coral, FL

7 of 14

Performance comparison - Rapid v2, Cape Coral, FL

So we really needed to take a pause and rewrite Rapid’s core rendering technology.

For version 2 we replaced that legacy SVG rendering code

with a JavaScript video game library called PixiJS

and now we use WebGL to render the map.

It looks almost exactly like it did before, but now we’ve replaced all that SVG document manipulation

with a proper scene graph, and the GPU is rendering the map for us now and everything is much smoother.

Now even in this densely mapped place, we’re able to show everything with extra computing cycles to spare.

It’s a great example of really - the topic of this panel -

how the tool can influence or limit the kinds of data that we could collect.

As our mapmaking tools improve, we can build maps at higher levels of detail than was previously thought possible.

8 of 14

Rapid Data Catalog

So now, what kinds of open data can we work with?

I’ll show another example of the data catalog in action

Here we’re looking at Chester County, Pennsylvania, just outside of Philadelphia

And the map here only has roads and not much else.

I opened the data catalog and searched for “Chester”

And I enabled the address and building datasets provided by Chester County GIS

through the Esri Community data sharing program

Now the map shows all of these other features that can be accepted into OpenStreetMap.

Buildings in teal, addresses in yellow,

We also see some footway centerlines shown in magenta,

provided by the Delaware Valley Regional Planning Commission

Partnerships like these really make a better map possible!

If you have data in ArcGIS already, and the license is open,

it might be very easy to get it added to this catalog

9 of 14

Open Data Footways around Denver, CO

Finally one more example, I thought this audience might want to see a bit more detail about the Open Footways data.

This time we’re looking around Denver, Colorado.

This is a good place to look because the data is mostly good and the sidewalks are mostly unmapped in OpenStreetMap.

Here the magenta linestrings contain both sidewalks and crossings, provided by the Denver Regional Council of Governments.

Some people call them “doctor cog”, which I think is fun

But what’s great is that this dataset has already been matched up to the appropriate OpenStreetMap tags,

in the left sidebar it shows the tags like `highway footway` `footway sidewalk`

And the data also includes a common source tag, in this case they wanted to go with “Denver Open Data Catalog”

Finally one other useful thing to point out is that

the feature service we fetch this data from has conflated the linestrings against OpenStreetmap.

So in this screenshot, you might be able to see along the West side of Autumn Sage Street,

That yellow arterial road in the center,

there already is an OpenStreetMap sidewalk, so we don’t show an additional magenta sidewalk suggestion there

10 of 14

Combining Editing and Tasking

Rapid now has an integration with MapRoulette: https://maproulette.org/
MapRoulette is an Open Source microtasking tool for OpenStreetMap
Anyone can create “challenges” for teams of volunteers to work on
Great way get crossings and sidewalks added to OpenStreetMap!
Check out “Walkabout” project by Meta

🫶

So as we’ve seen, Rapid makes the open data available,

but we also need a community of people to review that data and actually add it to the map.

We’ve partnered with another great open source tool called MapRoulette.

MapRoulette is a microtasking tool.

And the way this tool works is: anybody can create challenges on the MapRoulette site for volunteers to work on.

A challenge might be something like “open the map in this spot and review certain features for correctness”

And the user is presented with some options like “I fixed it” or “I couldn’t complete it”

Meta has created an initiative called the Walkabout project,

where we partnered with YouthMappers and other local community groups

to improve the pedestrian network in cities of interest,

and this has been an valuable tool to organize volunteer mappers.

Last year the Rapid team added a built-in integration between Rapid and MapRoulette to make this even easier.

You can complete MapRoulette tasks without even leaving the Rapid tool.

And we’ve had a lot of success in getting pedestrian features mapped this way.

11 of 14

MapRoulette challenges in Rapid

Here’s a short video of the Maproulette integration in action

Meta used machine learning to detect places around Philadelphia

where we think that a crossing is mistagged or missing.

Our team created some MapRoulette challenges to investigate these intersections.

So here I enable the MapRoulette layer and then we see green markers where all the MapRoulette tasks exist

If you click on them, in the sidebar you’ll see some details about what the task is.

Here’s a place where some marked crossings are visible on the imagery but they are missing from the map.

And we can use Rapid to quickly add the missing crossings, and then click back on the green marker and say “I fixed it”

It kind of turns mapping into a game, but also gives our project coordinators some statistics on how much work is being done.

12 of 14

Results - Footways mapped

Source: https://openstreetmap.us/news/2025/03/pedestrian-data-trends/

“In 2024, OpenStreetMap experienced its largest recorded increase in pedestrian mapping, including footways and crossings. Across the top 10 U.S. cities, contributors added 9,896 km of footways and 62,153 individual crossings.”

So lets see some results!

Though the Walkabout project, Meta has been running challenges like this for a while

and we’ve seen some real success in improving the pedestrian network in OpenStreetMap,

in the cities that we focused on.

About a month ago, my colleague Chad Blevins published a post on the OpenStreetMap US blog,

where he shared the results of this work.

You should check out the post but I wanted to summarize some of the results for this audience.

This first chart shows an increase in kilometers of footways added to OpenStreetMap.

We tracked the increase in footway coverage in these cities and found that,

not only are footways being added, but we are speeding up!

In every city, we saw a bunch more added in 2024 vs 2023.

Some cities like Phoenix saw huge increases -

I think in that case a few very committed volunteers were able to make a huge difference.

13 of 14

Results - Crossings mapped

Source: https://openstreetmap.us/news/2025/03/pedestrian-data-trends/

“In 2024, OpenStreetMap experienced its largest recorded increase in pedestrian mapping, including footways and crossings. Across the top 10 U.S. cities, contributors added 9,896 km of footways and 62,153 individual crossings.”

We also have similar data on crossings:

We saw a similar trend in almost every city where

not only are we adding crossings, but the number added in 2024 is higher than the number added in 2023

everywhere except Los Angeles.

Crossings are super important to pedestrian safety and navigation but they can be a bit more difficult to map properly,

and adding them can be more time consuming.

They require more knowledge to identify the important tags such as markings, signals, and curbs.

And unlike sidewalks that are mostly visible from aerial imagery,

Crossings often require quality street level imagery to get a complete perspective of the crossing situation.

But! We are making progress in getting sidewalks and crossings mapped

and these are encouraging numbers that we hope to see increase.

14 of 14

🙏 Thank you

Get involved with the OSM US Pedestrian Working group�https://openstreetmap.us/�https://openstreetmap.us/get-involved/working-groups/�
Open your data!�
Support Open Source tools�Rapid: https://rapideditor.org/�MapRoulette: https://maproulette.org/

So to wrap up - thanks for listening!

I always like to end a talk with an “ask”, and for this one it’s to get involved!

Or - if you’re already involved, thank you for keeping these projects alive!

There are lots of ways to help out - probably the best resource right now is the OpenStreetMap US Pedestrian Working Group.

They are coordinating efforts on open data and mapping standards.

If you are a data owner, or even just know of any open data that you want to see added to tools like Rapid,

you can reach out to myself or the people in the Pedestrian Working group to take next steps in putting that data to use.

Finally support open source tools like Rapid and MapRoulette..

Support can come in many forms - don’t feel like you need to code..

Even if you just use the tools and provide feedback to the developers about what could be improved, that’s important too.

Thank you!