Introducing Smart Edits: AI-assisted map-editing for point features with JOSM and Mapillary

Today, we’re introducing a new capability in the Mapillary plugin for JOSM, which allows for adding AI-detected map features using an accelerated workflow. Objects that were detected from Mapillary street level imagery can now be added to the map with just a few clicks.

Two of the fundamental challenges for any map are to provide rich detail and to stay up-to-date. In recent years, leveraging machine learning has helped address these problems by automatically processing numerous data sources and feeding them into the map building and map update process. Extracting information from street level imagery using computer vision is one of these approaches. Mapillary has provided machine-extracted data alongside our imagery for several years. We have also been enthusiastic about supporting OpenStreetMap (OSM) with the data that we could extract from the imagery contributed by all our users and partners across the world. To that end, since 2014, we have provided map data in some of the most popular editors for OSM, including iD, JOSM, and RapiD. And just since this year, we also made all our data, including point features. available free to the OSM community through these tools.

With these integrations, you were able to browse the map features detected in Mapillary imagery overlaid on the map. You could then use that data to add a missing feature to the map. For example, if you found a missing bike rack, you could then open the dialogue to add a point feature, choose “bicycle parking”, enter the relevant details, and add the source Mapillary with the ID of the imagery. As you can see, while the AI-detected features accelerate the process of spotting missed objects, there are still a lot of steps to add the feature to the map. The obvious goal is to eventually be able to add and verify a Mapillary detected feature with a single click. And this is exactly what our new Smart Edits capability for the Mapillary JOSM plugin does.

When you encounter a missing feature, you can now simply click on it, choose “Add” and the relevant pre-filled edit dialogue opens. We are launching Smart Edits today for the following point feature types (we plan to expand the set of features soon):

  • Fire hydrants
  • Bike racks
  • Benches
  • Street lights
  • Traffic lights

Additionally, those who wish to enable expert mode in JOSM can access crosswalks as an experimental Smart Edit data class.

In the video below, Mapillary’s Chris Beddow walks you through editing with Smart Edits in JOSM in more detail:

We hope you are as excited as we are about this new capability. Try it out, and we welcome your feedback.

Happy holidays!

/ Chris & Till

Continue the conversation