New Possibilities for Mapping Communities – Virtual Street Views and Machine Learning

Wheelmap relies on the local knowledge of many to provide data on the wheelchair accessibility of public places. But there are other approaches to mapping, too: The Project Sidewalk in the U.S., for example, uses the photos of Google Street View to identify barriers for wheelchair drivers.

Locate, review, tag – Wheelmap is based on a mapping strategy whereby the reality check of mappers on location is the crucial element in the collection of data.

Are all rooms of a location accessible without steps? Does a store have a mobile ramp that is not visible from the outside because it is only laid down when needed? These criteria, which determine the wheelchair accessibility of a place on Wheelmap, can only be verified on site. (With this, Wheelmap adopts the same mapping strategy that the OpenStreetMap community has been using to create a complete open data map of the world.)

But with regard to routing, there are other ways of determining how best to get from one place to another in a wheelchair. Thanks to advances in technology, there are other sources that can be used in addition to the information provided by mappers on site.

Project Sidewalk uses the virtual world of Google

One example is “Project Sidewalk“: A research team of the University of Maryland and the University of Washington specializes in virtual mapping and has collected unprecedented amounts of accessibility information about streets and sidewalks. In Project Sidewalk, volunteer users *virtually* walk through cities in Google Street View to label and assess sidewalk accessibility – a bit like a first-person video game.

Project Sidewalk is particularly unique because you do not need to be physically present to assess the accessibility of a street or sidewalk, instead the tool transports you there virtually via immersive imagery (from Google Street View).

Since its beta launch in Washington, D.C. in the fall of 2016, more than 700 users have contributed 150,000 accessibility labels for Washington DC streets! You can see some results on the website.

The collected data is shared with city authorities and used to develop new accessibility-friendly mapping tools (e.g., route planners for people in wheelchairs, new types of maps visualizing access), to train machine learning algorithms to automatically assess accessibility, and to create more transparency regarding accessible infrastructure (imagine a walkscore.com for sidewalk accessibility!).

Mapillary: Street view photos with the help of the crowd

Another example of the potential of photos is Mapillary. This application offers an alternative to Google Street View photos. It relies on crowdsourcing globally to create photo series that are linked to geo positions. The result is images that depict sidewalks in even greater detail than the current Google Street View because the photo producers are on foot or by bicycle.

Using machine learning and object recognition, the Mapillary application automatically identifies curbs, surfaces, and even steps at entrances. The OpenStreetMap community is using Mapillary’s photos more and more as a supplemental tool to mapping. With it, the information in the OSM becomes more comprehensible and accurate.

Communities and technology combined holds the biggest potential

Conclusion: Different strategies and technologies are necessary for different mapping requirements. The greatest opportunity to achieve a wide coverage of accessibility information and a good balance between data quality and data quantity is to combine virtual methods with on-site methods.

All these projects rely on the help of many volunteers. So, join in and rate Project Sidewalk walkways, create alternate street-view photos with Mapillary, complete OpenStreetMap and tag locations on Wheelmap.org!

All these projects rely on the help of many volunteers. So, join in and rate sidewalks on Project Sidewalk, create alternate Street View photos on Mapillary, complete OpenStreetMap and tag locations on Wheelmap.org.