Detect sidewalk information from street level images

When computing optimal routes, considering the existence and conditions of sidewalks can be of great benefit for pedestrian and wheelchair route planning. You can contribute to improve relevant data by classifying images taken from

Done! View results


published results
tasks done
pending tasks

Although OpenStreetMap is a great source of open geographical information which can be used for a number of applications including scientific research and route planning, one aspect that is falling behind in the availability of information is that of sidewalks. A huge amount of information is available about things such as street names, the location of hotels and whether a particular pub is accessible for wheelchair users, but in terms of the surfaces we walk along there is surprisingly little information. Can we walk along the side of the road? Do we have to go up a large curb when crossing the road? Am I likely to trip over damaged paving blocks? All of these questions are currently almost impossible to even get a suggested answer based on what information is available in OpenStreetMap.

The purpose of this study is to start at the basic level of mapping sidewalk information – is there a sidewalk present or not? Presence of even this most basic of information allows a much wider variety of applications to make use of OpenStreetMap data effectively, such as planning routes for pedestrians rather than just cars.

The project aims at raising awareness of the accessibility of places for the general public, in particular those people with reduced mobility such as wheelchair users. One aspect of that project is the development of routing and navigation services which take into account particular restrictions encountered by such users including the incline of surfaces and curb height. But first things first – the systems need to know whether a there is a safe place for people to walk rather than directing them down an Autobahn in the hope that there is a sidewalk there!

To contribute, all you need to do is look at an image (obtained from Mapillary) and say whether you can see a sidewalk on the left side, right side, both sides of the street or whether you see no sidewalk at all. If there is no street on the picture (e.g. a hiking path, a cycleway), you can also indicate this. That’s it – no trying to work out the surface material or incline, just whether there is a sidewalk there or not. Even though this is simple information, it will go a long way to adding this missing information to the OpenStreetMap and so (in our case) help to generate routing instructions that can be tailored to those people with reduced mobility.