Besides making street-level imagery available to anyone, Mapillary also uses computer vision to automatically extract map data from images. You can use two types of data.
1) Object detections are instances of different objects that have been detected in images. Since Mapillary images are geotagged, you can get a dataset of image locations for a particular object that interests you, and use that as a filter to quickly find and look at all images where that object is present (learn more).
2) Map features entail different objects positioned on the map. If the same object has been detected in multiple geotagged images, we can use triangulation to estimate the location of the object (learn more).
- Points—point features for 42 object classes
- Traffic signs—also point features but broken down to 1,500 different traffic sign classes
Object detections and point features (including traffic signs) are freely available to view in the web app, along with the ability to download geodata in the form of GeoJSON files.
How the data is created
Mapillary’s computer vision technology automatically detects what kind of objects have been captured in images, such as buildings, cars, pedestrians, traffic signs, bicycle racks, and much more (altogether 97 object classes). This is based on a method called semantic segmentation—an algorithm is trained to detect and assign a category label to every pixel in the image.
Example of object detections in an image
Another part of our computer vision technology deals with 3D reconstruction of places from the images. By connecting object detection and 3D reconstruction, we can triangulate the location of an object that has been detected in several images. That way, we can estimate the location of that object in the real world and place it on the map as a map feature.
In mapping terminology, map features are divided into three categories based on how an object is represented on the map: as a point, a line, or a polygon. Mapillary generates map features as points.
Example of map features generated for traffic signs
Our point features include 42 different objects plus 1,500 different types of traffic signs.
To clarify, traffic signs are point features. But because Mapillary is able to distinguish between different subclasses of traffic signs, they're often brought out separately, e.g. in the interface of the Mapillary web.
A selection of Mapillary's map data offering, displayed in the web app