Mapillary

Anyone can photograph streets for Mapillary with a smartphone or action camera and upload the pictures.¹ Computer vision connects the images across time and space into street-level views and extracts map data from them.¹ It started in 2013 and joined Facebook, now Meta, in 2020; its about page says members have contributed more than two billion images.¹ The service is provided by Meta Platforms Ireland Limited.²

Each image carries its capture time, compass angle and position, both as the camera recorded it and as computed after image processing.³ Each image belongs to a sequence, and images can be fetched as thumbnails up to 2,048 pixels wide or at original size.³ From the images Mapillary detects objects such as traffic signs and other street features; a map feature is an object located from detections in several images, and the detections in a single image, with their outlines in the picture, can also be read.³ The full lists of point features and traffic signs are published with the API documentation.³

The API has two parts: vector tiles for drawing coverage and map features on a map, and entity endpoints for the details of each image, sequence, map feature or detection.³ Every request needs an access token, and a developer must register each application to get one.^2,3 The entity endpoints search only small areas: since January 2026 a bounding box must be smaller than 0.01 degrees square, and the documentation points to the Mapillary Python SDK for larger areas.³ A radius search added in April 2026 finds images within 50 metres of a point.³ Without code, points and traffic signs visible on the web map can be downloaded as GeoJSON.³

Images are shared under CC BY-SA 4.0 unless Mapillary indicates otherwise, and the terms say some content may be offered under other licences such as CC BY-NC-SA.² The help centre gives the attribution it expects: the image title linked to the image, the contributor's username linked to their profile, and the licence.⁴ An application that serves Mapillary images from its own servers, or shows data extracted from them, must display the Mapillary logo and link back.² Applications must add something of their own; redistributing the content alone is not allowed.² Commercial use is allowed only for the purposes the terms list, such as building products, maps, studies and datasets or doing work for clients, and needs safeguards against re-identifying people or unblurring images.² Mapillary blurs faces and vehicle licence plates, and the terms forbid unblurring or re-identifying anyone.² Anyone may use any Mapillary image to derive data for OpenStreetMap, and its imagery and detections appear in the Rapid, iD and JOSM editors.⁵

For someone modelling a street, neighbourhood or road network, Mapillary is a way to see what is on the ground and when it was photographed: kerbs, signs, shopfronts, crossings, street furniture, the state of a road. The detected traffic signs and points can seed a layer of street assets, and the photos let a modeller check a map or an OpenStreetMap edit against the street itself. The photos come from whoever has contributed them, so their dates and gaps vary from street to street.