How Mapillary works
What do we do with computer vision?
Find out what happens behind the scenes to turn images into connected street-level views and extract map data at scale.
Mapillary uses technology designed to blur any faces or license plates our algorithms detect to help protect privacy. Tools are available for manual edits if needed.
Each image is pixel-wise labeled with 97 object classes. We have the best-performing algorithm for semantic segmentation of road imagery, based on Cityscapes and Mapillary Vistas benchmarks
Mapillary uses Structure from Motion to find the relative positions of images and create smooth transitions between them. Our OpenSfM pipeline is available as open source.
Traffic sign recognition
We detect more than 1,500 types of traffic signs in 100 countries. When the same sign appears in multiple images, its position in space is triangulated and the traffic sign is extracted as a map feature.