Mapillary’s computer vision extracts point data from street-level images, with these detections being used to fix and update maps. Earlier this year, Mapillary worked with OpenStreetMap communities to turn automatic computer vision detections into better map data. This was an insightful project that showed us which detections were useful for the OpenStreetMap community and end users. It’s also an early step to create a feedback loop where humans provide input to help us improve the quality and relevance of automatically derived features.