How we leverage AI to build great maps at massive scale
summary
Claims an automated pipeline that creates and updates300 million signs and 10 million traffic-light junctions worldwide by fusing fleet panoramic imagery, aerial/satellite data, OpenStreetMap, LIDAR and sensor feeds. Uses clustering and nearest-neighbor matching, state-of-the-art computer vision, NLP for harmful edits, and calibrated confidence scores to output single-feature positions and types. Publishes hundreds of layered outputs to a map platform, applies human-in-the-loop feedback and quality rules, and reports scalable, realtime updates though some cases still require manual correction.
AI-written summary · about
chapters (9)
- 0:00AI-driven map features
- 1:57Manual map creation challenges
- 4:51Accurate maps for navigation
- 6:43Automated map features needed
- 8:57AI for automatic sign creation
- 13:26Fusion and confidence scoring
- 16:43Map layers to end-user features
- 17:24Automating map feature complexities
- 18:09Closing remarks
related talks
10mMapping sidewalks and street furniture with AI and computer vision
15mTowards automated mapping of pedestrian path networks at scale
22mUrban Artificial Intelligence in Amsterdam
19mChallenges & Opportunities for Building Crowdsourced Mapping Services for Autonomous Driving
02mWe painted the whole town
32mData&AI in TomTom: How are are transforming mobility
22mIsland Mapping: Measuring Urban Footprints with Mapillary & Open Source Tools
44mMaps & Data