SpaceNet: Accelerating Machine Learning for Foundational Mapping Challenges

Опубликовано 20 июня 2019, 20:14

SpaceNet is a nonprofit LLC designed to accelerate machine learning against geospatial problems, such as mapping road network routes after a natural disaster using exclusively remote sensing data. Over the last two and half years, SpaceNet has released over 6500 sq km of high-resolution satellite imagery, with ~800,000 building footprint labels and 8000 sq km of road network labels. In addition to open sourcing a large, curated data set, SpaceNet has developed and administered four data science challenges to solve the problem of extracting building footprint and road networks from satellite imagery at scale. We will discuss the challenges of deploying these machine learning algorithms in operational timelines, and how AWS products be used to accelerate delivery of timely information derived from satellite imagery after a natural disaster. We will also highlight upcoming analytic challenges.