Machine Learning Applications in Mapping and Satellite Image Processing
In this presentation, we overview the ML-assisted techniques for the satellite image processing: improving the open source OpenStreetMap and generating the state-of-the-art population density map. The key part of our pipelines are classification and segmentation deep neural networks.
Danil Kirsanov works on ML problems in Facebook Spatial Computing group. He holds PhD in Applied Math from Harvard and prior to Facebook worked on scalable computing at Microsoft and MathWorks. His primary interests are in signal/image processing, numerical modeling and distributed machine leaning applications.