“Mapping the World with Deep Learning” – Facebook Tech Talk by Derrick Bonafilia ’17, Yitong Tseo ’17, Sam Blackshear ’10, and Baoxiang Yiang.
Satellites today are capable of capturing geo-referenced imagery of almost the entire globe at resolutions ups to 50cm per pixel. Using this imagery, things like buildings and roads can be located, usually within cm to meters of their actual location. However, this imagery is a massive dataset and it would take many lifetimes to map it all out by hand. Learn how Facebook is leveraging deep learning to massively speed up the process, and how we’re working with partners like CIESIN, the Bill and Melinda Gates Foundation, Humanitarian OpenStreetMap and the Red Cross to use this data for humanitarian efforts and to get it widely and freely available to anyone looking to make the world a better place.