Historic Earth

deep learning ai remote sensing historical reconstruction stable diffusion cultural preservation

Proposed by University of Warsaw

A public library University located in Poland

Description

Imagine Google Earth but with a very dense time layer that can be set to any period between 3000 bc and today. Such a tool would greatly enhance learning across the glob by providing real-world context to events of the past. Currently there are many 10,000's of archival maps online, of those only a small subset have been digitized to the point that they can be searched semantically. Semantic search for maps in this case would look like this: "Every church fire taking place within the borders of modern-day Germany in the period 1000 AD to 1400 AD." The goal is to collect any historical data that includes reference to a place, be that photos, phonebooks, or maps and convert it all into the same data format so it can be searched and processed at scale.