Researchers at the Polytechnic University of Pernambuco in Brazil developed a machine learning algorithm that can transform ancient maps into Google-style satellite images, which could help determine how land use has changed over time, as well as displaying the social and economic impacts of urbanization.
The researchers used an existing artificial intelligence (AI) tool, Pix2pix, which uses two neural networks—one that creates images based on the input set, and another that determines whether the generated image is fake or not—that ultimately create realistic-looking images based on the historical data.
The researchers said they are working to improve the resolution of the resulting images.
From IEEE Spectrum
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA