Researchers in Brazil have presented a way to turn old maps into satellite imagery. This approach will allow you to take into account the historical context when exploring individual parts of the planet.
Researchers at the Polytechnic University of Pernambuco have presented a machine learning algorithm that transforms old maps into Google satellite images. This development, according to scientists, can inform people about how the planet is changing over time, as well as inform about the social and economic consequences of urbanization.
To implement the project, they used an AI tool called Pix2pix, which relies on two neural networks. The first one creates images based on the input set, and the second network decides whether the generated image is real or not. The networks are then trained to trick each other and ultimately create realistic images based on the historical data provided.
Scientists describe their approach in a study published in the journal Geoscience and Remote Sensing Letters. In this study, they took an 1808 map of Recife (Brazil) and created modern images of the area.
“When you look at these images, you better understand how the city has changed over 200 years,” the scientists explain. “The city’s geography has changed dramatically – landfills have reduced water bodies, and green areas have disappeared as a result of human activity.”
The advantage of this approach to AI is that it requires relatively little data entry; however, data entry requires historical documentation and the resulting image resolution is lower than researchers would like.
“Next, we will work to improve image resolution and experiment with different input data,” the scientists note. The researchers hope that their development can be applied by city planners, anthropologists and historians.