Researchers have developed a novel deep-learning architecture that generates high-quality synthetic images of rare land covers to eliminate bias and drastically improve the accuracy of environmental monitoring systems.
