I always chortled when the Good Guys were looking at some grainy surveillance footage and they “enhanced” the image to read a license plate. But it’s a major research area – self-driving cars – and a new paper takes image super-resolution, the term of art, to a new level. See the picture after the jump.

In new paper, href=”” target=”_blank”>=”https://arxiv.org/pdf/1809.00219.pdf” target=”_blank”>ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks, researchers demo some amazing results.


The StorageMojo take
The problem with making stuff up to create details that can only be inferred from the original is noted by the researchers: “. . . hallucinated details are often accompanied with unpleasant artifacts.” As Morpheus might say, “I know exactly what you mean.” Trippy!

Courteous comments welcome, of course.