New Google tech brings CSI-style photo enhancement to life
New Google tech brings CSI-style photo enhancement to life
Enhancing a pixelated image CSI-style may no longer be completely far-fetched thanks to some new research from the Google Brain team. The tech uses a pair of neural networks to process a 8 x 8 pixels image and generates an approximation of the original.
One of the networks works as a "conditioning" element that maps the lower-res shot to similar higher-res examples to get a basic idea of what the image should look like. The other, the "prior" network, models sharper details to make the final result more plausible.
The results aren't perfect, and to be clear, the faces generated aren't those of the initial person, but an approximate recreation that represents the computer's best guess at the original. And even if the AI system gets the finer details wrong, it could be close enough to give you a better idea of what you're looking at. An indistinguishable blob might become clear enough to identify a location for example.
The police probably won't be using the tech to definitively identify a suspect just yet, but it's a promising start. In the meantime enjoy one of the greatest moments in television history:
Source: ArXiv.org (PDF) via Ars Technica