Remember all those crime fighting TV shows and movies where they are somehow capable of zooming in onto a low-resolution image, and with the magic of software they are able to sharpen the images and reveal certain things about the suspect? We guess to be fair it is fiction so anything’s possible, but Google is close to making that a reality.
Announced by the Google Brain team, it seems that they have managed to develop new software that uses AI to help sharpen low-resolution image files. How does this work? Basically with low-resolution images, there are less pixels involved which means that not all the details are there. However with the use of neural networks, the AI is able to fill in the details by mapping the lower-resolution image to that of higher-resolution examples.
This allows the system to get an idea of what the image should look like. This means that while painting an accurate picture is probably not the best idea or the most reliable, it could recreate the image to the point where a person’s face, object, or location is sort of identifiable. It could also be used in photo editing or the restoration of old photos where the AI could be used to enhance certain aspects of the image.
If you’re interested in learning more about Google Brain’s latest efforts, you can head on over to the Cornell University Library’s website and check out the paper they wrote about it.