5 ways how neural networks help creative community.
One can think that neural networks and AI technology is a thing far away from designers, photographers and creators. This is dead wrong.
When we launched a simple one-button super-resolution service we had no idea who is going to use it, if anyone. In less than 2 months we processed more than 500k real users images, using cutting-edge neural networks. I was quite curious about use cases, so I contacted some of the people who twitted or posted about the service to get an idea of the real-world usage of neural networks technology.
Here’s the top 5 applications (sometimes surprising) of super-resoltution technology in the hands of graphic professionals.
Upscaling posters and photos for printing
One of the first industries that comes to mind, thinking about image quality is printing.
Have you ever downloaded an image from the internet and then printed it, only to get results that were, well, less than you expected? The image looked great on your computer screen, but when you printed it, it either printed at the size of a postage stamp or it printed at a decent size but looks blurry or “blocky”? The reason for all this is image resolution. There simply are not enough pixels in most internet images to allow them to print at high quality, at least not without printing them in the small size.
The only way out of this situation is to increase the number of pixels in the picture, which was technically not possible without super-resolution technology which lies as the background of Let’s Enhance. Or simply finding another image :)
Increasing logos and images for design
Design is all about high quality and visual content and pictures need to be the proper size to make it work. We talked with almost 100 designers, since we started working on the project and most of them are constantly getting images from customers or the Web, that…well, they just can’t use. This problem is so widespread, it was named in “Top 3 things most designers hate” list.
“Often, clients will send us an image that they pulled from Google or Facebook, without paying much attention to the image’s size. Then, they’ll ask us to “blow it up” or “make it bigger.” The problem is, while we can stretch something to be much larger on a computer screen, this does not increase the number of pixels in the image — it simply makes the pixels (or dots) larger. This stretched image may look fine on the computer screen, but it will come out grainy and unclear later. And no one wants that!” ©
Keeping old photos looking good
Modern world and technology development completely changed photo industry in the recent decade. Thanks to our digital cameras and high capacity memory cards, we can now document our family activities on a daily basis, without worrying about limiting our shots. But it wasn’t always like that. Old photos are still playing the role of keepsakes of the past that evoke memories of wonderful times spent with families and friends. However, these photos can fade, tear, crease, or get stained with time.
Old games and museums
Old family photo albums are not the only sphere that we want to keep alive. How about game industry? I’m sure, we all feel this tempting nostalgia, thinking about playing original “Super Mario” or “Pacman” game. With modern state of neural networks it’s absolutely possible to restore these interfaces and adapt to modern screen resolutions.
Creating Art objects
We did not really envision this scenario but it turns out it’s possible to generate unique art objects. Results look pretty amazing.
There are artists who use neural networks especially GANs to generate modern art (Quasimondo | Mario Klingemann, Artist working with Code, AI and Data). We at LetsEnhance are really excited opportunities which GANs bring to modern art. In fact our main upscaling neural network is also a GAN. You can see how intelligently it generates details such as leaves on a tree or ripples of water when they are missing in low-res picture.
And how do you use neural networks? Feel free to tell us or show your results, using #letsenhance hashtag :) And you can check out this service at letsenhance.io and get your stack of free processing.
