TensorZoom — scale photo using neural networks

Currently, manufacturers for the quality of approximation of pictures using optical zoom. Digital zoom is not able to zoom the picture. However, companies can save a lot of money, if you are developing a software function approximation based on artificial intelligence. App TensorZoom is an excellent example of such an implementation of neural networks.

In TensorZoom you need to download a photo from the gallery or use the camera, there is a demo picture with the kitten. We uploaded a picture of Nissan GTR for analyzing algorithm performance.

Zooming, we compared it without TensorZoom and algorithm enabled algorithm. The result is pleasantly surprising. Of course, there is dampness, the algorithm works well though, but not trying to approximate the pixels in a very strong approximation.

Our goal is to show that in the future using neural networks it will be possible to zoom the image like in the movies. Until, of course, the quality of the algorithms on a rather average level. By the way, not so long ago Nvidia introduced a new feature Noise2Noise – based AI, which cleans noise in photographs.

Share your opinion in the Telegram chat or in the comments below.

Application: TensorZoom
Developer: machrisaa
Category: Art & design
Version: v0.2.126
Price: Free
Download: Google Play

0 Comments on “TensorZoom — scale photo using neural networks”

  1. I like the helpful info you provide in your articles I will bookmark your weblog and check again here regularly

Leave a Reply

Your email address will not be published. Required fields are marked *