The neural network learned how to transfer the style from one photo...

The neural network learned how to transfer the style from one photo to another


A group of scientists from Adobe Systems (Creator of Photoshop) and Cornell University (USA) has developed a new technology that transfers the graphical style from one image to another. In contrast to Prisma, which takes as a sample the “art filters” based on the paintings of famous artists, the program Deep Photo Style Transfer imposes on the image of color and light, “borrowed” from the picture standard.

Combining images to make it look like the photo was taken at the other weather conditions or time of day. In this case the algorithm does not change the composition of the original image, leaving lines and shapes intact, due to what the final work looks very natural.

Left image is the original, in the centre a reference to the right — the end result

Read also: Russian programmers have created a neural network that evaluates the quality of the photo

Based on Deep Photo Style Transfer algorithms are the deep learning that “capture” the artistic style of the reference image. The neural network, for example, for a few minutes to make sure the Windows of a skyscraper was falling light, and then the man-Retoucher can spend on this task for several hours.

Maybe someday Adobe will make a new development available as a commercial product like Photoshop or Lightroom. And yet to see the source code of the program can any programmer downloading it from the repository on GitHub.

Source: The Next Web