| CPC G06T 19/20 (2013.01) [G06N 3/08 (2013.01); G06T 11/001 (2013.01); G06T 15/06 (2013.01); G06T 17/00 (2013.01); G06T 2219/2024 (2013.01)] | 4 Claims |
|
1. A non-transitory computer readable medium comprising a style transfer program that, when executed, causes a server to perform:
receiving first image data from a buffer;
applying style transfer to the first image data based on one or more style images to generate second image data; and
providing the second image data,
wherein the program causes a server to further perform:
obtaining a neural network comprising:
mixing parameters based on a plurality of style images into a predetermined layer of a neural network;
optimizing the neural network based on the plurality of style images to provide a learned neural network; and
providing the buffer data into a learned neural network,
wherein applying in the style transfer function comprises:
providing the first image data into the learned neural network; and
applying the style transfer based on the plurality of style images to the first image data,
wherein optimizing comprises:
optimizing based on two style images selected from the plurality of style images; and
optimizing based on one style image of the plurality of style images,
wherein optimizing based on the two style images selected from the plurality of style images comprises:
obtaining a first value by performing a predetermined calculation on a generated image denoted by p;
obtaining second and third values by performing the predetermined calculation on the two style images denoted by q and r;
obtaining an average value of the second and third values; and
adding norms between the first value and the average value, and
wherein a function optimizing based on the two style images is represented by an equation:
 |