Stylizing input images

A computer-implemented method comprising:

• maintaining, by a computing system, data specifying respective parameter values for each image style in a set of image styles;
• responsive to receiving, by the computing system, an input comprising an input image and input style data identifying at least one image style to be applied to the input image to generate a stylized image that is in an input style:
• determining, by the computing system from the maintained data and the input style data, parameter values for the input style;
• and
• generating, by the computing system, the stylized image by processing the input image using a style transfer neural network that is configured to process the input image to generate the stylized image,
• wherein the style transfer neural network comprises a conditional instance normalization layer between a first neural network layer and a second neural network layer,
• wherein the conditional instance normalization layer is configured to, during processing of the input image by the style transfer neural network:
• receive a first layer output generated by the first neural network layer,
• normalize the first layer output to generate a normalized layer output;
• and
• transform the normalized layer output in accordance with the current values of the conditional instance normalization layer parameters to generate a conditional instance normalization layer output, and
• provide the conditional instance normalization layer output as an input to the second neural network layer;
• and
• wherein generating the stylized image comprises setting, for the processing of the input image by the style transfer neural network, the current values of the conditional instance normalization layer parameters to be the parameter values for the input style.