Machine learning-based 2D structured image generation

One or more non-transitory computer-readable media storing one or more sequences of instructions that, when executed by one or more processors, cause:

• using a training dataset, training a machine learning model to produce a first trained machine learning model;
• wherein the training dataset comprises one or more pairs of images, wherein each pair of images, of the one or more pairs of images, comprises an input image that visually defines a plurality of distinct image areas and a target image that depicts a texturized version of each distinct image area of the plurality of distinct image areas;
• receiving a request to texturize a particular input image that visually defines a particular plurality of distinct image areas;
• performing a multiple-phase operation to produce a styled texturized output image based on the particular input image;
• during a first phase of the multiple-phase operation:
• based on the particular input image, using the first trained machine learning model to infer a textured output image that depicts an automatically-generated texture for each distinct image area of the particular plurality of distinct image areas;
• during a second phase of the multiple-phase operation:
• based on (a) the textured output image produced during the first phase of the multiple-phase operation and (b) a style input image, using a second trained machine learning model configured to perform neural style transfer to infer the styled texturized output image that incorporates content of the textured output image and reflects one or more style aspects of the style input image;
• returning, as a response to the request, the styled texturized output image.