Semantic preserved style transfer

A system for image style transfer using Semantic Preserved Generative Adversarial Network (SPGAN), comprising:

• a plurality of sensors, wherein at least one of the sensors is a camera;
• a processor in communication with the plurality of sensors;
• wherein the processor is programmed:
• receive a source image, wherein the source image was captured by a camera, the source image includes a source semantic content, and the source semantic content is objects in the source image and an arrangement of the objects in the source image;
• input the source image into the SPGAN, wherein the SPGAN includes a first generator network including a first encoder and a first decoder, and a first discriminator network, wherein the SPGAN runs on the processor;
• extracting, by the first encoder of the first generator network, a source-semantic feature data from the source semantic content of the source image;
• generate, by the first decoder of the first generator network, a first synthetic image including the source semantic content of the source image in a target style of a target image using the source-semantic feature data extracted by the first encoder of the first generator network, wherein the target style is a spatially-average colors and a texture of the target image, and the first synthetic image includes first-synthetic feature data;
• determine a first encoder loss using the source-semantic feature data and the first-synthetic feature data;
• discriminate, using the first discriminator network, the first synthetic image generated by the first generator network against the target image to determine a GAN loss;
• determine a total loss as a function of the first encoder loss and the GAN loss;
• and
• train the first generator network and the first discriminator network using the first encoder loss and the GAN loss until the total loss is equal to or less than a predetermined loss threshold or reach the maximum number of training iterations in order to minimize image distortion during the image style transfer.