Aesthetics-guided image enhancement
Methods and systems are provided for generating enhanced image. A neural network system is trained where the training includes training a first neural network that generates enhanced images conditioned on content of an image undergoing enhancement and training a second neural network that designates realism of the enhanced images generated by the first neural network. The neural network system is trained by determine loss and accordingly adjusting the appropriate neural network(s). The trained neural network system is used to generate an enhanced aesthetic image from a selected image where the output enhanced aesthetic image has increased aesthetics when compared to the selected image.
1. A computer-implemented method for generating images with enhanced aesthetics, the method comprising:
selecting an image to aesthetically enhance;
obtaining a segmentation map corresponding to the image;
converting, by a first neural network of a neural network system based on a generative adversarial type architecture, the image into an enhanced aesthetic image based on the image and the segmentation map, wherein the neural network system includes a second neural network for discriminating, based on the segmentation map and the enhanced aesthetic image, the enhanced aesthetic image from a reference image that has an aesthetic score greater than a threshold;
outputting the enhanced aesthetic image generated using the first neural network of the neural network system; and
outputting, using the second neural network, a binary evaluation of the enhanced aesthetic image.
2. The computer-implemented method of claim 1 , wherein the segmentation map is obtained by parsing the selected image to map image content for pixels in the image.
3. The computer-implemented method of claim 1 , wherein the first neural network acts as a generator in the trained neural network system based on the generative adversarial type architecture.
4. The computer-implemented method of claim 1 , wherein the trained neural network system further includes a deactivated second neural network that acts as a discriminator during training.
5. The computer-implemented method of claim 1 , wherein the image is selected from a set of images stored on a user device, the set of images taken using an imaging function of the user device.
6. The computer-implemented method of claim 1 , wherein the image is selected using a graphical user interface of an application running on a user device.
7. The computer-implemented method of claim 1 , wherein the enhanced aesthetic image is displayed via a display screen of a user device.
8. One or more non-transitory computer-readable media having a plurality of executable instructions embodied thereon, which, when executed by one or more processors, cause the one or more processors to perform a method, the method comprising:
selecting an image to aesthetically enhance;
obtaining a segmentation map corresponding to the image;
converting, by a first neural network of a neural network system based on a generative adversarial type architecture, the image into an enhanced image based on the image and the segmentation map, wherein the neural network system includes a second neural network for discriminating, based on the segmentation map and the enhanced image, the enhanced image from a reference image that has an aesthetic score greater than a threshold;
outputting the enhanced image generated using the first neural network of the neural network system; and
outputting, using the second neural network, an indication that the enhanced image is real.
9. The one or more non-transitory computer-readable media of claim 8 , wherein the method further comprising:
outputting the enhanced image, wherein the enhanced image has increased aesthetics when compared with the image input into the neural network system.
10. The one or more non-transitory computer-readable media of claim 8 , wherein the segmentation map is obtained by parsing the image to map image content for pixels in the image.
11. The one or more non-transitory computer-readable media of claim 8 , wherein the first neural network acts as a generator in the neural network system based on the generative adversarial type architecture.
12. The one or more non-transitory computer-readable media of claim 8 , wherein the neural network system further includes a deactivated second neural network that acts as a discriminator during training.
13. The one or more non-transitory computer-readable media of claim 8 , wherein the image is selected from a set of images stored on a device, the set of images taken using an imaging function of the device.
14. The one or more non-transitory computer-readable media of claim 8 , wherein the image is selected using a graphical user interface of an application running on a user device.
15. The one or more non-transitory computer-readable media of claim 8 , wherein the enhanced image is displayed via a display screen of a user device.
16. A computing system comprising:
means for obtaining a segmentation map corresponding to a selected image;
means for converting, by a first neural network of a neural network system based on a generative adversarial type architecture, the image into an enhanced aesthetic image based on the selected image and the segmentation map, and a second neural network for discriminating, based on the segmentation map and the enhanced aesthetic image, the enhanced aesthetic image from a reference image that has an aesthetic score greater than a threshold;
means for outputting the enhanced aesthetic image generated using the first neural network; and
means for outputting, using the second neural network, an indication that the enhanced aesthetic image is fake.
17. The system of claim 16 , wherein the segmentation map is obtained by parsing the selected image to map image content for pixels in the selected image.
18. The system of claim 16 , wherein a first neural network acts as a generator in the neural network system based on the generative adversarial type architecture.
19. The system of claim 16 , further comprising:
means for selecting the selected image, wherein the selected image is selected from a set of images stored on a device, the set of images taken using a camera function of the device.
20. The system of claim 16 , further comprising:
means for selecting the selected image, wherein the selected image is selected using a graphical user interface of an application running on a user device.