Utilizing an image exposure transformation neural network to generate a long-exposure image from a single short-exposure image

1. A non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause a computer system to:

receive a short-exposure target image;

determine an attention map for the short-exposure target image utilizing an appearance guided attention neural network, the attention map comprising transformation regions and corresponding magnitudes for the short-exposure target image;

generate a synthesized long-exposure image from the short-exposure target image and the attention map utilizing an image exposure generator neural network by applying long-exposure effects to the short-exposure target image based on the transformation regions and the corresponding magnitudes; and

display, to a client device, the synthesized long-exposure image within a graphical user interface.

2. The non-transitory computer-readable medium of claim 1 , further comprising additional instructions that, when executed by the at least one processor, cause the computer system to receive the short-exposure target image from a live camera capture of a mobile client device.

3. The non-transitory computer-readable medium of claim 1 , further comprising additional instructions that, when executed by the at least one processor, cause the computer system to receive the short-exposure target image from a storage medium.

4. The non-transitory computer-readable medium of claim 1 , further comprising additional instructions that, when executed by the at least one processor, cause the computer system to determine the attention map for the short-exposure target image by generating spatial context predictions of the short-exposure target image utilizing an appearance guided attention neural network.

5. The non-transitory computer-readable medium of claim 1 , further comprising additional instructions that, when executed by the at least one processor, cause the computer system to determine magnitudes corresponding to the transformation regions within the attention map that comprise weighted parameters indicating motions influences within the transformation regions.

6. The non-transitory computer-readable medium of claim 5 , further comprising added instructions that, when executed by the at least one processor, cause the computer system to generate the synthesized long-exposure image utilizing the image exposure generator neural network by applying the long-exposure effects to the transformation regions of the short-exposure target image according to the weighted parameters of the corresponding magnitudes.

7. The non-transitory computer-readable medium of claim 1 , further comprising additional instructions that, when executed by the at least one processor, cause the computer system to generate the synthesized long-exposure image utilizing the image exposure generator neural network further based on input modifying an exposure magnitude to one or more transformation regions of the attention map.

8. The non-transitory computer-readable medium of claim 1 , further comprising additional instructions that, when executed by the at least one processor, cause the computer system to generate the synthesized long-exposure image utilizing the image exposure generator neural network by applying the long-exposure effects to learned features of the short-exposure target image based on attention weights within the attention map in accordance with the transformation regions and the corresponding magnitudes.

9. The non-transitory computer-readable medium of claim 1 , further comprising additional instructions that, when executed by the at least one processor, cause the computer system to generate the synthesized long-exposure image utilizing the image exposure generator neural network further based on an optical flow prediction.

10. The non-transitory computer-readable medium of claim 9 , further comprising added instructions that, when executed by the at least one processor, cause the computer system to generate the optical flow prediction utilizing an optical flow prediction network that indicates pixel movement and direction data in the short-exposure target image.

11. A system for synthesizing long-exposure images from single short-exposure images comprising:

a memory device comprising:

a short-exposure target image;

an appearance guided attention neural network; and

an image exposure generator neural network;

at least one computing device configured to cause the system to:

generate an attention map that comprises a spatial context of the short-exposure target image utilizing the appearance guided attention neural network on the short-exposure target image;

generate a synthesized long-exposure image from the short-exposure target image and the attention map, utilizing the image exposure generator neural network, by applying long-exposure effects to learned features of the short-exposure target image based on attention weights within the attention map; and

provide the synthesized long-exposure image for display.

12. The system of claim 11 , wherein the at least one computing device is further configured to cause the system to determine the attention map for the short-exposure target image by generating spatial context predictions of the short-exposure target image utilizing an appearance guided attention neural network.

13. The system of claim 11 , wherein the at least one computing device is further configured to cause the system to determine the attention weights within the attention map to indicate magnitudes for corresponding transformation regions within the attention map.

14. The system of claim 13 , wherein the at least one computing device is further configured to cause the system to generate the synthesized long-exposure image utilizing the image exposure generator neural network by applying the long-exposure effects to the transformation regions of the short-exposure target image according to corresponding magnitudes as indicated by the attention weights.

15. The system of claim 11 , wherein the at least one computing device is further configured to cause the system to provide the synthesized long-exposure image for display on a graphical user interface of a mobile client device.

16. In a digital medium environment for editing electronic images, a computer-implemented method of transforming exposure time of images, comprising:

receiving a short-exposure target image;

determining an attention map for the short-exposure target image utilizing an appearance guided attention neural network that indicates transformation regions and corresponding magnitudes in the short-exposure target image;

generating a synthesized long-exposure image from the short-exposure target image, utilizing an image exposure generator neural network, by applying long-exposure effects to the transformation regions of the short-exposure target image in accordance with the corresponding magnitudes as indicated in the attention map; and

providing the synthesized long-exposure image for display within a graphical user interface.

17. The method of claim 16 , further comprising receiving the short-exposure target image from a live camera capture of a mobile client device.

18. The method of claim 16 , further comprising generating the synthesized long-exposure image utilizing the image exposure generator neural network further based on input modifying an exposure magnitude to one or more transformation regions of the attention map.

19. The method of claim 16 , further comprising generating the synthesized long-exposure image utilizing the image exposure generator neural network further based on an optical flow prediction.

20. The method of claim 16 , further comprising determining, for each transformation region of the transformation regions, one or more features, wherein the corresponding magnitudes comprise an long-exposure importance of the one or more features within a transformation region.