Systems and methods for generating holograms using deep learning

1. A method comprising, by a computing system:

generating a hologram by processing a first natural image using a machine-learning model, wherein the first natural image that is provided to the machine-learning model is a photograph of a real world;

generating a second image that is a reproduced representation of the first natural image by processing at least a portion of the hologram using a processing model that is configured to simulate interactions between a light source and the hologram;

comparing (1) the second image generated after the processing of the hologram and (2) the first natural image resulting before the processing of the hologram to calculate a loss based on a loss function; and

training the machine-learning model by iteratively updating the machine-learning model based on the loss between the first natural image and the second image, wherein the trained machine-learning model is configured to receive as an input a single natural image on which to add holographic effect and the trained machine-learning model is further configured to generate a final hologram based on the single natural image for display via a holographic display device.

2. The method of claim 1 , wherein the machine-learning model comprises one or more dilated convolutional blocks, and wherein the one or more dilated convolutional blocks are configured to utilize a propagation kernel to generate the final hologram.

3. The method of claim 1 ,

wherein the machine-learning model comprises at least one residual block that comprises sequentially connected layers, the sequentially connected layers comprising a first layer and a last layer; and

wherein the residual block is configured to add an input of the first layer to an output of the last layer.

4. The method of claim 1 ,

wherein the machine-learning model comprises a plurality of layers that includes a last layer and a next-to-last layer;

wherein the machine-learning model is configured to generate a concatenated result by concatenating an input of the machine-learning model with an output of the next-to-last layer; and

wherein the last layer is configured to use the concatenated result as input.

5. The method of claim 1 ,

wherein the hologram is a sub-hologram of a larger hologram;

wherein the first natural image is a portion of a larger natural image.

6. The method of claim 1 ,

wherein the processing model is configured to perform convolution of a phase definition on the hologram.

7. The method of claim 6 ,

wherein the convolution performed by the processing model is based on a propagation kernel.

8. The method of claim 7 , wherein the propagation kernel is defined based on an optical system designed for viewing the final hologram generated by the machine-learning model.

9. The method of claim 5 , wherein the sub-hologram has a size less than 50% of a resolution of the first natural image.

10. The method of claim 1 , wherein a resolution of the first natural image is at least 720p.

11. The method of claim 1 , wherein the trained machine-learning model is configured to generate images for wide-field-of-view displays which have a field of view larger than 40 degrees.

12. The method of claim 1 , wherein the trained machine-learning model is configured to generate images for near-eye displays, television monitors, cinema screens, computer monitors, mobile phones, or tablets.

13. One or more computer-readable non-transitory storage media embodying software that is operable when executed to:

generate a hologram by processing a first natural image using a machine-learning model, wherein the first natural image that is provided to the machine-learning model is a photograph of a real world;

generate a second image that is a reproduced representation of the first natural image by processing at least a portion of the hologram using a processing model that is configured to simulate interactions between a light source and the hologram;

compare (1) the second image generated after the processing of the hologram and (2) the first natural image resulting before the processing of the hologram to calculate a loss based on a loss function; and

train the machine-learning model by iteratively updating the machine-learning model based on the loss between the first natural image and the second image, wherein the trained machine-learning model is configured to receive as an input a single natural image on which to add holographic effect and the trained machine-learning model is further configured to generate a final hologram based on the single natural image for display via a holographic display device.

14. The media of claim 13 , wherein the machine-learning model comprises one or more dilated convolutional blocks, and wherein the one or more dilated convolutional blocks are configured to utilize a propagation kernel to generate the final hologram.

15. The media of claim 13 ,

wherein the hologram is a sub-hologram of a larger hologram;

wherein the first natural image is a portion of a larger natural image.

16. The media of claim 13 ,

wherein the processing model is configured to perform convolution of a phase definition on the hologram.

17. The media of claim 16 ,

wherein the convolution performed by the processing model is based on a propagation kernel, and

wherein the propagation kernel is defined based on an optical system designed for viewing the final hologram generated by the machine-learning model.

18. The media of claim 15 , wherein the sub-hologram has a size less than 50% of a resolution of the first natural image.

19. The media of claim 13 , wherein a resolution of the first natural image is at least 720p.

20. A system comprising: one or more processors; and one or more computer-readable non-transitory storage media coupled to one or more of the processors and comprising instructions operable when executed by one or more of the processors to cause the system to:

generate a hologram by processing a first natural image using a machine-learning model, wherein the first natural image that is provided to the machine-learning model is a photograph of a real world;

generate a second image that is a reproduced representation of the first natural image by processing least a portion of the hologram using a processing model that is configured to simulate interactions between a light source and the hologram;

compare (1) the second image generated after the processing of the hologram and (2) the first natural image resulting before the processing of the hologram to calculate a loss based on a loss function; and

train the machine-learning model by iteratively updating the machine-learning model based on the loss between the first natural image and the second image, wherein the trained machine-learning model is configured to receive as an input a single natural image on which to add holographic effect and the trained machine-learning model is further configured to generate a final hologram based on the single natural image for display via a holographic display device.