Line drawing generation
Computing systems and computer-implemented methods can be used for automatically generating a digital line drawing of the contents of a photograph. In various examples, these techniques include use of a neural network, referred to as a generator network, that is trained on a dataset of photographs and human-generated line drawings of the photographs. The training data set teaches the neural network to trace the edges and features of objects in the photographs, as well as which edges or features can be ignored. The output of the generator network is a two-tone digital image, where the background of the image is one tone, and the contents in the input photographs are represented by lines drawn in the second tone. In some examples, a second neural network, referred to as a restorer network, can further process the output of the generator network, and remove visual artifacts and clean up the lines.
1. A computer-implemented method performed by one or more processing devices, the method comprising:
inputting a digital image into a first neural network;
generating, using the first neural network, a first digital line drawing based on contents of the digital image;
inputting the first digital line drawing, rather than the digital image, into a second neural network;
generating, using the second neural network, a second digital line drawing based on the first digital line drawing, wherein the second digital line drawing is a two-tone image and is different from the first digital line drawing; and
outputting the second digital line drawing as a line drawing corresponding to the digital image.
2. The computer-implemented method of claim 1 , wherein the second digital line drawing includes edges of objects in the digital image and one or more lines corresponding to specific features of the objects.
3. The computer-implemented method of claim 1 , wherein the first digital line drawing is a greyscale digital image, and generating the second digital line drawing comprises removing, by the second neural network, one or more digital artifacts from the first digital line drawing and outputting the two-tone image.
4. The computer-implemented method of claim 3 , wherein the second neural network is trained separately from the first neural network on a data set of line drawings and copies of the line drawings, wherein the copies of the line drawings include artifacts digitally added to the line drawings in the data set of line drawings, and the second neural network is trained to use the copies of the line drawings as input and output the line drawings.
5. The computer-implemented method of claim 1 , wherein the first neural network is trained on a data set including digital images and digital line drawings of the digital images, and wherein the digital line drawings include lines for edges and specific features of objects in the digital images.
6. The computer-implemented method of claim 5 , wherein lines for the specific features aid in recognizing the objects.
7. The computer-implemented method of claim 5 , wherein lines for the specific features provide shape or definition to parts of the objects.
8. The computer-implemented method of claim 5 , wherein the specific features include texture when the objects are in foregrounds of the digital images.
9. The computer-implemented method of claim 5 , wherein the specific features exclude textures.
10. The computer-implemented method of claim 1 , wherein the digital image depicts a part of a person or an architectural structure.
11. The computer-implemented method of claim 1 , wherein the digital image comprises a photograph of an outdoor scene or a building interior.
12. The computer-implemented method of claim 1 , wherein the second digital line drawing comprises a set of lines in a first tone and a background in a second tone, and wherein the set of lines are substantially uniform in width.
13. A computing device, comprising:
one or more processors; and
a non-transitory computer-readable medium including instructions that, when executed by the one or more processors, cause the one or more processors to perform operations including:
inputting a digital image into a first neural network;
generating, using the first neural network, a first digital line drawing based on contents of the digital image;
inputting the first digital line drawing, rather than the digital image, into a second neural network;
generating, using the second neural network, a second digital line drawing based on the first digital line drawing, wherein the second digital line drawing is a two-tone image and is different from the first digital line drawing; and
outputting the second digital line drawing as, a line drawing corresponding to the digital image.
14. The computing device of claim 13 , wherein the second digital line drawing includes edges of objects in the digital image and one or more lines corresponding to specific features of the objects.
15. The computing device of claim 13 , wherein the first digital line drawing is a greyscale digital image, and generating the second digital line drawing comprises removing, by the second neural network, one or more digital artifacts from the first digital line drawing and outputting the two-tone image.
16. The computing device of claim 13 , wherein the first neural network is trained on a data set including digital images and digital line drawings of the digital images, and wherein the digital line drawings include lines for edges and specific features of objects in the digital images.
17. The computing device of claim 13 , wherein
content of the digital image includes a part of a person or an architectural structure.
18. The computing device of claim 13 , wherein the digital image is a photograph of an outdoor scene or a building interior.
19. A non-transitory computer-readable medium having stored thereon instructions that, when executed by one or more processors of a computing device, cause the one or more processors to perform operations including:
inputting a digital image into a first neural network;
generating, using the first neural network, a first digital line drawing based on contents of the digital image;
inputting the first digital line drawing, rather than the digital image, into a second neural network;
generating, using the second neural network, a second digital line drawing based on the first digital line drawing, wherein the second digital line drawing is a two-tone image and is different from the first digital line drawing; and
outputting the second digital line drawing as a line drawing corresponding to the digital image.
20. The non-transitory computer-readable medium of claim 19 , wherein the second digital line drawing comprises a set of lines in a first tone and a background in a second tone, and wherein the set of lines are substantially uniform in width.