Method and system for automated generation of representative icons from images

Patent metadata

Field	Value
Publication number	US-12136139-B2
Application number	US-202217710720-A
Country	US
Kind code	B2
Filing date	Mar 31, 2022
Priority date	Mar 31, 2022
Publication date	Nov 5, 2024
Grant date	Nov 5, 2024

Abstract

A method for automated generation of representative icons from images involves obtaining an image, generating an image vector from the image using an encoder deep neural network, generating an icon based on the image vector using a generative adversarial network (GAN), and outputting the icon.

First claim

What is claimed:

1. A method, comprising:
obtaining an image;
generating an image vector from the image using an encoder deep neural network;
generating an icon based on the image vector using a generative adversarial network (GAN);
training at least one of:
the encoder deep neural network to generate the image vector from the image using a set of labeled training images, and
the GAN to generate the icon based on the image vector using a set of labeled training icons; and

outputting the icon.

2. The method of claim 1 ,
wherein the icon is one selected from a group consisting of a shape, a logo, and an emoji.

3. The method of claim 1 , wherein the icon depicts at least one selected from a group consisting of an object, an action, and an emotion that is present in the image.

4. The method of claim 1 , wherein the image vector establishes a joint image-icon vector space.

5. The method of claim 1 , further comprising:
translating the image vector into a caption.

6. The method of claim 1 , further comprising, prior to generating the image vector:
preprocessing the image, comprising: transforming the image to a format compatible with the encoder deep neural network.

7. The method of claim 1 , wherein outputting the icon comprises one selected from a group consisting of displaying the icon to a user, and storing the icon.

8. A system, comprising:
an icon generation engine comprising at least one processor, the icon generation engine configured to:
obtain an image;
generate an image vector from the image using an encoder deep neural network;
generate an icon based on the image vector using a generative adversarial network (GAN);
train at least one of:
the encoder deep neural network to generate the image vector from the image using a set of labeled training images, and
the GAN to generate the icon based on the image vector using a set of labeled training icons; and

output the icon.

9. The system of claim 8 ,
wherein the icon is one selected from a group consisting of a shape, a logo, and an emoji.

10. The system of claim 8 , wherein the icon depicts at least one selected from a group consisting of an object, an action, and an emotion that is present in the image.

11. The system of claim 8 , wherein the image vector establishes a joint image-icon vector space.

12. The system of claim 8 , wherein the icon generation engine is further configured to:
translate the image vector into a caption.

13. The system of claim 8 , wherein the icon generation engine is further configured to, prior to generating the image vector:
preprocess the image, comprising: transforming the image to a format compatible with the encoder deep neural network.

14. The system of claim 8 , further comprising a user interface,
wherein outputting the icon comprises displaying the icon to a user, in the user interface.

15. A non-transitory computer readable medium storing instructions executable by a computer processor, the instructions comprising functionality for:
obtaining an image;
generating an image vector from the image using an encoder deep neural network;
generating an icon based on the image vector using a generative adversarial network (GAN);
training at least one of:
the encoder deep neural network to generate the image vector from the image using a set of labeled training images, and
the GAN to generate the icon based on the image vector using a set of labeled training icons; and

outputting the icon.

16. The non-transitory computer readable medium of claim 15 , wherein the icon is one selected from a group consisting of a shape, a logo, and an emoji.

17. The non-transitory computer readable medium of claim 15 , wherein the icon depicts at least one selected from a group consisting of an object, an action, and an emotion that is present in the image.

18. The non-transitory computer readable medium of claim 15 , wherein the image vector establishes a joint image-icon vector space.

19. The non-transitory computer readable medium of claim 15 , further comprising:
translating the image vector into a caption.

20. The non-transitory computer readable medium of claim 15 , further comprising, prior to generating the image vector:
preprocessing the image, comprising: transforming the image to a format compatible with the encoder deep neural network.

Assignees

• Konica Minolta Business Solutions Usa Inc

Inventors

• Stopa Michael
• Amano Jun

Classifications

• G06N3/08

  Learning methods · CPC title

• G06N3/082

  modifying the architecture, e.g. adding, deleting or silencing nodes or connections · CPC title

• G06N3/048

  Activation functions · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title