Inferring the gender of a face in an image

• US 8,041,082 B1
• Filed: 11/02/2007
• Issued: 10/18/2011
• Est. Priority Date: 11/02/2007
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Jing, Yushi, Baluja, Shumeet
• Primary Examiner(s): CHAWAN, SHEELA C
• Application Number:
  US11/934,547
• Time in Patent Office: 1,446 Days
• Field of Search: 382/100, 382/115, 382/118, 382/155, 382/159, 382/181, 382/190, 382/276, 382/224, 382/103, 382/192, 382/274, 382/305, 705/14.4, 705/319, 706/47, 704/232, 704/E15.042, 704/E15.017, 704/231, 345/426, 345/589, 345/420, 358/1.17, 358/1.18, 358/538
• US Class Current: 382/118
• CPC Class Codes: G06Q 50/01 Social networking
• CPC Class Codes: G06V 40/16 Human faces, e.g. facial parts, sketches or expressions
• CPC Class Codes: G06V 40/172 Classification, e.g. identification

Abstract

The subject matter of this specification can be embodied in, among other things, a computer-implemented method that includes receiving a plurality of images having human faces. The method further includes generating a data structure having representations of the faces and associations that link the representations based on similarities in appearance between the faces. The method further includes outputting a first gender value for a first representation of a first face that indicates a gender of the first face based on one or more other gender values of one or more other representations of one or more other faces that are linked to the first representation.

Claims

1. A computer-implemented method comprising:

receiving a plurality of images having human faces;

generating a data structure having representations of the faces and associations that link the representations based on similarities in appearance between the faces; and

outputting a first gender value for a first representation of a first face that indicates a gender of the first face based on one or more other gender values of one or more other representations of one or more other faces that are linked to the first representation,

wherein the first gender value is determined by comparing an area of an image having the first face with areas of images having the one or more other faces, and selecting a gender value that matches known gender values for areas that are more similar to the area of the image having the first face, in opposition to known gender values for areas that are less similar to the area of the image having the first face.

2. The method of claim 1, further comprising comparing the first face and a second face having a second gender value to determine a similarity score between the first face and the second face.

3. The method of claim 2, wherein outputting the first gender value based on the one or more other gender values comprises weighting a contribution of the second gender value based on the similarity score.

4. The method of claim 3, further comprising discarding the contribution of the second gender value if the similarity score is lower than a threshold similarity score.

5. The method of claim 3, further comprising discarding the contribution of the second gender value if the similarity score is lower than a predetermined number or percentage of other similarity scores.

6. The method of claim 1, wherein outputting the first gender value based on the one or more other gender values comprises performing a random walk algorithm on the data structure to determine the first gender value.

7. The method of claim 1, wherein outputting the first gender value based on the one or more other gender values comprises performing an adsorption algorithm on the data structure to determine the first gender value.

8. The method of claim 7, wherein outputting the first gender value occurs at an iteration of the algorithm selected from the group consisting of an iteration where the first gender value converges, an iteration where the first gender value changes by less than a threshold amount over a previous iteration, and an iteration occurring after a predetermined number of iterations.

9. The method of claim 1, further comprising seeding a subset of the one or more other gender values with values indicative of male or female gender.

10. The method of claim 9, wherein the seeding is based on user input.

11. The method of claim 9, wherein the seeding is based on a computer algorithm.

12. The method of claim 1, wherein each of the one or more other faces is associated with both a male gender value and a female gender value.

13. The method of claim 12, further comprising determining a final gender value for each of the one or more other faces based on whether the male gender value or the female gender value is greater.

14. The method of claim 12, further comprising normalizing the male gender values and the female gender values associated with the one or more other faces.

15. The method of claim 14, wherein the normalizing is based on a contribution of the male and female gender values relative to each other for each of the one or more other faces or the normalizing is based on a contribution of a gender value for a selected gender relative to other gender values associated with the one or more other faces for the selected gender.

16. The method of claim 1, further comprising generating a confidence value for the first gender value based on a distribution of the one or more other gender values.

17. The method of claim 16, further comprising weighting the first gender value based on the confidence value.

18. The method of claim 1, further comprising identifying the faces within the images using a facial recognition algorithm.

19. A computer-implemented method comprising:

receiving a plurality of images having animal faces;

generating a data structure that associates a first face with one or more second faces based on similarities in appearance between the first face and the one or more second faces; and

outputting a first gender value for the first face based on one or more gender values previously associated with the one or more second faces,

wherein the first gender value is determined by comparing an area of an image having the first face with areas of images having the one or more second faces, and selecting a gender value that matches known gender values for areas that are more similar to the area of the image having the first face, in opposition to known gender values for areas that are less similar to the area of the image having the first face.

20. The method of claim 19, wherein outputting the first gender value based on the one or more other gender values comprises performing a random walk algorithm on the data structure to determine the first gender value.

21. The method of claim 19, further comprising seeding a subset of the one or more second gender values with values indicative of male or female gender, using manual user input or a computer algorithm.

22. The method of claim 19, further comprising generating a confidence value for the first gender value based on a distribution of the one or more second gender values.

23. The method of claim 22, further comprising weighting the first gender value based on the confidence value.

24. The method of claim 19, further comprising identifying the faces within the images using a facial recognition algorithm.

25. A system comprising:

a face detector for identifying faces within images;

means for generating a data structure configured to associate a first face with one or more second faces based on similarities in appearance between the first face and the one or more second faces; and

an interface to output a first gender value for the first face based on one or more gender values previously associated with the one or more second faces,

wherein the first gender value is determined by comparing an area of an image having the first face with areas of images having the one or more second faces, and selecting a gender value that matches known gender values for areas that are more similar to the area of the image having the first face, in opposition to known gender values for areas that are less similar to the area of the image having the first face.

26. The system of claim 25, wherein outputting the first gender value based on the one or more other gender values comprises performing a random walk algorithm on the data structure to determine the first gender value.

27. The system of claim 25, further comprising memory storing instructions that, when performed on one or more processors of the system, seed a subset of the one or more second gender values with values indicative of male or female gender, using manual user input or a computer algorithm.

28. The system of claim 25, further comprising memory storing instructions that, when performed on one or more processors of the system, generate a confidence value for the first gender value based on a distribution of the one or more second gender values.

29. The system of claim 28, further comprising memory storing instructions that, when performed on one or more processors of the system, weight the first gender value based on the confidence value.

30. The system of claim 25, further comprising memory storing instructions that, when performed on one or more processors of the system, identify the faces within the images using a facial recognition algorithm.