Automatically Mining Person Models of Celebrities for Visual Search Applications

• US 20110116690A1
• Filed: 08/19/2010
• Published: 05/19/2011
• Est. Priority Date: 11/18/2009
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: RABINOVICH, Andrew, PILLAI, Anand, ROSS, David, ADAM, Hartwig
• Granted Patent: US 8,605,956 B2
• Time in Patent Office: Days
• US Class Current: 382/118
• CPC Class Codes: G06V 10/763 Non-hierarchical techniques, e.g. based on statistics of modelling distributions
• CPC Class Codes: G06V 40/173 face re-identification, e.g. recognising unknown faces across different face tracks

Abstract

Methods and systems for automated identification of celebrity face images are provided that generate a name list of prominent celebrities, obtain a set of images and corresponding feature vectors for each name, detect faces within the set of images, and remove non-face images. An analysis of the images is performed using an intra-model analysis, an inter-model analysis, and a spectral analysis to return highly accurate biometric models for each of the individuals present in the name list. Recognition is then performed based on precision and recall to identify the face images as belonging to a celebrity or indicate that the face is unknown.

Claims

1. A computer-implemented method of automatic face recognition, comprising:

(a) generating one or more names based on one or more articles;

(b) obtaining one or more images purporting to correspond to the one or more names;

(c) selecting one or more face images from the one or more images;

(d) associating the one or more face images with the one or more names; and

(e) removing incorrectly associated face images using intra-model, inter-model, and spectral analysis.

2. The computer-implemented method of claim 1, further comprising determining a best matching name with a face image.

3. The computer-implemented method of claim 1, further comprising determining a representative image of a person.

4. The computer-implemented method of claim 1, wherein the one or more articles are filtered to retain only articles that contain names of people.

5. The computer-implemented method of claim 1, wherein the spectral analysis is based upon iterative binary clustering.

6. The computer-implemented method of claim 1, further comprising failing to determine a best matching name with a face image when a recognition likelihood threshold is not exceeded.

7. The computer-implemented method of claim 1, wherein the spectral analysis is performed after the inter-model analysis and the inter-model analysis is performed after the intra-model analysis.

8. The computer-implemented method of claim 1, further comprising ranking a name based on a quantity of associated face images.

9. The computer-implemented method of claim 1, further comprising detecting a feature vector for the one or more face images.

10. The computer-implemented method of claim 8, wherein the detecting identifies a facial feature location within the one or more face images.

11. A system, comprising:

(a) a face image database;

(b) a name database; and

(c) a computer-based face recognition system, comprising;

(i) a name list generator configured to generate one or more names in a name list based on an article and to retrieve one or more images associated with the one or more names;

(ii) a face signature detector configured to detect and associate face images within the one or more images corresponding with the one or more names in the name list;

(iii) an intra-model analyzer configured to remove incorrectly associated face images based on face images associated with a single name in the name list;

(iv) an inter-model analyzer configured to remove incorrectly associated face images based on face images associated with a different name in the name list; and

(v) a spectral analyzer configured to remove incorrectly associated face images based on a similarity matrix.

12. The system of claim 11, further comprising a recognizer configured to determine whether a particular image is associated with a particular name in the name list.

13. The system of claim 11, wherein the name list generator further comprises a name ranker configured to rank the one or more names in the name list based on a quantity of associated face images.

14. The system of claim 11, wherein the face signature detector comprises a feature detector to detect face images based on Gabor wavelets.

15. The system of claim 11, wherein the face signature detector comprises a feature detector to detect face images based on a facial feature location within the one or more face images.

16. The system of claim 12, wherein the recognizer determines that there is no matching name associated with a face image.

17. The system of claim 11, wherein the intra-model analyzer uses an intra-person comparison based on the associated face images of all the names in the name list.

18. The system of claim 11, wherein the inter-model analyzer is uses an inter-person comparison based on a recursive similarity comparison.

19. A computer program product, comprising a non-transitory computer readable storage medium, the non-transitory computer readable storage medium having embodied thereon computer readable program code to realize an automated face recognition matching, the computer control logic comprising:

first computer readable program code for causing the computer to generate one or more names in a name list based on an article and to retrieve one or more images associated with the one or more names;

second computer readable program code for causing the computer to detect and associate face images within the one or more images corresponding with the one or more names in the name list;

third computer readable program code for causing the computer to remove incorrectly associated face images based on face images corresponding with a single name in the name list;

fourth computer readable program code for causing the computer to perform an inter-model analysis to remove incorrectly associated face images based on face images associated with a different name in the name list;

fifth computer readable program code for causing the computer to perform a spectral analysis to remove incorrectly associated face images based on a distance matrix; and

sixth computer readable program code for causing the computer to determine whether a particular image is associated with a particular name in the name list.

20. The computer program product of claim 19, further comprising further a seventh computer readable program code for causing the computer to determine that there is no matching name associated with a face image.

21. A device for communicating with an automatic face recognition system that determines whether a particular image is associated with a particular name in a name list using intra-model, inter-model, and spectral analysis, comprising:

a client-based interface configured to input one or more names to the name list of the automatic face recognition system; and

a client-based interface configured to receive a particular image that is associated with a particular name in the name list from the automatic face recognition system.

22. A method of automatic face recognition including generating one or more names based on one or more articles, obtaining one or more images purporting to correspond to the one or more names, selecting one or more face images from the one or more images, associating the one or more face images with the one or more names, and removing incorrectly associated face images using intra-model, inter-model, and spectral analysis, determining a best matching face image with a particular name, comprising:

inputting one or more names; and

receiving the best matching face image with the particular name.