Method and system for age estimation based on relative ages of pairwise facial images of people

• US 8,520,906 B1
• Filed: 09/12/2008
• Issued: 08/27/2013
• Est. Priority Date: 09/24/2007
• Status: Active Grant

• Current Assignee: Videomining Corporation
• Original Assignee: Videomining Corporation
• Inventors: Moon, Hankyu, Sharma, Rajeev, Jung, Namsoon
• Primary Examiner(s): Alavi, Amir
• Assistant Examiner(s): CESE, KENNY A
• Application Number:
  US12/283,595
• Time in Patent Office: 1,810 Days
• Field of Search: 382115-118
• US Class Current: 382/118
• CPC Class Codes: G06F 18/2178 based on feedback of a supervisor
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data
• CPC Class Codes: G06V 10/7784 based on feedback from supervisors
• CPC Class Codes: G06V 40/178 estimating age from face image; using age information for improving recognition

Abstract

The present invention is a system and method for estimating the age of people based on their facial images. It addresses the difficulty of annotating the age of a person from facial image by utilizing relative age (such as older than, or younger than) and face-based class similarity (gender, ethnicity or appearance-based cluster) of sampled pair-wise facial images. It involves a unique method for the pair-wise face training and a learning machine (or multiple learning machines) which output the relative age along with the face-based class similarity, of the pairwise facial images. At the testing stage, the given input face image is paired with some number of reference images to be fed to the trained machines. The age of the input face is determined by comparing the estimated relative ages of the pairwise facial images to the ages of reference face images. Because age comparison is more meaningful when the pair belongs to the same demographics category (such as gender and ethnicity) or when the pair has similar appearance, the estimated relative ages are weighted according to the face-based class similarity score between the reference face and the input face.

Claims

1. A method for automatically performing age estimation based on the facial image of people using the notion of pairwise facial image and relative age, comprising the following steps of:

a) generating first pairwise facial images along with reference faces from a database of facial images, and annotating the first pairwise facial images for their relative ages and the reference faces for their absolute ages,

b) determining face-based class similarity for the first pairwise facial images by appearance-based face clusters or demographic categories,

c) training learning machines using the first pairwise facial images so that the learning machines estimate the relative age and face-based class similarity score of any pairwise facial images,

d) constructing second pairwise facial images from an input face and the reference faces,

e) estimating the relative ages of the second pairwise facial images, using the trained learning machines, and

f) estimating the age of the input face using the relative ages of the second pairwise facial images,

wherein each of the learning machines represents a face-based class among pre-determined face-based classes,

wherein first face of each pair in the pairwise facial images belongs to the face-based class, and

wherein the face-based class similarity represents whether two faces in the pairwise facial images belong to one of predetermined face-based classes.

2. The method according to claim 1, wherein the method further comprises a step of combining the estimated relative ages from the second pairwise facial images to estimate the age of the input face,

wherein the estimated relative ages are weighted by the face-based class similarity scores.

3. The method according to claim 1, wherein the method further comprises a step of selecting the reference faces for which the ages are determined with high confidence,

wherein the input face is paired with the reference faces to form the second pairwise facial images.

4. The method according to claim 1, wherein the method further comprises a step of generating class-dependent pairwise facial images,

wherein the class-dependent pairwise facial images comprise subsets of faces in the database of facial images and each subset represents a pre-determined face-based class.

5. The method according to claim 1, wherein the method further comprises a step of pairing the input face with the references faces in each face-based class of the pre-determined face-based classes to form the second pairwise facial images, wherein first face in a pairwise facial image is a reference face from the given face-based class and second face is the input face.

6. The method according to claim 1, wherein the method further comprises a step of estimating the relative age of each of the second pairwise facial images using the learning machines that are class-dependent, wherein the estimated relative age casts vote to an input face age interval vote for each reference face, with a voting weight given by face-based class similarity score.

7. An apparatus for automatically performing age estimation based on the facial image of people using the notion of pairwise facial image and relative age, comprising:

a) means for generating first pairwise facial images along with reference faces from a database of facial images, and annotating the first pairwise facial images for their relative ages and the reference faces for their absolute ages,

b) means for determining face-based class similarity for the first pairwise facial images by appearance-based face clusters or demographic categories,

c) means for training learning machines using the first pairwise facial images so that the learning machines estimate the relative age and face-based class similarity score of any pairwise facial images,

d) means for constructing second pairwise facial images from an input face and the reference faces,

e) means for estimating the relative ages of second pairwise facial images, using the trained learning machines, and

f) means for estimating the age of the input face using the relative ages of the second pairwise facial images,

wherein each of the learning machines represents a face-based class among pre-determined face-based classes,

wherein first face of each pair in the pairwise facial images belongs to the face-based class, and wherein the face-based class similarity represents whether two faces in the pairwise facial images belong to one of predetermined face-based classes.

8. The apparatus according to claim 7, wherein the apparatus further comprises means for combining the estimated relative ages from the second pairwise facial images to estimate the age of the input face,

wherein the estimated relative ages are weighted by the face-based class similarity scores.

9. The apparatus according to claim 7, wherein the apparatus further comprises means for selecting the reference faces for which the ages are determined with high confidence,

wherein the input face is paired with the reference faces to form the second pairwise facial images.

10. The apparatus according to claim 7, wherein the apparatus further comprises means for generating class-dependent pairwise facial images, wherein the class-dependent pairwise facial images comprise subsets of faces in the database of facial images and each subset represents a pre-determined face-based class.

11. The apparatus according to claim 7, wherein the apparatus further comprises means for pairing the input face with the references faces in each face-based class of the pre-determined face-based classes to form the second pairwise facial images, wherein first face in a pairwise facial image is a reference face from the given face-based class and second face is the input face.

12. The apparatus according to claim 7, wherein the apparatus further comprises means for estimating the relative age of each of the second pairwise facial images using the learning machines that are class-dependent, wherein the estimated relative age casts vote to an input face age interval vote for each reference face, with a voting weight given by face-based class similarity score.