Method and device for face recognition, storage medium, and electronic device

• US 11,380,131 B2
• Filed: 12/30/2019
• Issued: 07/05/2022
• Est. Priority Date: 06/30/2017
• Status: Active Grant

• Current Assignee: Guangdong OPPO Mobile Telecommunications Corporation Limited
• Original Assignee: Guangdong OPPO Mobile Telecommunications Corporation Limited
• Inventors: Liang, Kun
• Primary Examiner(s): Shin, Soo
• Application Number:
  US16/730,571
• Publication Number:
  US 20200134294A1
• Time in Patent Office: 918 Days
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06V 40/16 Human faces, e.g. facial parts, sketches or expressions
• CPC Class Codes: G06V 40/168 Feature extraction; Face representation
• CPC Class Codes: G06V 40/171 Local features and components; Facial parts eye characteristics G06V40/18; Occluding parts, e.g. glasses; Geometrical relationships
• CPC Class Codes: G06V 40/178 estimating age from face image; using age information for improving recognition
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

A method and device for face recognition, a storage medium, and an electronic device are provided. The method includes the following. Face data to-be-tested is obtained. A first derived face data set related to the face data to-be-tested is generated according to the face data to-be-tested. For each of multiple derived face data in the first derived face data set, perform age determination, and generate an age distribution interval corresponding to the first derived face data set. Whether the age distribution interval matches a first reference age interval is determined. Upon determining that the age distribution interval matches the first reference age interval, age data corresponding to the face data to-be-tested is obtained according to the age distribution interval.

Claims

1. A method for face recognition, comprising:

obtaining face data to-be-tested;

performing face aging and face rejuvenation on the face data to-be-tested in terms of different preset ages, and generating multiple derived face data, the multiple derived face data representing different ages and being a set of synthetically generated faces obtained from the face data to-be-tested, with faces looking older and younger than a face in the face data to-be-tested;

obtaining a first derived face data set by distributing the multiple derived face data sequentially according to ages corresponding to the multiple derived face data;

for each of the multiple derived face data in the first derived face data set, performing age determination, and generating an age distribution interval corresponding to the first derived face data set;

determining whether the age distribution interval matches a first reference age interval comprising;

determining whether a degree of correlation between the age distribution interval obtained by performing age determination on each of the multiple derived face data in the first derived face data set and the first reference age interval is greater than a preset threshold, wherein the degree of correlation is a ratio of an overlap portion of the two intervals to an entire range of the age distribution interval; and

obtaining, according to the age distribution interval, age data corresponding to the face data to-be-tested, upon determining that the age distribution interval matches the first reference age interval, comprising;

upon determining that the degree of correlation is greater than the preset threshold, obtaining, according to the age distribution interval, the age data corresponding to the face data to-be-tested;

or

upon determining that the degree of correlation is less than the preset threshold, adjusting endpoint values of the age distribution interval, until the age distribution interval matches the first reference age interval.

2. The method of claim 1, wherein

the first derived face data set is generated according to the face data to-be-tested based on a predetermined face data generating model;

and

the method further comprises;

prior to performing the face aging and the face rejuvenation on the face data to-be-tested, generating the multiple derived face data, and obtaining the first derived face data set;

obtaining reference face data;

generating, according to the reference face data, a second derived face data set related to the reference face data, wherein the second derived face data set comprises multiple different derived face data;

for each of the multiple different derived face data in the second derived face data set, performing age determination, and generating an age distribution interval corresponding to the second derived face data set;

determining whether the age distribution interval corresponding to the second derived face data set matches a second reference age interval; and

upon determining that the age distribution interval corresponding to the second derived face data set does not match the second reference age interval, updating model parameters of the predetermined face data generating model, until the age distribution interval corresponding to the second derived face data set and obtained by performing age determination on each of the multiple different derived face data in the second derived face data set matches the second reference age interval.

3. The method of claim 1, wherein for each of the multiple derived face data in the first derived face data set, performing the age determination, and generating the age distribution interval corresponding to the first derived face data set comprise:

for each of the multiple derived face data in the first derived face data set, determining an age corresponding thereto based on a predetermined age determining model, and generating the age distribution interval corresponding to the first derived face data set;

wherein

the predetermined age determining model is configured to extract feature data from the derived face data and determine age data corresponding to the feature data.

4. The method of claim 3, wherein the predetermined age determining model comprises a plurality of feature-data determining models, and each of the plurality of feature-data determining models is configured to determine feature data of a preset type and obtain, according to the feature data of the preset type, age data corresponding to the feature data of the preset type.

5. The method of claim 3, further comprising:

prior to, for each of the multiple derived face data in the first derived face data set, determining the age corresponding thereto based on the predetermined age determining model, and generating the age distribution interval corresponding to the first derived face data set;

obtaining reference face data, wherein the reference face data comprises age data;

extracting, according to the reference face data, feature data corresponding to the reference face data; and

generating the predetermined age determining model by associating the feature data with the age data corresponding to the feature data.

6. The method of claim 3, wherein the feature data is obtained from face data by deep learning and a convolutional neural network.

7. The method of claim 1, wherein obtaining, according to the age distribution interval, the age data corresponding to the face data to-be-tested comprises:

obtaining, according to the age distribution interval, age median data corresponding to the age distribution interval; and

using the age median data as the age data corresponding to the face data to-be-tested.

8. The method of claim 1, wherein for each of the multiple derived face data in the first derived face data set, performing age determination, and generating the age distribution interval corresponding to the first derived face data set comprise:

for each of the multiple derived face data in the first derived face data set,

extracting a plurality of feature data from the derived face data based on a predetermined age determining model, wherein the predetermined age determining model is generated by associating a plurality of feature-data determining models according to preset weights; and

for each of the plurality of feature data, performing age determination on the feature data based on the predetermined age determining model, to obtain ages corresponding to the plurality of feature data, and determining an age corresponding to the derived face data by aggregating the ages corresponding to the plurality of feature data; and

the method further comprises;

generating, according to a plurality of ages determined, the age distribution interval corresponding to the first derived face data set.

9. The method of claim 1, wherein the first reference age interval is obtained by performing age determination on the face data to-be-tested.

10. A non-transitory computer readable storage medium configured to store instructions which, when executed by a processor, cause the processor to carry out actions, comprising:

obtaining face data to-be-tested;

performing face aging and face rejuvenation on the face data to-be-tested in terms of different preset ages, and generating multiple derived face data, the multiple derived face data representing different ages and being a set of synthetically generated faces obtained from the face data to-be-tested, with faces looking older and younger than a face in the face data-to-be tested;

obtaining a first derived face data set by distributing the multiple derived face data sequentially according to ages corresponding to the multiple derived face data;

for each of the multiple derived face data in the first derived face data set, performing age determination, and generating an age distribution interval corresponding to the first derived face data set;

determining whether the age distribution interval matches a first reference age interval comprising;

determining whether a degree of correlation between the age distribution interval obtained by performing age determination on each of the multiple derived face data in the first derived face data set and the first reference age interval is greater than a preset threshold, wherein the degree of correlation is a ratio of an overlap portion of the two intervals to an entire range of the age distribution interval; and

obtaining, according to the age distribution interval, age data corresponding to the face data to-be-tested, upon determining that the age distribution interval matches the first reference age interval, comprising;

upon determining that the degree of correlation is greater than the preset threshold obtaining, according to the age distribution interval, the age data corresponding to the face data to-be-tested;

or

upon determining that the degree of correlation is less than the preset threshold, adjusting endpoint values of the age distribution interval until the age distribution interval matches the first reference age interval.

11. The non-transitory computer readable storage medium of claim 10, wherein the instructions are further executed by the processor to carry out actions, comprising:

obtaining reference face data, wherein the first derived face data set is generated according to the face data to-be-tested based on a predetermined face data generating model;

generating, according to the reference face data, a second derived face data set related to the reference face data, wherein the second derived face data set comprises multiple different derived face data;

for each of the multiple different derived face data in the second derived face data set, performing age determination, and generating an age distribution interval corresponding to the second derived face data set;

determining whether the age distribution interval corresponding to the second derived face data set matches a second reference age interval; and

upon determining that the age distribution interval corresponding to the second derived face data set does not match the second reference age interval, updating model parameters of the predetermined face data generating model, until the age distribution interval corresponding to the second derived face data set and obtained by performing age determination on each of the multiple different derived face data in the second derived face data set matches the second reference age interval.

12. The non-transitory computer readable storage medium of claim 10, wherein the instructions executed by the processor to carry out the action of performing, for each of the multiple derived face data in the first derived face data set, age determination, and generating the age distribution interval corresponding to the first derived face data set are executed by the processor to carry out actions, comprising:

for each of the multiple derived face data in the first derived face data set, determining an age corresponding thereto based on a predetermined age determining model, and generating the age distribution interval corresponding to the first derived face data set;

wherein

the predetermined age determining model is configured to extract feature data from the derived face data and determine age data corresponding to the feature data.

13. The non-transitory computer readable storage medium of claim 12, wherein the predetermined age determining model comprises a plurality of feature-data determining models, and each of the plurality of feature-data determining models is configured to determine feature data of a preset type and obtain, according to the feature data of the preset type, age data corresponding to the feature data of the preset type.

14. The non-transitory computer readable storage medium of claim 10, wherein the instructions executed by the processor to carry out the action of obtaining, according to the age distribution interval, the age data corresponding to the face data to-be-tested are executed by the processor to carry out actions, comprising:

obtaining, according to the age distribution interval, age median data corresponding to the age distribution interval; and

using the age median data as the age data corresponding to the face data to-be-tested.

15. An electronic device, comprising:

at least one processor; and

a computer readable storage, coupled to the at least one processor and storing at least one computer executable instruction thereon which, when executed by the at least one processor, is operable with the at least one processor to;

obtain face data to-be-tested;

perform face aging and face rejuvenation on the face data to-be-tested in terms of different preset ages, and generate multiple derived face data, the multiple derived face data representing different ages and being a set of synthetically generated faces obtained from the face data to-be-tested, with faces looking older and younger than a face in the data to-be-tested;

obtain a first derived face data set by distributing the multiple derived face data sequentially according to ages corresponding to the multiple derived face data;

for each of the multiple derived face data in the first derived face data set, perform age determination, and generate an age distribution interval corresponding to the first derived face data set;

determine whether the age distribution interval matches a first reference age interval, comprising;

determine whether a degree of correlation between the age distribution interval obtained by performing age determination on each of the multiple derived face data in the first derived face data set and the first reference age interval is greater than a preset threshold, wherein the degree of correlation is a ratio of an overlap portion of the two intervals to an entire range of the age distribution interval; and

obtain, according to the age distribution interval, age data corresponding to the face data to-be-tested, upon determining that the age distribution interval matches the first reference age interval, comprising;

upon determining that the degree of correlation is greater than the preset threshold, obtain, according to the age distribution interval, the age data corresponding to the face data to-be-tested;

or

upon determining that the degree of correlation is less than the preset threshold, adjust endpoint values of the age distribution interval, until the age distribution interval matches the first reference age interval.

16. The electronic device of claim 15, wherein the at least one computer executable instruction is further operable with the at least one processor to:

obtain reference face data, wherein the first derived face data set is generated according to the face data to-be-tested based on a predetermined face data generating model;

generate, according to the reference face data, a second derived face data set related to the reference face data, wherein the second derived face data set comprises multiple different derived face data;

for each of the multiple different derived face data in the second derived face data set, perform age determination, and generate an age distribution interval corresponding to the second derived face data set;

determine whether the age distribution interval corresponding to the second derived face data set matches a second reference age interval; and

upon determining that the age distribution interval corresponding to the second derived face data set does not match the second reference age interval, update model parameters of the predetermined face data generating model, until the age distribution interval corresponding to the second derived face data set and obtained by performing age determination on each of the multiple different derived face data in the second derived face data set matches the second reference age interval.

17. The electronic device of claim 15, wherein the at least one computer executable instruction operable with the at least one processor to perform, for each of the multiple derived face data in the first derived face data set, age determination, and generate the age distribution interval corresponding to the first derived face data set is operable with the at least one processor to:

for each of the multiple derived face data in the first derived face data set, determine an age corresponding thereto based on a predetermined age determining model, and generate the age distribution interval corresponding to the first derived face data set;

wherein

the predetermined age determining model is configured to extract feature data from the derived face data and determine age data corresponding to the feature data.

18. The electronic device of claim 17, wherein the at least one computer executable instruction is further operable with the at least one processor to:

obtain reference face data, wherein the reference face data comprises age data;

extract, according to the reference face data, feature data corresponding to the reference face data; and

generate the predetermined age determining model by associating the feature data with the age data corresponding to the feature data.

19. The electronic device of claim 17, wherein the predetermined age determining model comprises a plurality of feature-data determining models, and each of the plurality of feature-data determining models is configured to determine feature data of a preset type and obtain, according to the feature data of the preset type, age data corresponding to the feature data of the preset type.

20. The electronic device of claim 15, wherein the at least one computer executable instruction operable with the at least one processor to obtain, according to the age distribution interval, the age data corresponding to the face data to-be-tested is operable with the at least one processor to:

obtain, according to the age distribution interval, age median data corresponding to the age distribution interval; and

use the age median data as the age data corresponding to the face data to-be-tested.