IMAGE IDENTIFICATION APPARATUS AND IMAGE IDENTIFICATION METHOD

• US 20170154209A1
• Filed: 11/28/2016
• Published: 06/01/2017
• Est. Priority Date: 12/01/2015
• Status: Active Grant

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Kaneda, Yuji, Nomoto, Atsuo, Nishino, Daisuke, Yamamoto, Takahisa, Nakano, Shunsuke, Sato, Hiroshi
• Granted Patent: US 10,699,102 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 18/2411 based on the proximity to a decision surface, e.g. support vector machines
• CPC Class Codes: G06F 18/2431 Multiple classes
• CPC Class Codes: G06F 18/2451 linear, e.g. hyperplane
• CPC Class Codes: G06V 10/454 Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
• CPC Class Codes: G06V 40/166 using acquisition arrangements
• CPC Class Codes: G06V 40/172 Classification, e.g. identification
• CPC Class Codes: G06V 40/178 estimating age from face image; using age information for improving recognition

Abstract

An identification apparatus includes an acquisition unit configured to acquire input data and an identification unit configured to identify a class to which the acquired input data belongs based on a classifier configured to identify a class to which the data belongs from among a plurality of classes that includes two or more classes having been set beforehand.

Claims

1. An identification apparatus comprising:

an acquisition unit configured to acquire input data; and

an identification unit configured to identify a class to which the acquired input data belongs based on a classifier configured to identify which class the input data belongs to among a plurality of classes that includes two or more classes having been set beforehand.

2. The identification apparatus according to claim 1,

wherein the classifier includes a plurality of classifiers configured to identify whether the data belongs to a first class including the two or more classes having been set beforehand or a second class,

wherein the plurality of classifiers are different from each other in class to be identified as including the data,

wherein the acquisition unit acquires first input data and second input data as the input data, and

wherein the identification unit identifies classes to which the first input data and the second input data acquired by the acquisition unit belong based on the plurality of classifiers,

the identification apparatus further comprising a determination unit configured to determine whether the class of the first input data is equal to or different from the class of the second input data based on the class of the first input data and the class of the second input data identified by the identification unit.

3. The identification apparatus according to claim 2,

wherein each of the plurality of classifiers is configured to identify whether the data belongs to the first class or the second class including all data not belonging to the first class, and

wherein the identification unit identifies the classes to which the first input data and the second input data acquired by the acquisition unit belong based on the plurality of classifiers.

4. The identification apparatus according to claim 2,

wherein the plurality of classifiers are each configured to identify whether the data belongs to the first class or the second class including at least one class having been set beforehand, and

wherein the identification unit identifies the classes to which the first input data and the second input data acquired by the acquisition unit belong based on the plurality of classifiers.

5. The identification apparatus according to claim 2, wherein the identification unit identifies the classes to which the acquired first input data and the second input data belong by using the plurality of classifiers that are linear classifiers.

6. The identification apparatus according to claim 2, wherein the identification unit identifies the classes to which the acquired first input data and the second input data belong by using the plurality of classifiers that include neural networks.

7. The identification apparatus according to claim 2, wherein the identification unit identifies the classes to which the acquired first input data and the second input data belong based on a first classifier included in the plurality of classifiers, and a second classifier included in the plurality of classifiers identifies an identification result obtained by the first classifier.

8. The identification apparatus according to claim 2, wherein the determination unit determines whether the class of the first input data is equal to or different from the class of the second input data based on a comparison between a vector obtained by connecting class identification results of the first input data respectively identified by the plurality of classifiers and a vector obtained by connecting class identification results of the second input data respectively identified by the plurality of classifiers.

9. The identification apparatus according to claim 1,

wherein the acquisition unit acquires image data of a captured human face as the input data, and

wherein the identification unit identifies the class to which the acquired input data belongs based on the classifier configured to identify one of a plurality of classes, to which the data belongs, that includes classes indicating two or more individual persons having been set beforehand.

10. An adjustment apparatus comprising:

an acquisition unit configured to acquire learning data belonging to a class having been set beforehand; and

an adjustment unit configured to identify a class to which the data belongs, from among a plurality of classes that includes two or more classes having been set beforehand, based on the acquired learning data, and configured to adjust a parameter of the classifier in which the class having been set beforehand is included in any of the plurality of classes.

11. An information processing method for an identification apparatus, the method comprising:

acquiring input data; and

identifying a class to which the acquired input data belongs based on a classifier configured to identify which class the acquired input data belongs to from a plurality of classes that includes two or more classes having been set beforehand.

12. An information processing method for an adjustment apparatus, comprising:

acquiring learning data belonging to a class having been set beforehand;

identifying a class to which the data belongs, from among a plurality of classes that includes two or more classes having been set beforehand, based on the acquired learning data; and

adjusting a parameter of the classifier in which the class having been set beforehand is included in the plurality of classes.

13. A non-transitory computer-readable storage medium storing a program that causes a computer to function as:

an acquisition unit configured to acquire input data; and

an identification unit configured to identify a class to which the acquired input data belongs based on a classifier configured to identify a class to which the data belongs from among a plurality of classes that includes two or more classes having been set beforehand.

14. A non-transitory computer-readable storage medium storing a program that causes a computer to function as:

an acquisition unit configured to acquire learning data belonging to a class having been set beforehand; and

an adjustment unit configured to identify a class to which the data belongs, from among a plurality of classes that includes two or more classes having been set beforehand, based on the acquired learning data, and configured to adjust a parameter of the classifier in which the class having been set beforehand is included in the plurality of classes.