APPARATUS AND METHOD FOR CLASSIFYING IMAGE

• US 20200134386A1
• Filed: 10/24/2019
• Published: 04/30/2020
• Est. Priority Date: 10/24/2018
• Status: Active Grant

• Current Assignee: Samsung SDS (China) Co., Ltd. (Samsung Group)
• Original Assignee: Samsung SDS (China) Co., Ltd. (Samsung Group)
• Inventors: Lee, JoonHo, Song, HyoSeob, Song, JiEun
• Granted Patent: US 11,468,267 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 16/51 Indexing; Data structures therefor; Storage structures
• CPC Class Codes: G06F 18/2113 by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
• CPC Class Codes: G06F 18/217 Validation; Performance evaluation; Active pattern learning techniques
• CPC Class Codes: G06F 18/2431 Multiple classes
• CPC Class Codes: G06F 18/285 Selection of pattern recognition techniques, e.g. of classifiers in a multi-classifier system
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 20/20 Ensemble learning
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 5/01 Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: G06V 10/40 Extraction of image or video features
• CPC Class Codes: G06V 10/764 using classification, e.g. of video objects
• CPC Class Codes: G06V 10/7747 Organisation of the process, e.g. bagging or boosting
• CPC Class Codes: G06V 10/776 Validation; Performance evaluation
• CPC Class Codes: G06V 10/87 using selection of the recognition techniques, e.g. of a classifier in a multiple classifier system

Abstract

An image classification apparatus includes an image segmentation module configured to segment a learning image into a plurality of segment images, a primary classification module configured to perform machine learning on a primary classifier using the plurality of segment images, and a secondary classification module configured to calculate a weight value combination for creating a secondary classification estimation value for the learning image from a plurality of primary classification estimation values generated by passing the plurality of segment images to the trained primary classifier, or a machine learning-based learning parameter.

Claims

1. An image classification apparatus comprising:

an image segmentation module configured to segment a learning image into a plurality of segment images;

a primary classification module configured to perform machine learning on a primary classifier using the plurality of segment images; and

a secondary classification module configured to calculate a weight value combination for creating a secondary classification estimation value for the learning image from a plurality of primary classification estimation values generated by passing the plurality of segment images to the trained primary classifier, or a machine learning-based learning parameter.

2. The apparatus according to claim 1, wherein the primary classification module is further configured to perform machine learning on the primary classifier by individually inputting the plurality of segment images into the primary classifier.

3. The apparatus according to claim 2, wherein a classification value of each of the plurality of segment images inputted into the primary classifier is set to be equal to a classification value of the learning image.

4. The apparatus according to claim 1, wherein the primary classification estimation value is any one among a regression value, a score value and a class value of each classification target class.

5. The apparatus according to claim 1, wherein the secondary classification module is further configured to receive an initial weight value and a weight value search range for each of the plurality of primary classification estimation values, and search for an optimum weight value combination satisfying a preset condition while changing the initial weight value within the inputted weight value search range.

6. The apparatus according to claim 5, wherein if a loss is smaller than a preset threshold value when a specific weight value combination is applied to the plurality of primary classification estimation values, the secondary classification module is further configured to determine the specific weight value combination as the optimum weight value combination.

7. The apparatus according to claim 1, wherein the secondary classification module is further configured to perform machine learning on a secondary classifier by simultaneously inputting the plurality of primary classification estimation values into the secondary classifier.

8. The apparatus according to claim 7, wherein the secondary classification module is further configured to update a learning parameter of the secondary classifier so that a loss of the secondary classification estimation value generated by the trained secondary classifier may be smaller than a preset threshold value.

9. An image classification method performed in a computing device including one or more processors and a memory for storing one or more programs executed by the one or more processors, the method comprising:

a step of segmenting a learning image into a plurality of segment images;

a primary learning step of performing machine learning on a primary classifier using the plurality of segment images; and

a secondary learning step of calculating a weight value combination for creating a secondary classification estimation value for the learning image from a plurality of primary classification estimation values generated by passing the plurality of segment images to the trained primary classifier, or a machine learning-based learning parameter.

10. The method according to claim 9, wherein the primary learning step comprises:

performing machine learning on the primary classifier by individually inputting the plurality of segment images into the primary classifier.

11. The method according to claim 10, wherein a classification value of each of the plurality of segment images inputted into the primary classifier is set to be equal to a classification value of the learning image.

12. The method according to claim 9, wherein the primary classification estimation value is any one among a regression value, a score value and a class value of each classification target class.

13. The method according to claim 9, wherein the secondary learning step comprises:

receiving an initial weight value and a weight value search range for each of the plurality of primary classification estimation values; and

searching for an optimum weight value combination satisfying a preset condition while changing the initial weight value within the inputted weight value search range.

14. The method according to claim 13, wherein if a loss is smaller than a preset threshold value when a specific weight value combination is applied to the plurality of primary classification estimation values, the secondary learning step further comprises:

determining the specific weight value combination as the optimum weight value combination.

15. The method according to claim 9, wherein the secondary learning step comprises:

performing machine learning on a secondary classifier by simultaneously inputting the plurality of primary classification estimation values into the secondary classifier.

16. The method according to claim 15, wherein the secondary learning step is further comprises:

updating a learning parameter of the secondary classifier so that a loss of the secondary classification estimation value generated by the trained secondary classifier may be smaller than a preset threshold value.