Systems and methods for identifying objects in media contents

Patent metadata

Field	Value
Publication number	US-10013621-B2
Application number	US-201615285377-A
Country	US
Kind code	B2
Filing date	Oct 4, 2016
Priority date	Oct 4, 2016
Publication date	Jul 3, 2018
Grant date	Jul 3, 2018

Abstract

There is provided a system including a memory and a processor to receive images. The processor is further to train independent detectors for identifying first objects in the images based on individual attributes including a first attribute and a second attribute. The processor is also to train joint detectors for identifying first objects in the images based on composite attributes including a composite attributes each including the first attribute and the second attribute. The processor is to analyze features of the images to determine a difference between a first training performance of the independent detectors and a second training performance of the joint detectors. Lastly, the processor is to select, based on the analyzing, between using the independent detectors and using the joint detectors for identifying second objects in the images using a third attribute and a fourth attribute in the attribute database.

First claim

What is claimed is:

1. A system comprising:
a non-transitory memory storing an attribute database; and
a hardware processor executing an executable code to:
receive a plurality of images;
train, using the plurality of images, a plurality of independent detectors for identifying a first plurality of objects in the plurality of images based on a first set of individual attributes in the attribute database including a first attribute and a second attribute;
train, using the plurality of images, a plurality of joint detectors for identifying the first plurality of objects in the plurality of images based on a first set of composite attributes in the attribute database including a plurality of composite attributes, each of the plurality of composite attributes including the first attribute and the second attribute;
analyze a set of features of the plurality of images to determine a difference between a first training performance of the plurality of independent detectors and a second training performance of the plurality of joint detectors; and
select, based on the analyzing, between using the plurality of independent detectors and using the plurality of joint detectors for identifying a second plurality of objects in the plurality of images using at least a first new attribute and a second new attribute in the attribute database.

2. The system of claim 1 , wherein the first attribute is one of a first noun, a first adjective, a first preposition, and a first verb.

3. The system of claim 2 , wherein the second attribute is one of a second noun, a second adjective, a second preposition, and a second verb.

4. The system of claim 3 , wherein the first set of individual attributes and the first set of composite attributes further include a third attribute.

5. The system of claim 4 , wherein the third attribute is one of a third noun and a third adjective.

6. The system of claim 1 , wherein the hardware processor further executes the executable code to:
identify the second plurality of objects in the plurality of images based on the selection.

7. The system of claim 1 , wherein the plurality of independent detectors and the plurality of joint detectors are trained on less than 50% of the attributes in the attribute database.

8. The system of claim 1 , wherein the set of features includes at least one of a number of training images, a separability of the first attribute and the second attribute, and an entropy of the plurality of images.

9. The system of claim 8 , wherein each feature in the set of features is weighted.

10. The system of claim 1 , wherein the first attribute is a scene attribute.

11. A method for use with a system including a non-transitory memory and a hardware processor, the method comprising:
receiving, using the hardware processor, a plurality of images;
training, using the hardware processor, using the plurality of images, a plurality of independent detectors for identifying a first plurality of objects in the plurality of images based on a first set of individual attributes in the attribute database including a first attribute and a second attribute;
training, using the hardware processor, using the plurality of images, a plurality of joint detectors for identifying the first plurality of objects in the plurality of images based on a first set of composite attributes in the attribute database including a plurality of composite attributes, each of the plurality of composite attributes including the first attribute and the second attribute;
analyzing, using the hardware processor, a set of features of the plurality of images to determine a difference between a first training performance of the plurality of independent detectors and a second training performance of the plurality of joint detectors;
selecting, using the hardware processor and based on the analyzing, between using the plurality of independent detectors and using the plurality of joint detectors for identifying a second plurality of objects in the plurality of images using at least a first new attribute and a second new attribute in the attribute database.

12. The method of claim 11 , wherein the first attribute is one of a first noun, a first adjective, a first preposition, and a first verb.

13. The method of claim 12 , wherein the second attribute is one of a second noun, a second adjective, a second preposition, and a second verb.

14. The method of claim 13 , wherein the first set of individual attributes and the first set of composite attributes further include a third attribute.

15. The method of claim 14 , wherein the third attribute is one of a third noun and a third adjective.

16. The method of claim 11 , further comprising:
identifying, using the hardware processor, the second plurality of objects in the plurality of images based on the selection.

17. The method of claim 11 , wherein the plurality of independent detectors and the plurality of joint detectors are trained on less than 50% of the attributes in the attribute database.

18. The method of claim 11 , wherein the set of features includes at least one of a number of training images, a separability of the first attribute and the second attribute, and an entropy of the plurality of images.

19. The method of claim 18 , wherein each feature in the set of features is weighted.

20. The method of claim 11 , wherein the first attribute is a scene attribute.

Assignees

• Disney Entpr Inc

Inventors

• Sigal Leonid
• Shamir Arik
• Rosenthal Guy

Classifications

• G06V10/82 Primary

  using neural networks · CPC title

• G06V10/955

  using specific electronic processors · CPC title

• G06V10/454

  Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN] · CPC title

• G06V10/75

  Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries · CPC title

• G06K9/00986 Primary

  Physics · mapped topic