Identification of item attributes using artificial intelligence

Patent metadata

Field	Value
Publication number	US-9798960-B2
Application number	US-201715414433-A
Country	US
Kind code	B2
Filing date	Jan 24, 2017
Priority date	Dec 17, 2014
Publication date	Oct 24, 2017
Grant date	Oct 24, 2017

Abstract

A system that identifies attributes of an item depicted in an image using artificial intelligence is provided. For example, the system may use one or more deep belief networks (DBNs) or convolution neural networks (CNNs) trained to analyze images and identify attributes in items depicted in the images. A first artificial intelligence module may analyze an image to determine a type of item depicted in the image. The system may then select a second artificial intelligence module that is associated with the type of item and use the second artificial intelligence module to identify attributes in the item depicted in the image. Identified attributes, if associated with a confidence level over a threshold value, may be provided to a user. The user may provide feedback on the accuracy of the identified attributes, which can be used to further train the first and/or second artificial intelligence modules.

First claim

What is claimed is:

1. A system for analyzing an image in real-time using artificial intelligence comprising:
a catalog database that stores images and attributes; and
a server system configured to:
receive an image and attributes associated with the image that are stored in the catalog database;
determine a plurality of attributes depicted in the image using an artificial intelligence system, wherein the artificial intelligence system is trained using a plurality of images, and wherein individual attributes of the plurality of attributes are associated with a confidence level;
identify a first subset of attributes in the plurality of attributes, wherein the first subset comprises attributes in the plurality that are associated with a confidence level greater than a threshold value; and
compare the attributes associated with the image with the first subset of attributes to determine whether an item description in the catalog database accurately describes an item depicted in the image.

2. The system of claim 1 , wherein the server system is further configured to:
determine a second plurality of attributes depicted in the image using a second artificial intelligence system, wherein attributes of the second plurality are associated with a second confidence level;
identify a third subset of attributes in the second plurality of attributes, wherein the third subset comprises attributes in the second plurality that are associated with a second confidence level greater than the threshold value; and
update the catalog database to include at least one attribute in the third subset of attributes that is not included in attributes associated with the image.

3. The system of claim 1 , wherein the catalog database comprises signatures associated with the images, and wherein the server system is further configured to:
generate a signature using the image;
compare the generated signature with signatures in the catalog database; and
identify a type of item depicted in the image using a second artificial intelligence system in response to a determination that no signature in the catalog database matches the generated signature.

4. The system of claim 3 , wherein the server system is further configured to generate the signature by calculating a hash value of the image.

5. The system of claim 1 , wherein the server system is further configured to update the catalog database to include at least one attribute in the first subset of attributes that is not included in the attributes associated with the image.

6. The system of claim 1 , wherein the artificial intelligence system comprises a network of hierarchical artificial intelligence systems, and wherein an artificial intelligence system in the network of hierarchical artificial intelligence systems identifies at least one attribute in the determined plurality of attributes.

7. A computer-implemented method for analyzing an image in real-time using artificial intelligence, the method comprising:
as implemented by a computer system comprising one or more computing devices, the computer system configured with specific executable instructions,
receiving an image and attributes associated with the image that are stored in a catalog database;
determining a plurality of attributes depicted in the image using an artificial intelligence system, wherein the artificial intelligence system is trained using a plurality of images, and wherein individual attributes of the plurality of attributes are associated with a confidence level;
identifying a first subset of attributes in the plurality of attributes, wherein the first subset comprises attributes in the plurality that are associated with a confidence level greater than a threshold value; and
comparing the attributes associated with the image with the first subset of attributes to determine whether an item description in the catalog database accurately describes an item depicted in the image.

8. The computer-implemented method of claim 7 , further comprising updating the catalog database to include at least one attribute in the first subset of attributes that is not included in the attributes associated with the image.

9. The computer-implemented method of claim 7 , further comprising transmitting, to a user device, an identification of at least one attributes in the first subset of attributes that is not included in the attributes associated with the image.

10. The computer-implemented method of claim 7 , further comprising transmitting, to a user device, an identification of an attribute in the first subset of attributes that conflicts with the attributes associated with the image.

11. The computer-implemented method of claim 7 , further comprising:
receiving a confirmation from a user device indicating whether an attribute in the first subset of attributes transmitted to the user device is associated with the image; and
updating the catalog database to include an attribute in the first subset of attributes for which a confirmation is received that the attribute is associated with the image.

12. The computer-implemented method of claim 7 , further comprising:
determining a second plurality of attributes depicted in the image using a second artificial intelligence system, wherein attributes of the second plurality are associated with a second confidence level;
identifying a third subset of attributes in the second plurality of attributes, wherein the third subset comprises attributes in the second plurality that are associated with a second confidence level greater than the threshold value; and
updating the catalog database to include at least one attribute in the third subset of attributes that is not included in attributes associated with the image.

13. The computer-implemented method of claim 7 , further comprising:
generating a signature based on the first subset of attributes, wherein the image is associated with a first listing;
comparing the generated signature with signatures stored in the catalog database;
retrieving an identification of a second listing associated with a second signature that matches the generated signature; and
transmitting, to a user device, the identification of the second listing and a suggestion to associate the first listing with the second listing.

14. The computer-implemented method of claim 7 , wherein the plurality of attributes comprises at least one of size attributes, color attributes, or material attributes.

15. A non-transitory computer-readable medium having stored thereon an attribute identification module for analyzing an image in real-time using artificial intelligence, the attribute identification module comprising executable code that, when executed on a computing device, implements a process comprising:
receiving an image and attributes associated with the image that are stored in a database;
determining a plurality of attributes depicted in the image using an artificial intelligence system, wherein the artificial intelligence system is trained using a plurality of images, and wherein attributes of the plurality of attributes are associated with a confidence level;
identifying a first subset of attributes in the plurality of attributes, wherein the first subset comprises attributes in the plurality that are associated with a confidence level greater than a threshold value; and
comparing the attributes associated with the image with the first subset of attributes to determine whether an item description in the database accurately describes an item depicted in the image.

16. The non-transitory computer-readable medium of claim 15 , wherein th

Assignees

• Amazon Tech Inc

Inventors

• Santos Anthony Alexander
• Finkelstein Adam James
• Gingrich David Akira
• Hurley David Michael
• Zhao Siqi

Classifications

• G06V10/82 Primary

  using neural networks · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title

• G06N3/042

  Knowledge-based neural networks; Logical representations of neural networks · CPC title

• G06F18/2113

  by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation · CPC title

• G06N3/045

  Combinations of networks · CPC title