Color-based recommendations using product images

• US 10,296,960 B1
• Filed: 05/19/2015
• Issued: 05/21/2019
• Est. Priority Date: 05/19/2015
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Gyllstrom, Karl Anders
• Primary Examiner(s): Dunham, Jason B
• Assistant Examiner(s): Sutch, Jr., Norman Donald
• Application Number:
  US14/716,779
• Time in Patent Office: 1,463 Days
• Field of Search: None
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06Q 30/0631 Item recommendations
• CPC Class Codes: G06T 7/90 Determination of colour characteristics

Abstract

Techniques described herein include a system and method for identifying color complements from an electronic marketplace catalog. In particular, the disclosure is directed to extracting color palette information from a variety of item images in the electronic catalog and creating clusters into which separate items are placed based on their similarity in colors. Multi-colored items may belong to more than one cluster (they may belong to a cluster for each color associated with the item). The clusters that share multi-colored items may be analyzed to determine the strength of a relationship between the two clusters. Clusters that share a significant number of items may be associated with complementary colors. In this disclosure, a service provider may receive a request related to an item, and may subsequently identify and recommend a complementary item.

Claims

1. A computer-implemented method, comprising:

maintaining, in a catalog of an electronic marketplace, image information related to a plurality of clothing items;

identifying, from the image information related to the plurality of clothing items, color information of the plurality of clothing items by color sampling pixels within the image information such that the color information is determined as a percentage of the pixels sampled;

generating, based at least in part on the identified color information, a plurality of color groupings to which the plurality of clothing items are associated;

identifying, from the plurality of clothing items, a number of clothing items each associated with at least two color groupings of the plurality of color groupings;

identifying, based at least in part on the on the number of clothing items being greater than a threshold value, a complementary relationship between the at least two color groupings;

receiving a request related to a first item of the plurality of clothing items;

determining that the first item is associated with a first color grouping of the at least two color groupings;

identifying a number of second items associated with a second color grouping of the at least two color groupings;

determining, based at least in part on the complementary relationship between the at least two color groupings, that a second item of the number of second items is a complementary item; and

presenting a recommendation related to the second item.

2. The computer-implemented method of claim 1, wherein the color information is identified from the image information by extracting color values in a color space from an image using machine vision.

3. The computer-implemented method of claim 1, wherein the image information is received from a third party associated with an item of the plurality of items.

4. The computer-implemented method of claim 1, wherein an item of the plurality of items is associated with a color grouping of the plurality of color groupings based at least in part on a similarity between the color grouping and the color information for the item.

5. A system, comprising:

a processor; and

a memory including instructions that, when executed with the processor, cause the system to, at least;

identify, from image information related to a plurality of clothing items, color information of the plurality of clothing items by color sampling pixels within the image information;

associate individual clothing items of the plurality of clothing items with one or more color groups based on the color information;

upon determining that a number of the same individual clothing items are associated with two color groups, create a complementary relationship between the two color groups;

receive information related to a first item, the information including color data;

determine, based at least in part on color data, that the item is to be assigned to a first color group of the two color groups;

identify one or more second items associated with a second color group of the two color groups; and

provide a recommendation for the one or more second items.

6. The system of claim 5, wherein the information related to the first item includes a digital image of the item.

7. The system of claim 5, wherein the first color group for the first item to be assigned is determined based at least in part on a distance between a cluster point of the first color group and at least one color of the color data included in the information related to the first item.

8. The system of claim 5, wherein the complementary relationship is created based at least in part on an number of the same individual clothing items shared by the second color group and the first color group exceeding a threshold value.

9. The system of claim 5, wherein the complementary relationship is created based at least in part on an item rating, a category associated with at least one shared item, or a percentage of the at least one shared item that corresponds to the first color group.

10. The system of claim 5, wherein the one or more second items associated with the second color group includes at least one item is worn with the first item.

11. The system of claim 5, wherein the recommendation is presented within a user interface as a notification message that includes an option to order the one or more second items.

12. The system of claim 5, further comprising instructions that cause the system to, at least provide a second recommendation for a third item, the third item being associated with the first color group.

13. A non-transitory computer readable storage medium storing computer-executable instructions that, when executed by a processor, cause a computer system to at least perform operations comprising:

identifying, from image information related to a plurality of clothing items, color information of the plurality of clothing items by color sampling pixels within the image information;

associating individual clothing items of the plurality of clothing items with one or more color groupings based on the color information;

upon determining that two color groupings are associated with a number of the same individual clothing items, identifying a complementary relationship between the two color groupings;

receiving a request related to a first item;

determining the first item is associated with a first color grouping of the two color groupings;

identifying a second item based at least in part on the second item being associated with a second color grouping of the two color groupings;

and

providing a recommendation for the second item.

14. The computer readable storage medium of claim 13, wherein the request related to the first item is the request to order the first item.

15. The computer readable storage medium of claim 13, wherein the second item is further identified based at least in part on the second item corresponding to a purchase history.

16. The computer readable storage medium of claim 13, wherein the received request includes an image, and the determined color associated with the first item is identified from the image.

17. The computer readable storage medium of claim 16, wherein the first item is identified from the image based at least in part on an object recognition process.

18. The computer readable storage medium of claim 13, wherein the plurality of clothing items includes at least a multi-colored clothing item received from a third-party.

19. The computer readable storage medium of claim 13, wherein the first color grouping and the second color grouping are a minimum distance apart.