Image segmentation and vectorization system for complementary styling products

Patent metadata

Field	Value
Publication number	US-12469067-B2
Application number	US-202318112870-A
Country	US
Kind code	B2
Filing date	Feb 22, 2023
Priority date	Feb 22, 2023
Publication date	Nov 11, 2025
Grant date	Nov 11, 2025

Abstract

Systems and methods are provided for segmenting each image of a first plurality of images, using an image processing segmentation technique, into one or more category of a plurality of predefined categories to generate a set of image segments for the image. A numerical vector representation is generated for each image segment and for each image in a second plurality of images and used to determine a similarity between image segments and images in the second plurality of images. Each image segment in each set of image segment are replaced with an image in the second plurality of images that is similar to the image segment to generate a recommendation catalog including a plurality of sets of recommendation images.

First claim

What is claimed is:

1 . A computer-implemented method comprising:
for each image of a first plurality of images:
segmenting the image, using an image processing segmentation technique, into one or more category of a plurality of predefined categories to generate image segments for the image; and
generating a set identifier for the image and associating the set identifier with each image segment for the image to generate a set of image segments for the set identifier;
generating a numerical vector representation for each image segment; generating a numerical vector representation for each image in a second plurality of images, each image in the second plurality of images comprising a specific product; determining images in the second plurality of images that are similar to image segments from the first plurality of images by comparing the numerical vector representation for each image segment to the numerical vector representation for each image in the second plurality of images; replacing each image segment in each set of image segments with an image in the second plurality of images that is similar to the image segment to generate a recommendation catalog comprising a plurality of sets of recommendation images; searching the recommendation catalog to find a first image in a first set of recommendation images that corresponds to a selected product; and providing recommendation images, other than the first image, in the first set of recommendation images as complementary products to the selected product.

2 . The computer-implemented method of claim 1 , wherein the image segments are extracted from a respective image and stored individually, and
wherein the set of image segments for an image comprises the extracted image segments.

3 . The computer-implemented method of claim 1 , wherein the predefined categories correspond to clothing and accessories or to home décor products, and
wherein the second plurality of images corresponds to images in a merchant catalog and the specific product is an item of clothing or accessory or a home décor product.

4 . The computer-implemented method of claim 1 , wherein segmenting the image, using an image processing segmentation technique, into one or more category of a plurality of predefined categories to generate image segments for the image comprises:
analyzing the image using a machine learning model trained to segment and categorize objects in an image to generate a bounding box and category for each object recognized in the image.

5 . The computer-implemented method of claim 1 , wherein generating a numerical vector representation for each image segment and generating a numerical vector representation for each image in a second plurality of images comprises generating a representation of each image segment and each image in the second plurality of images as a point in n-dimensional space.

6 . The computer-implemented method of claim 1 , wherein determining images in the second plurality of images that are similar to image segments from the first plurality of images by comparing the numerical vector representation for each image segment to the numerical vector representation for each image in the second plurality of images comprises determining a distance score for each pair of image in the second plurality of images and image segment.

7 . The computer-implemented method of claim 1 , wherein an image in the second plurality of images is determined to be similar to an image segment when a distance score is greater than a predefined threshold value.

8 . The computer-implemented method of claim 1 , wherein the first plurality of images comprise images from at least one public source of images including at least one social media source.

9 . The computer-implemented method of claim 8 , further comprising:
detecting new images from the at least one public source of images; and updating the recommendation catalog based on the new images.

10 . The computer-implemented method of claim 1 , further comprising:
receiving a captured image that was captured by a computing device; segmenting the captured image, using an image processing segmentation technique, into one or more category of the plurality of predefined categories to generate a set of image segments for the captured image; comparing each image segment of the set of image segments for the captured image to each image in the second plurality of images to find at least one matching image in the second plurality of images; and providing the at least one matching image and product information about the at least one matching image to the computing device.

11 . The computer-implemented method of claim 10 , further comprising:
searching the recommendation catalog to find a second image in a second set of recommendation images that corresponds to the at least one matching image; and providing recommendation images, other than the second image, in the second set of recommendation images as complementary products to the at least one matching image.

12 . A system comprising:
a memory that stores instructions; and one or more processors configured by the instructions to perform operations comprising: for each image of a first plurality of images:
segmenting the image, using an image processing segmentation technique, into one or more category of a plurality of predefined categories to generate image segments for the image; and
generating a set identifier for the image and associating the set identifier with each image segment for the image to generate a set of image segments for the set identifier;
generating a numerical vector representation for each image segment; generating a numerical vector representation for each image in a second plurality of images, each image in the second plurality of images comprising a specific product; determining images in the second plurality of images that are similar to image segments from the first plurality of images by comparing the numerical vector representation for each image segment to the numerical vector representation for each image in the second plurality of images; replacing each image segment in each set of image segments with an image in the second plurality of images that is similar to the image segment to generate a recommendation catalog comprising a plurality of sets of recommendation images; searching the recommendation catalog to find a first image in a first set of recommendation images that corresponds to a selected product; and providing recommendation images, other than the first image, in the first set of recommendation images as complementary products to the selected product.

13 . The system of claim 12 , wherein the image segments are extracted from a respective image and stored individually, and
wherein the set of image segments for an image comprises the extracted image segments.

14 . The system of claim 12 , wherein segmenting the image, using an image processing segmentation technique, into one or more category of a plurality of predefined categories to generate image segments for the image comprises:
analyzing the image using a machine learning model trained to segment and categorize objects in an image to generate a bounding box and category for each object recognized in the image.

15 . The system of claim 12 , wherein generating a numerical vector representation for each image segment and generating a numerical vector representation for each image in a second plurality of images comprises generating a representation of each image segment and each image in the second plurality of images as a point in n-

Assignees

• Sap Se

Inventors

• Rao Arjun
• Pendar Nick

Classifications

• G06Q30/0603 Primary

  Catalogue creation or management · CPC title

• G06V10/26

  Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion · CPC title

• G06V10/765

  using rules for classification or partitioning the feature space · CPC title

• G06V10/454

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

• G06V10/255

  Detecting or recognising potential candidate objects based on visual cues, e.g. shapes · CPC title