Method, medium, and system for intelligent online personal assistant with image text localization

Patent metadata

Field	Value
Publication number	US-12223533-B2
Application number	US-202117222251-A
Country	US
Kind code	B2
Filing date	Apr 5, 2021
Priority date	Nov 11, 2016
Publication date	Feb 11, 2025
Grant date	Feb 11, 2025

Abstract

Systems, methods, and computer program products for identifying a candidate product in an electronic marketplace based on a visual comparison between candidate product image visual text content and input query image visual text content. Unlike conventional optical character recognition (OCR) based systems, embodiments automatically localize and isolate portions of a candidate product image and an input query image that each contain visual text content, and calculate a visual similarity measure between the respective portions. A trained neural network may be re-trained to more effectively find visual text content by using the localized and isolated visual text content portions as additional ground truths. The visual similarity measure serves as a visual search result score for the candidate product. Any number of images of any number of candidate products may be compared to an input query image to enable text-in-image based product searching without resorting to conventional OCR techniques.

First claim

What is claimed is:

1. A method comprising:
localizing and isolating a candidate portion of a candidate product image with candidate visual text content;
retraining a neural network using the candidate portion of the candidate product image;
localizing and isolating, via the retrained neural network, an input query portion of an input query image with input query visual text content;
calculating a visual similarity measure between the candidate portion and the input query portion based on a distance value between image signatures associated with the candidate product image and the input query image;
ranking a candidate product in a product list based on the visual similarity measure; and
causing presentation of the product list on a graphical user interface of a client device.

2. The method of claim 1 , wherein the input query image comprises one of:
a photograph;
a video frame;
a sketch; or
a diagram.

3. The method of claim 1 , wherein the candidate product image is associated with an electronic marketplace.

4. A non-transitory computer-readable storage medium having embedded therein a set of instructions which, when executed by one or more processors of a computer, causes the computer to execute operations comprising:
localizing and isolating a candidate portion of a candidate product image with candidate visual text content;
retraining a neural network using the candidate portion of the candidate product image;
localizing and isolating, via the retrained neural network, an input query portion of an input query image with input query visual text content;
calculating a visual similarity measure between the respective portions of the candidate product image and the input query image based on a distance value between image signatures associated with the candidate product image and the input query image;
ranking a candidate product in a product list based on the visual similarity measure; and
causing presentation of the product list on a graphical user interface of a client device.

5. A system comprising:
at least one processor; and
memory encoding computer-executable instructions that, when executed by the at least one processor, cause the system to perform operations comprising:
localizing and isolating a candidate portion of a candidate product image with candidate visual text content;
retraining a neural network using the candidate portion of the candidate product image;
localizing and isolating, via the retrained neural network, an input query portion of an input query image with input query visual text content;
calculating a visual similarity measure between the candidate portion and the input query portion based on a distance value between image signatures associated with the candidate product image and the input query image;
ranking a candidate product in a product list based on the visual similarity measure; and
causing presentation of the product list on a graphical user interface of a client device.

6. The system of claim 5 , wherein the input query image comprises one of:
a photograph;
a video frame;
a sketch; or
a diagram.

7. The method of claim 1 , wherein localizing and isolating the respective portions of the candidate product image and the input query image comprises localizing and isolating, using a machine learning model, the respective portions of the candidate product image and the input query image.

8. The method of claim 7 , wherein the machine learning model comprises a deep neural network.

9. The method of claim 7 , further comprising retraining the machine learning model using a new image of a new product provided to an electronic marketplace.

10. The method of claim 1 , wherein the image signatures comprise binary vectors, and the distance value comprises one or more bits that are different in the binary vectors.

11. The non-transitory computer-readable storage medium of claim 4 , wherein the input query image comprises one of:
a photograph;
a video frame;
a sketch; or
a diagram.

12. The non-transitory computer-readable storage medium of claim 4 , wherein the candidate product image is associated with an electronic marketplace.

13. The non-transitory computer-readable storage medium of claim 4 , wherein localizing and isolating the respective portions of the candidate product image and the input query image comprises localizing and isolating, using a machine learning model, the respective portions of the candidate product image and the input query image.

14. The non-transitory computer-readable storage medium of claim 13 , wherein the machine learning model comprises a deep neural network.

15. The non-transitory computer-readable storage medium of claim 13 , wherein the operations further comprise retraining the machine learning model using a new image of a new product provided to an electronic marketplace.

16. The non-transitory computer-readable storage medium of claim 4 , wherein the image signatures comprise binary vectors, and the distance value comprises one or more bits that are different in the binary vectors.

17. The system of claim 5 , wherein the candidate product image is associated with an electronic marketplace.

18. The system of claim 5 , wherein localizing and isolating the respective portions of the candidate product image and the input query image comprises localizing and isolating, using a machine learning model, the respective portions of the candidate product image and the input query image.

19. The system of claim 18 , wherein the operations further comprise retraining the machine learning model using a new image of a new product provided to an electronic marketplace.

20. The system of claim 5 , wherein the image signatures comprise binary vectors, and the distance value comprises one or more bits that are different in the binary vectors.

Assignees

• Ebay Inc

Inventors

• Zheng Shuai
• Piramuthu Robinson

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06T2207/30108

  Industrial image inspection · CPC title

• G06T7/0004

  Industrial image inspection · CPC title