logo

For the Inventor, By the Inventor

The patent badge is an abbreviated version of the USPTO patent document. The patent badge does contain a link to the full patent document.

The patent badge is an abbreviated version of the USPTO patent document. The patent badge covers the following: Patent number, Date patent was issued, Date patent was filed, Title of the patent, Applicant, Inventor, Assignee, Attorney firm, Primary examiner, Assistant examiner, CPCs, and Abstract. The patent badge does contain a link to the full patent document (in Adobe Acrobat format, aka pdf). To download or print any patent click here.

Patent No.:

US 12235892 B1

PDFFull Text
Date of Patent:
Feb. 25, 2025

Filed:

May. 19, 2022

Data cleaning device, data cleaning method and face verification method

Applicant:

Guangdong Oppo Mobile Telecommunications Corp., Ltd., Guangdong, CN;

Inventors:

Jie Liu, Palo Alto, CA (US);

Yang Zhou, Palo Alto, CA (US);

Assignee:

GUANGDONG OPPO MOBILE TELECOMMUNICATIONS CORP., LTD., Guangdong, CN;

Attorney:

Hodgson Russ LLP

Primary Examiner:

Quang N Vo

Int. Cl.
CPC ...
G06V 10/74 (2022.01); G06F 16/55 (2019.01); G06F 18/2413 (2023.01); G06K 9/62 (2022.01); G06V 10/30 (2022.01); G06V 10/44 (2022.01); G06V 10/774 (2022.01); G06V 10/778 (2022.01); G06V 10/82 (2022.01); G06V 40/16 (2022.01); G06V 40/50 (2022.01);
U.S. Cl.
CPC ...
G06F 16/55 (2019.01); G06F 18/2413 (2023.01); G06V 10/30 (2022.01); G06V 10/454 (2022.01); G06V 10/761 (2022.01); G06V 10/774 (2022.01); G06V 10/7784 (2022.01); G06V 10/82 (2022.01); G06V 40/161 (2022.01); G06V 40/171 (2022.01); G06V 40/172 (2022.01); G06V 40/50 (2022.01);
Abstract

A data cleaning method is provided. The method incudes: acquiring a training face dataset, the training face dataset including a plurality of training images each labeled with a person ID; acquiring a candidate face verification model and training the candidate face verification model by the plurality of training images; acquiring a plurality of feature embeddings from the candidate face verification model after training, and determining a similarity between the average feature embedding of one person ID and each image labeled as the same person ID; extracting at least one training image whose similarity is smaller than a similarity threshold from the plurality of training images; and excluding the at least one extracted training image in response to receiving a confirmation signal.

Find Patent Forward Citations

Loading…