Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Face image quality labeling method and device based on face recognition system

A face recognition system and face image technology, applied in character and pattern recognition, instruments, computer parts, etc., can solve the problem of high difficulty in data collection, achieve high labeling efficiency, accurate labeling quality, and reduce difficulty Effect

Pending Publication Date: 2021-11-23
杭州英歌智达科技有限公司
View PDF1 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0011] Aiming at the problem of high difficulty in data collection in the prior art, the present invention provides a face image quality labeling method and device based on a face recognition system

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Face image quality labeling method and device based on face recognition system
  • Face image quality labeling method and device based on face recognition system
  • Face image quality labeling method and device based on face recognition system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0044] A face image quality labeling method based on a face recognition system, the method comprising;

[0045] The image set is obtained by obtaining the first image set and the second image set;

[0046] The acquisition of the face feature vector, the first image set and the second image set are input to the face recognition network, so as to obtain the face feature vector;

[0047] The calculation of the similarity of the image set, the calculation of the similarity between the Jth human face to be marked in the second image set and the reference photo of the first image set subset I;

[0048]The calculation of the similarity variance of the image set, the similarity of the Jth face to be marked in the second image set and all subsets in the first image set is calculated for the similarity variance;

[0049] The determination of the face image quality score label value is based on the variance of the image similarity to determine the face image quality score.

[0050] By ...

Embodiment 2

[0062] On the basis of embodiment 1, the face image quality labeling device realized based on the face image quality labeling method of the face recognition system; it includes an image collection unit, a feature extraction unit, a first calculation unit and a score fusion unit;

[0063] An image collection unit, used to collect the first image set and the second image set in the face recognition system;

[0064] The feature extraction unit is used to obtain the face feature vector of each sample in the image set based on the face recognition network for the image set obtained by the image collection unit;

[0065] The first calculation unit is used to calculate the variance of the similarity distribution between its feature vector and the reference photo feature vector in the first image set for each sample to be labeled;

[0066] The score fusion unit is configured to, for each sample to be labeled, map the output score of the first calculation unit based on a certain functi...

Embodiment 3

[0068] On the basis of the above embodiments, in the recognition process of the face recognition system, for a snapshot,

[0069] First, it will calculate its similarity with each bottom library (I=1,2,...,N) in the bottom library set with a quantity of N, and finally assign it to the bottom library I with the largest similarity;

[0070] At this time, if the allocation is correct, that is, the snapshot and the bottom library are the same person, which is called recall;

[0071] If the assignment is wrong, that is, the snapshot and the assigned bottom library are not the same person, which is called a false positive.

[0072] The criterion for evaluating a face recognition system is that the recall rate should be large and the false positive rate should be small.

[0073] Reflected on the face vector features, that is to say, the distance between the face feature vectors under the same ID (intra-class) should be as close as possible, and the distance between the face feature ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention relates to the field of deep learning, and discloses a face image quality labeling method and device based on a face recognition system. The method comprises the steps of obtaining a first image set and a second image set; inputting the first image set and the second image set into a face recognition network to obtain a face feature vector; calculating the similarity of the image sets, and calculating the similarity of the jth to-be-labeled face in the second image set and the reference photo of the subset i of the first image set; calculating the similarity variance of the image sets, namely calculating the similarity variance of the jth to-be-labeled face in the second image set and the similarity of all subsets in the first image set; and determining the quality score of the face image according to the variance of the image similarity. In the method, the variance is utilized to reflect the quality of the to-be-labeled sample, the method is coupled with the recognition capability of a face recognition system, and the labeling result is more objective by only considering the similarity distribution with other people; therefore, it is not required that the to-be-labeled sample needs a reference photo of the person, and the difficulty degree of data collection is reduced.

Description

technical field [0001] The invention relates to the field of deep learning, in particular to a face image quality labeling method and device based on a face recognition system. Background technique [0002] Face image quality analysis (FIQA) is an essential front-end processing module for face recognition systems to be put into practical use. [0003] By filtering out inappropriate captured faces, it can ensure the high quality of the face images entering the subsequent face recognition process, thereby reducing false positives (that is, matching the captured faces to the wrong bottom library photos) and introducing the optimal frame person face so as to save the computing power required for subsequent face recognition. [0004] At present, the biggest problem in face quality analysis lies in the formulation of quality score labels. Different from the image quality analysis task, it removes the noise, codec, and out-of-focus problems of the image itself, and more important...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06K9/62G06K9/00
CPCG06F18/22
Inventor 卿敏敏徐联伯
Owner 杭州英歌智达科技有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com