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

A real-time face recognition method and device

A face recognition and recognition method technology, applied in neural learning methods, character and pattern recognition, instruments, etc., can solve the problems of the real-time impact of face recognition network models, and achieve obvious advantages, improve real-time performance, and improve The effect of face recognition speed

Active Publication Date: 2022-04-08
FENGHUO COMM SCI & TECH CO LTD
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In particular, the deletion of face quality assessment will have a greater impact on the real-time performance of deep face recognition network models

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
  • A real-time face recognition method and device
  • A real-time face recognition method and device
  • A real-time face recognition method and device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0046]FaceQnet is a face quality assessment model based on face recognition, which is more robust than the current deep learning for face angle, ambiguity and light condition judgment or single quality feature. Non-face and low-quality face evaluation scores are relatively low, and the accuracy of face quality evaluation is higher than that of other face quality evaluation methods and algorithm models. FaceQnet is a face quality assessment model trained by ISO / ICAO to generate the highest-scoring face image as ground truth, using the highest-scoring face image as a benchmark, and using Facenet face comparison similarity scoring as a label. The biggest advantage of this model is based on the face quality evaluation model trained by face recognition. The evaluated high-quality faces are more suitable for face recognition and can improve the accuracy of face recognition.

[0047] The face detection model yolov3-tiny backbone network is based on darknet19, which is a lightweight n...

Embodiment 2

[0051] Such as figure 2 As shown, the embodiment of the present invention provides a method for real-time face recognition, including:

[0052] S1, the face data training set is obtained after the original data set is evaluated based on the FaceQnet face quality evaluation model;

[0053] Specifically, in order to increase the data volume of the data set and the wide representativeness of the data, the widerface and vggface2 data sets were combined as the original data set to calculate the face pictures selected in the widerface and vggface2 data sets through the FaceQnet face quality assessment model The face quality evaluation score FaceQualityScore is used as the training label of the selected face picture to generate a face data training set. The face data set in the present invention is not limited to widerface and vggface2, and other face detection data sets or data collected and tagged according to actual use scenarios can be used.

[0054] The face quality evaluatio...

Embodiment 3

[0069] Such as image 3 As shown, the embodiment of the present invention provides a real-time face recognition method. When the face category confidence is higher than the preset threshold and the FaceQualityScore is smaller than the preset face quality evaluation score threshold, it is considered that the non-human face is directly detected in the face detection process. throw away.

[0070] For example, in the embodiment of the present invention, two human face quality assessment score thresholds can be set: the first human face quality assessment score threshold and the second human face quality assessment score threshold, wherein the first human face quality assessment score threshold is less than the second human face quality assessment score threshold Evaluation score threshold. First judge whether the confidence of the face category is higher than the preset confidence threshold of the face category and the face quality evaluation score is greater than the first face ...

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 discloses a real-time face recognition method, comprising: evaluating the original data set based on the FaceQnet face quality evaluation model to obtain a face data training set; using the training face data to train an optimized yolov3-tiny network Model, obtains the optimized yolov3-tiny network model through training; Utilize the optimized yolov3-tiny network model through training to detect the picture to be detected, obtain the face category confidence and the face quality evaluation score of each picture to be detected; Using the face category confidence and the face quality assessment score of the picture to be detected to determine whether to perform face recognition processing on the picture to be detected. The technical solution of the present invention combines the two models of FaceQnet and yolov3-tiny into one model, which reduces the calculation amount of the AI ​​chip, thereby further improving the real-time performance of face recognition. The invention also provides a corresponding real-time face recognition device.

Description

technical field [0001] The invention belongs to the technical field of image recognition, and more particularly relates to a method and device for real-time face recognition. Background technique [0002] Embedded face recognition is to use the edge AI chip acceleration engine on the embedded terminal to complete a large number of calculation tasks of the face algorithm, and finally complete the process of face recognition. Embedded face recognition has the advantages of small size, low cost, easy deployment, and convenient distributed computing. [0003] At present, for AI chips with low computing power at the embedded end, in order to reduce the calculation amount of the face algorithm and shorten the calculation time at the embedded end, conventional practice 1: pruning, compressing, and quantizing the model to reduce the model volume, but at the same time resulting in a corresponding decrease in accuracy. Conventional practice two: reduce unnecessary links of face reco...

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
Patent Type & Authority Patents(China)
IPC IPC(8): G06V40/16G06V10/774G06V10/82G06N3/04G06N3/08
CPCG06N3/08G06V40/161G06V40/172G06N3/045G06F18/214
Inventor 聂建平
Owner FENGHUO COMM SCI & TECH CO LTD
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