Deep learning-based face recognition and face verification supervised learning method

A technology of supervised learning and face recognition, applied in the field of supervised learning, can solve the problems of increased computational complexity and inconvenient training process, and achieve the effect of stable face recognition, reduced intra-class distance and simple model

Inactive Publication Date: 2018-07-06
TIANJIN UNIV
View PDF1 Cites 26 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

But it will significantly increase the computational complexity, and the training process becomes inconvenient

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
  • Deep learning-based face recognition and face verification supervised learning method
  • Deep learning-based face recognition and face verification supervised learning method
  • Deep learning-based face recognition and face verification supervised learning method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0052] In order to solve the problems revealed by the example model experiments, it is necessary to find a method that can reduce the intra-class distance and increase the inter-class distance of deep features. Studies have shown that increasing inter-class variables by extracting deep features from different people and reducing inter-class distance by clustering different features extracted from the same person are very important for face recognition and verification [11] . In this method, experiments are designed to explain how the parameter λ affects the outcome distribution in the proposed joint supervised learning, as Figure 4 demonstrates that different λ leads to different feature distributions. If the value of λ is appropriate, the distance between features of different categories increases, and the distance between features of the same category decreases, which significantly enhances the discriminative ability of deep features. Therefore, joint supervision benefits...

Embodiment 2

[0061] The scheme in embodiment 1 is further introduced below in conjunction with specific calculation formulas and examples, see the following description for details:

[0062] 201: Process large data sets, keep the number of characters in the data set unchanged, and reduce the number of pictures in the training data set to meet the protocol of small data sets;

[0063] The key feature points of faces and faces in all pictures are determined by the algorithm [13] detection. This method uses 5 key feature points of the face (two eyes, nose, two edges of the mouth) to do similarity transformation. For the training set, the image is simply discarded if the detection fails, but for the test set, the provided key features are used.

[0064] The image is cropped to a 112×96 RGB image. According to the convention of image preprocessing, each pixel in the RGB image (pixel value [0, 255]) is subtracted from 127.5, and then divided by 128 for normalization.

[0065] For the trainin...

Embodiment 3

[0102] The scheme in embodiment 1 and 2 is carried out feasibility verification below in conjunction with specific example, see the following description for details:

[0103] data set:

[0104] The LFW database is currently the most widely used face image database, such as Figure 5(a) and 5(b) shown. The database has a total of 13233 images, including 5749 different people. 1680 people had two or more images, and 4069 people had only one image. There are significant differences in pose, expression, and lighting between each image. Images are in JPEG format with a size of 250×250. The vast majority are color images, and a few are grayscale images. The database collects face pictures under natural conditions, and the purpose is to improve the accuracy of face recognition under natural conditions. There are 6 evaluation criteria for this dataset: Unsupervised; Image-restricted with no outside data; Unrestricted with no outside data; Unlabeled data Image-restricted with ...

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 deep learning-based face recognition and face verification supervised learning method. The method comprises the following steps: a soft maximum loss function is used to increase a between-class distance for full connection layer output characteristics of a convolutional neural network model, a center is learnt for the depth characteristics of each class through a centralloss function, a super parameter is used to balance the two functions to thus jointly supervise the learning characteristics; backward propagation of the convolution neural network model is calculated, a stochastic gradient descent algorithm based on minimum batch processing is adopted to optimize the convolutional neural network model, and a weight matrix and the depth characteristic center of each class are updated; and after the depth characteristics are subjected to principal component analysis and dimension reduction, the cosine distance between each two characteristics is calculated to calculate a score, wherein the score is used for target matching in nearest neighbor and threshold comparison, and a face is recognized and verified. The identification ability of the neural network learning characteristics can be effectively improved, and a face characteristic recognition and face verification mode with robustness is acquired.

Description

technical field [0001] The invention relates to the field of face recognition, in particular to a supervised learning method for face recognition and face verification based on deep learning. Background technique [0002] CNN (Convolutional Neural Network) has achieved great success in the field of machine vision and significantly consolidated the field of classification problems, including: object classification, scene classification, and action classification, etc. Mainly benefit from large training dataset and end-to-end learning structure. CNN is most commonly used for feature learning and label prediction, that is, to map input data to deep learning features (the output of the last hidden layer) to predict the label of input data. [0003] Recently, references [1] The learning process of the CNN model is supervised by modifying the identification signal of the softmax (soft maximum) loss function, which makes the deep learning features contain richer identification in...

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 Applications(China)
IPC IPC(8): G06K9/00G06K9/62G06N3/04G06N3/08
CPCG06N3/084G06V40/168G06N3/045G06F18/2135G06F18/214
Inventor 聂为之李梦洁
Owner TIANJIN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap