CN111582044A - Face recognition method based on convolutional neural network and attention model - Google Patents

Face recognition method based on convolutional neural network and attention model Download PDF

Info

Publication number
CN111582044A
CN111582044A CN202010295613.8A CN202010295613A CN111582044A CN 111582044 A CN111582044 A CN 111582044A CN 202010295613 A CN202010295613 A CN 202010295613A CN 111582044 A CN111582044 A CN 111582044A
Authority
CN
China
Prior art keywords
attention
feature
neural network
image
convolutional neural
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010295613.8A
Other languages
Chinese (zh)
Other versions
CN111582044B (en
Inventor
贺前华
杨泽睿
庞文丰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
South China University of Technology SCUT
Original Assignee
South China University of Technology SCUT
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by South China University of Technology SCUT filed Critical South China University of Technology SCUT
Priority to CN202010295613.8A priority Critical patent/CN111582044B/en
Publication of CN111582044A publication Critical patent/CN111582044A/en
Application granted granted Critical
Publication of CN111582044B publication Critical patent/CN111582044B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/12—Fingerprints or palmprints
    • G06V40/1365—Matching; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/12—Fingerprints or palmprints
    • G06V40/1347—Preprocessing; Feature extraction
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Human Computer Interaction (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Analysis (AREA)

Abstract

The invention relates to a face recognition technology, in particular to a face recognition method based on a convolutional neural network and an attention model, which comprises the following steps: preprocessing a face image; inputting the preprocessed image data into a convolutional neural network to extract high-dimensional features; inputting the high-dimensional features into an attention model, calculating an attention mask through a training method of a neural network, and obtaining the attention distribution features M (F) of a space domain and a channel domainc) (ii) a Feature M (F) of attention distributionc) Inputting the feature into a Bottleneck module, and acquiring a feature H (M (F) by using a short mechanismc) ); will be characterized by H (M (F)c) Input into the full convolution network classification structure, and use Dropout strategy and softmax function to obtain the final face recognition result. The invention adopts the Attention model and the Bottleneck module to replace the middle of the VGG modelThe high-level network greatly reduces the number of parameters required by the model, and effectively reduces the video memory and time consumption during training; meanwhile, the learning capability of the network to data is effectively improved, and the stability and the practicability of the system are improved.

Description

Face recognition method based on convolutional neural network and attention model
Technical Field
The invention relates to a face recognition technology, in particular to a face recognition method based on a convolutional neural network and an attention model.
Background
Face recognition refers to a technique of recognizing or verifying one or more faces by extracting face information in a still image or a video sequence and comparing the face information with information in an existing face database. Compared with biological characteristic identification means such as signature and fingerprint, the information extraction of the image-based identity identification system is more convenient. It is one of the most active research directions in the fields of computer vision and pattern recognition due to its wide application direction and commercial value. In recent years, face identification networks based on deep learning have achieved remarkable results. Meanwhile, the attention model is widely used in various fields of deep learning. The visual attention mechanism mimics the brain signal processing mechanisms characteristic of human vision. By quickly scanning the global image, a target area needing important attention is obtained, and then more attention resources are put into the area to obtain more detailed information of the target needing attention and suppress other useless information. The method greatly improves the efficiency and accuracy of visual information processing.
In recent years, the research work of deep learning in combination with visual attention mechanisms has mostly focused on the use of masks to form the attention mechanism. The principle of the mask is that key features in the image data are identified through another layer of new weight, and the deep neural network learns the region needing attention in each new image through learning training, so that attention is brought to the user. Attention has two major classifications: soft attentions (soft attentions) and hard attentions (hardattentions). Strong attention is a stochastic prediction process, emphasizing dynamic changes and being inconsequential, and training generally needs to be done by reinforcement learning. The key to soft attention is that it can be miniaturized, i.e. the gradient can be calculated, and it can be trained using neural network methods. In the CNN-based attention model, the two attention domains are mainly classified, namely, spatial domain (spatial domain) and channel domain (channel domain). Attention in the spatial domain is mainly distributed in space, and is expressed on an image that attention degrees of different areas on the image are different, and is reflected in mathematics, that is, for a certain feature map with the size of C × W × H, effective spatial attention corresponds to a matrix with the size of W × H, and each position is a weight for a pixel at a position corresponding to an original feature map. The method integrates and compresses the multi-channel numerical values of the same pixel, and differentiates the pixel when calculating the attention weight, so that the spatial attention of the features can be obtained. The attention of the channel domain is mainly distributed in the channels, which means that the attention of different image channels is different when the image is displayed. Mathematically, that is, for a feature map of size C × W × H, a certain channel attention corresponds to a matrix of size C × 1 × 1, and each position is a weight for all pixels of the corresponding channel of the original feature map. The method integrates and compresses a plurality of pixel values of the same channel, and differentiates the channel when calculating the attention weight, so that the characteristic channel attention can be obtained.
The existing face recognition method based on deep learning and attention model uses training samples to automatically learn the face features, and can extract excellent face features with discrimination. However, an efficient neural network often requires a large number of parameters, and the consumption of computing resources during training and testing is large, which also brings difficulties to the migration, overloading and embedded design of the network. There is a need to find a design solution for a miniaturized network.
Disclosure of Invention
The invention provides a face recognition method based on a convolutional neural network and an attention model, which aims to solve the problem that the existing face recognition neural network generally consumes large computing resources, thereby reducing the requirements of a network model on equipment and simultaneously improving the computing speed; by using the attention model and the Bottleneck structure, on the basis of further reducing network parameters, the defect that the shallow network is difficult to extract the characteristics with rich semantic information is overcome, the recognition task is completed by combining the semantic information of each scale, and a high-accuracy recognition result is obtained.
The invention is realized by the following technical scheme: the face recognition method based on the convolutional neural network and the attention model comprises the following steps:
s1, preprocessing the face image;
s2, inputting the preprocessed image data into a convolutional neural network to extract high-dimensional features to obtain high-dimensional features Fc；
S3, converting the high-dimensional feature FcInputting the data into an attention model, calculating an attention mask through a training method of a neural network, and obtaining an attention distribution characteristic M (F) of a space domain and a channel domainc)；
S4, feature M (F) of attention distributionc) Inputting the feature into a Bottleneck module, and acquiring a feature H (M (F) by using a short mechanismc))；
S5, converting the characteristic H (M (F)c) Input into the full convolution network classification structure, and use Dropout strategy and softmax function to obtain the final face recognition result.
According to the technical scheme, the Attention model (Attention model) and the Bottleneeck module are adopted to replace a middle-high network in the VGG model, so that the quantity of parameters required by the model is greatly reduced, and the video memory and the time consumption during training are effectively reduced; meanwhile, the learning capability of the network to data is effectively improved, and the stability and the practicability of the system are improved. Compared with the prior art, the invention has the following advantages and beneficial effects:
1. the integral neural network remarkably reduces the number of parameters while keeping the simple structure characteristic and good generalization performance of the VGG, greatly reduces the video memory and memory occupation during training, and simultaneously, the extremely small network model is also beneficial to embedded design.
2. The attention module is designed based on one-dimensional convolution, and the information of a pixel domain and a channel domain is fully used on the basis of further reducing the number of parameters.
3. According to the invention, the Bottleneck module is based on the design of a one-dimensional volume and a single Batchnorm layer, so that the calculation time is reduced on the basis of further reducing the number of parameters, and the capture of multi-dimensional information is completed.
4. The method has the capability of fast and efficient learning of the local data set, and still has high performance when processing data with illumination difference and posture change.
Drawings
Fig. 1 is a schematic flowchart of a face recognition method based on a convolutional neural network and an attention model according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a convolutional neural network feature extraction module according to an embodiment of the present invention;
fig. 3 is a schematic diagram of an attention model and a bottellnk module according to an embodiment of the present invention;
fig. 4 is a schematic application diagram of a face recognition neural network module based on a convolutional neural network and an attention model according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The embodiment provides a face recognition method based on a convolutional neural network and an Attention model, wherein an Attention model and a Bottleneeck module are adopted to process output characteristics of a shallow convolutional neural network, and high-dimensional semantic characteristics including pixel domain and channel domain information are established; as shown in fig. 1, the method specifically includes the following steps:
1) preprocessing the face image: the preprocessing comprises the steps of converting an RGB face image into a gray image, randomly turning the image left and right, standardizing an image tensor, and normalizing the image pixel size into a normalized image
Figure BDA0002452075810000032
Normalized to 256 × 256 in this embodiment;
2) and Convolutional Neural Network (CNN) feature extraction: inputting the image data preprocessed in the step 1) into a convolutional neural network to extract high-dimensional features to obtain high-dimensional features FcWhereinThe convolutional neural network structure adopts the design of an improved VGG network;
3) attention (Attention) model: the high-dimensional feature F output by the step 2) is processedcInputting the data into a soft attention model, calculating an attention mask through a training method of a neural network, and obtaining an attention distribution characteristic M (F) of a space domain and a channel domainc) Wherein the attention model is based on a CNN network design;
4) a Bottleneeck module: the attention distribution characteristics M (F) output by the step 3) are processedc) Inputting the feature into a Bottleneck module, and acquiring a feature H (M (F) by using a short mechanismc) Wherein the Bottleneeck module is based on one-dimensional convolution;
5) full Convolutional Network (FCN) classification: the characteristics H (M (F) output by the step 4) are processedc) Input into the FCN classification structure to obtain the final face recognition result. The FCN classification structure uses the Dropout policy and the softmax function.
Further, in the step 1) of preprocessing the face image, the RGB image is converted into a gray image, specifically:
after the RGB image is converted into the gray image, the single-channel gray image is expanded into an N-channel gray image, and the image conversion calculation formula and the channel expansion formula are as follows:
Figure BDA0002452075810000031
Grayi＝λiGray+biasi
in the formula, R, G, and B represent RGB three-channel pixel values of the original image, respectively. i denotes the ith gray channel,
Figure BDA0002452075810000041
biasiin order to be a term of the offset,
Figure BDA0002452075810000042
Grayi∈[0,255]. In this embodiment, the number of channels N is 3, λ1、λ2、λ3Respectively taking 0.37, 0.30 and 0.33; biasiIs a value generated randomly.
Further, the image tensor is normalized in step 1) as follows:
Figure BDA0002452075810000043
in the formula, XiInput data for i channel, E (X)i) Is a mean value, σ (X)i) Is the standard deviation. In the present embodiment, since the picture is mapped to the (0, 1) space when the tensor conversion is performed, E (X) is specified in order to reduce the amount of calculation and obtain more resolved input datai) 0.5 and σ (X)i) When X is equal to 0.5, mixingiMapping to the (-1, 1) space allows negative samples to be generated.
In step 2), the input image data is subjected to batch normalization (Batchnormalization) after passing through a first maximum pooling (MaxPooling) layer, and picture down-sampling is performed after each convolutional layer by using the maximum pooling, wherein the number of convolutional layers is 5, and the shrunk feature size W × H (8 × 8 in the embodiment) is the original picture size
Figure BDA0002452075810000044
High-dimensional feature of output FcThe form is as follows: (C)c,Wc,Hc) (ii) a The mean and variance of the batch normalization were as follows:
Figure BDA0002452075810000045
Figure BDA0002452075810000046
in the formula, WhFor convolutional neural network parameters, xiFor convolutional neural network input, m is the batch size. In this example, the batch size value is 32, and the output characteristic FcThe form is as follows: (32, 128,8,8) discarding the bias term bias when performing the batch normalization calculation.
As shown in fig. 3, Attention information is added to different channels and different pixel points of the output high-dimensional features by using the Attention model, and information of different dimensions is further captured by using a Bottleneck module, so that the performance of the convolutional neural network is improved, specifically:
firstly, the size output in the step 2) is (C)c,Wc,Hc) Is spliced into a length Cc×Wc×HcIs input vector Vc(ii) a The dimensions of the three-dimensional feature in this embodiment are (128,8,8), vector VcLength 128 × 8 × 8 ═ 8192;
second, adopting full-connection network structure to correlate input vector VcAnd a length Wc×HcIs output vector VpEstablishing the relationship between each high-dimensional characteristic pixel point and other pixel points and channels; output vector V in the present embodimentpLength 8 × 8-64;
thirdly, the convolution output is processed by a sigmoid activation function to generate an attention feature graph sigma, the size of the attention feature graph is 8 × 8-64, and when the mask display is carried out, the vector needs to be reconstructed into 8 × 8 image data;
fourthly, the obtained attention feature map and the output feature F of the step 2)cCarrying out tensor multiplication to form attention enhancement on a space domain and a channel domain to obtain the feature M (F) with enhanced attentionc)：
M(Fc)＝ψ(fC×W×H(Contact(Fc)))＝ψ(fC×W×H(Vc))
Where Contact () is the matrix splice, fC×W×H() For feature vector VcThe net computation performed, ψ () is the tensor multiplication (mask) computation of the attention feature map. In this embodiment, tensor multiplication (mask) calculation is based on a broadcast mechanism (broadcast); specifically, the high-dimensional feature has a scale of (32, 128, 64), and the mask having the scale of (32, 1, 64) is multiplied by the corresponding element having the scale of (32, 64) for each high-dimensional feature of the 128 channels to obtain a feature M (F)c) Has a dimension of (32, 128, 64).
Further, the process of using the short mechanism to acquire the features by the bottleeck module in the step 4) specifically includes:
firstly, using 1 × 1 one-dimensional convolution to convert the feature M (F) output by the step 3)c) Reducing the dimension and fusing information among different channels; the present embodiment reduces the number of channels from 128 to 32 by dimension reduction;
performing zero filling processing on two ends of the feature vector after dimension reduction, and then performing one-dimensional convolution with a convolution kernel of 3, namely performing 3 x 3 convolution to obtain a feature vector with a scale of (32, 32, 64);
thirdly, performing convolution 1 × 1 again to recover the dimension and obtain the feature F with the scale of (32, 128, 64)B(M(Fc))；
Fourth, output characteristic HB(M(Fc) In which H) isB(M(Fc))＝FB(M(Fc))+M(Fc) In this embodiment, the matrix alignment element addition is performed, feature HB(M(Fc) Dimension (32, 128, 64).
That is, the Bottleneck structure is aligned to the feature M (F)c) The feature F was obtained using batch normalization after 1 × 1 convolution followed by successive use of a one-dimensional convolution with a convolution kernel of 3 and a 1 × 1 convolution and a ReLU activation functionB(M(Fc))。
Referring to fig. 4, the FCN performs classification on the data output by the bottleeck using Dropout policy and softmax function, resulting in the final recognition result. In this embodiment, two fully-connected layers are adopted, the number of neurons is 512, and the Dropout coefficient takes 0.2.
The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto, and those skilled in the art should be able to equally replace or change the technical solution and the inventive concept of the present invention in the disclosure of the present invention, and all belong to the protection scope of the present invention.

Claims (8)

1. The face recognition method based on the convolutional neural network and the attention model is characterized by comprising the following steps of:
s1, preprocessing the face image;
s2, inputting the preprocessed image data into a convolutional neural network to extract high-dimensional features to obtain high-dimensional features Fc；
S3, converting the high-dimensional feature FcInputting the data into an attention model, calculating an attention mask through a training method of a neural network, and obtaining an attention distribution characteristic M (F) of a space domain and a channel domainc)；
S4, feature M (F) of attention distributionc) Inputting the feature into a Bottleneck module, and acquiring a feature H (M (F) by using a short mechanismc))；
S5, converting the characteristic H (M (F)c) Input into the full convolution network classification structure, and use Dropout strategy and softmax function to obtain the final face recognition result.
2. The face recognition method based on the convolutional neural network and the attention model as claimed in claim 1, wherein the preprocessing of step S1 comprises: normalizing face image size to
Figure FDA0002452075800000011
The face image is randomly turned left and right, the RGB image is converted into a gray level image, and the tensor of the face image is standardized.
3. The face recognition method based on the convolutional neural network and the attention model as claimed in claim 2, wherein the RGB image is converted into a gray image, specifically:
after the RGB image is converted into the gray image, the single-channel gray image is expanded into an N-channel gray image, and the image conversion calculation formula and the channel expansion formula are as follows:
Figure FDA0002452075800000012
Grayi＝λiGray+biasi
in the formula, R, G and B are eachRepresenting RGB three-channel pixel values of an original image; i denotes the ith gray channel,
Figure FDA0002452075800000013
biasiin order to be a term of the offset,
Figure FDA0002452075800000014
Grayi∈[0,255]。
4. the face recognition method based on convolutional neural network and attention model of claim 3, wherein the number of channels N is 3, λ1、λ2、λ3Respectively taking 0.37, 0.30 and 0.33; biasiIs a value generated randomly.
5. The face recognition method based on the convolutional neural network and the attention model as claimed in claim 2, wherein the tensor of the face image is normalized as follows:
Figure FDA0002452075800000015
in the formula, XiInput data for i channel, E (X)i) Is a mean value, σ (X)i) Is the standard deviation.
6. The face recognition method based on convolutional neural network and attention model of claim 1, wherein the image data inputted in step S2 is normalized by batch after passing through the first max pooling layer, the down-sampling of the picture is performed by max pooling after each convolutional layer, the number of convolutional layers is 5, and the shrunk feature size W × H is the original picture' S feature size
Figure FDA0002452075800000021
High-dimensional feature of output FcThe form is as follows: (C)c,Wc,Hc) (ii) a The mean and variance of the batch normalization were as follows:
Figure FDA0002452075800000022
Figure FDA0002452075800000023
in the formula, WhFor convolutional neural network parameters, xiFor convolutional neural network input, m is the batch size.
7. The face recognition method based on the convolutional neural network and the attention model as claimed in claim 1, wherein the step S3 comprises the steps of:
s31, outputting the size of (C)c,Wc,Hc) Is spliced into a length Cc×Wc×HcIs input vector Vc；
S32, adopting full-connection network structure to correlate input vector VcAnd a length Wc×HcIs output vector VpEstablishing the relationship between each high-dimensional characteristic pixel point and other pixel points and channels;
s33, generating an attention feature map by the convolution output through a sigmoid activation function;
s34, and combining the attention feature map with the high-dimensional feature F output in the step S2cCarrying out tensor multiplication to form attention enhancement on a space domain and a channel domain to obtain the feature M (F) with enhanced attentionc)：
M(Fc)＝ψ(fC×W×H(Contact(Fc)))＝ψ(fC×W×H(Vc))
Where Contact () is the matrix splice, fC×W×H() For feature vector VcThe network computation is performed, ψ () is a tensor multiplication computation of the attention feature map.
8. The face recognition method based on the convolutional neural network and the attention model as claimed in claim 7, wherein the bottleeck module in step S4 is based on one-dimensional convolution, and the process of using the Shortcut mechanism to obtain features comprises:
s41, using 1 × 1 convolution to output the feature M (F)c) Performing dimensionality reduction;
s42, performing zero filling processing on two ends of the feature vector after the dimension reduction, and then performing one-dimensional convolution with a convolution kernel of 3;
s43, carrying out convolution 1 × 1 again, recovering dimensionality and obtaining a feature FB(M(Fc))；
S44, outputting characteristic HB(M(Fc) In which H) isB(M(Fc))＝FB(M(Fc))+M(Fc)。
CN202010295613.8A 2020-04-15 2020-04-15 Face recognition method based on convolutional neural network and attention model Active CN111582044B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010295613.8A CN111582044B (en) 2020-04-15 2020-04-15 Face recognition method based on convolutional neural network and attention model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010295613.8A CN111582044B (en) 2020-04-15 2020-04-15 Face recognition method based on convolutional neural network and attention model

Publications (2)

Publication Number Publication Date
CN111582044A true CN111582044A (en) 2020-08-25
CN111582044B CN111582044B (en) 2023-06-20

Family

ID=72124407

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010295613.8A Active CN111582044B (en) 2020-04-15 2020-04-15 Face recognition method based on convolutional neural network and attention model

Country Status (1)

Country Link
CN (1) CN111582044B (en)

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112507995A (en) * 2021-02-05 2021-03-16 成都东方天呈智能科技有限公司 Cross-model face feature vector conversion system and method
CN112766176A (en) * 2021-01-21 2021-05-07 深圳市安软科技股份有限公司 Training method of lightweight convolutional neural network and face attribute recognition method
CN112766229A (en) * 2021-02-08 2021-05-07 南京林业大学 Human face point cloud image intelligent identification system and method based on attention mechanism
CN112801280A (en) * 2021-03-11 2021-05-14 东南大学 One-dimensional convolution position coding method of visual depth self-adaptive neural network
CN112800871A (en) * 2021-01-13 2021-05-14 南京邮电大学 Automatic driving image recognition method based on attention mechanism and relation network
CN113256629A (en) * 2021-07-05 2021-08-13 之江实验室 Image calibration error detection method and device
CN113469111A (en) * 2021-07-16 2021-10-01 中国银行股份有限公司 Image key point detection method and system, electronic device and storage medium
CN114898363A (en) * 2022-05-26 2022-08-12 浙江大学 Identity recognition method and system based on egg shell image features
WO2024045320A1 (en) * 2022-08-31 2024-03-07 北京龙智数科科技服务有限公司 Facial recognition method and apparatus
CN117894058A (en) * 2024-03-14 2024-04-16 山东远桥信息科技有限公司 Smart city camera face recognition method based on attention enhancement

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109543606A (en) * 2018-11-22 2019-03-29 中山大学 A kind of face identification method that attention mechanism is added
CN110059582A (en) * 2019-03-28 2019-07-26 东南大学 Driving behavior recognition methods based on multiple dimensioned attention convolutional neural networks
CN110427867A (en) * 2019-07-30 2019-11-08 华中科技大学 Human facial expression recognition method and system based on residual error attention mechanism
CN110781784A (en) * 2019-10-18 2020-02-11 高新兴科技集团股份有限公司 Face recognition method, device and equipment based on double-path attention mechanism
CN110781760A (en) * 2019-05-24 2020-02-11 西安电子科技大学 Facial expression recognition method and device based on space attention

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109543606A (en) * 2018-11-22 2019-03-29 中山大学 A kind of face identification method that attention mechanism is added
CN110059582A (en) * 2019-03-28 2019-07-26 东南大学 Driving behavior recognition methods based on multiple dimensioned attention convolutional neural networks
CN110781760A (en) * 2019-05-24 2020-02-11 西安电子科技大学 Facial expression recognition method and device based on space attention
CN110427867A (en) * 2019-07-30 2019-11-08 华中科技大学 Human facial expression recognition method and system based on residual error attention mechanism
CN110781784A (en) * 2019-10-18 2020-02-11 高新兴科技集团股份有限公司 Face recognition method, device and equipment based on double-path attention mechanism

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112800871B (en) * 2021-01-13 2022-08-26 南京邮电大学 Automatic driving image recognition method based on attention mechanism and relation network
CN112800871A (en) * 2021-01-13 2021-05-14 南京邮电大学 Automatic driving image recognition method based on attention mechanism and relation network
CN112766176A (en) * 2021-01-21 2021-05-07 深圳市安软科技股份有限公司 Training method of lightweight convolutional neural network and face attribute recognition method
CN112766176B (en) * 2021-01-21 2023-12-01 深圳市安软科技股份有限公司 Training method of lightweight convolutional neural network and face attribute recognition method
CN112507995A (en) * 2021-02-05 2021-03-16 成都东方天呈智能科技有限公司 Cross-model face feature vector conversion system and method
CN112766229A (en) * 2021-02-08 2021-05-07 南京林业大学 Human face point cloud image intelligent identification system and method based on attention mechanism
CN112801280A (en) * 2021-03-11 2021-05-14 东南大学 One-dimensional convolution position coding method of visual depth self-adaptive neural network
CN113256629A (en) * 2021-07-05 2021-08-13 之江实验室 Image calibration error detection method and device
CN113469111A (en) * 2021-07-16 2021-10-01 中国银行股份有限公司 Image key point detection method and system, electronic device and storage medium
CN114898363B (en) * 2022-05-26 2023-03-24 浙江大学 Identity recognition method and system based on egg shell image features
CN114898363A (en) * 2022-05-26 2022-08-12 浙江大学 Identity recognition method and system based on egg shell image features
WO2024045320A1 (en) * 2022-08-31 2024-03-07 北京龙智数科科技服务有限公司 Facial recognition method and apparatus
CN117894058A (en) * 2024-03-14 2024-04-16 山东远桥信息科技有限公司 Smart city camera face recognition method based on attention enhancement
CN117894058B (en) * 2024-03-14 2024-05-24 山东远桥信息科技有限公司 Smart city camera face recognition method based on attention enhancement

Also Published As

Publication number Publication date
CN111582044B (en) 2023-06-20

Similar Documents

Publication Publication Date Title
CN111582044A (en) Face recognition method based on convolutional neural network and attention model
CN112307958B (en) Micro-expression recognition method based on space-time appearance motion attention network
CN111639692B (en) Shadow detection method based on attention mechanism
WO2021042828A1 (en) Neural network model compression method and apparatus, and storage medium and chip
CN112926396B (en) Action identification method based on double-current convolution attention
CN111950649B (en) Attention mechanism and capsule network-based low-illumination image classification method
CN112288011B (en) Image matching method based on self-attention deep neural network
CN111444881A (en) Fake face video detection method and device
CN110163286B (en) Hybrid pooling-based domain adaptive image classification method
CN112232184B (en) Multi-angle face recognition method based on deep learning and space conversion network
CN111476806B (en) Image processing method, image processing device, computer equipment and storage medium
CN111368672A (en) Construction method and device for genetic disease facial recognition model
CN111709313B (en) Pedestrian re-identification method based on local and channel combination characteristics
CN114549913B (en) Semantic segmentation method and device, computer equipment and storage medium
CN112733665B (en) Face recognition method and system based on lightweight network structure design
CN113780249B (en) Expression recognition model processing method, device, equipment, medium and program product
Rao et al. Exploring deep learning techniques for Kannada handwritten character recognition: a boon for digitization
CN111899203A (en) Real image generation method based on label graph under unsupervised training and storage medium
CN115601820A (en) Face fake image detection method, device, terminal and storage medium
CN110188646B (en) Human ear identification method based on fusion of gradient direction histogram and local binary pattern
CN110135435B (en) Saliency detection method and device based on breadth learning system
CN111444957B (en) Image data processing method, device, computer equipment and storage medium
KR20180092453A (en) Face recognition method Using convolutional neural network and stereo image
CN116311345A (en) Transformer-based pedestrian shielding re-recognition method
Pei et al. FGO-Net: Feature and Gaussian Optimization Network for visual saliency prediction

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant