CN110909614B - Method for using self-coding network for fingerprint gender classification - Google Patents
Method for using self-coding network for fingerprint gender classification Download PDFInfo
- Publication number
- CN110909614B CN110909614B CN201911030724.XA CN201911030724A CN110909614B CN 110909614 B CN110909614 B CN 110909614B CN 201911030724 A CN201911030724 A CN 201911030724A CN 110909614 B CN110909614 B CN 110909614B
- Authority
- CN
- China
- Prior art keywords
- network
- fingerprint
- fingerprint image
- feature extraction
- gender
- 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.)
- Active
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
- G06V40/1347—Preprocessing; Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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
-
- 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
Abstract
The invention discloses a method for using a self-coding network for fingerprint gender classification, which specifically comprises the following steps: firstly, data acquisition and data preprocessing of a fingerprint image are carried out; then inputting the collected fingerprint image into a feature extraction network to extract a feature vector; then, performing up-sampling processing on the extracted feature vector of the fingerprint image to obtain an image of a predicted fingerprint; finally, classifying the gender of the image with the predicted fingerprint by using a classifier; compared with the traditional machine learning methods such as dictionary learning, wavelet analysis and the like, the method has higher accuracy, and has the advantages of smaller data volume requirement, easier data acquisition and the like compared with a method for integrating gender classification of human faces and fingerprint images.
Description
Technical Field
The invention belongs to the technical field of biological identification, and relates to a method for using a self-coding network for fingerprint gender classification.
Background
The finger has abundant characteristics, and the fingerprint is taken as the main biological characteristic of the finger and has several ideal characteristics of universality, significance, permanence, collectability, acceptability, avoidance resistance and the like; the existing method for integrating gender classification of the face and the fingerprint images cannot collect the face and the fingerprint images at the same time, so that gender classification can be carried out only by adopting the fingerprint images.
The fingerprint image contains a large number of features to classify the gender, and if the fingerprint image is put into a traditional convolution neural network to carry out deep learning, the phenomenon that the gradient disappears along with the increase of the network depth exists; for the present time, there is no method of applying a self-encoding network to fingerprint image recognition of gender. Aiming at the problem, a method for applying a self-coding network to fingerprint image gender classification is specially designed by a team, has higher accuracy compared with the traditional machine learning methods such as dictionary learning, wavelet analysis and the like, and has the advantages of smaller data volume requirement, easier data acquisition and the like compared with a method for integrating gender classification of human faces and fingerprint images.
Disclosure of Invention
The invention aims to provide a method for using a self-coding network for fingerprint gender classification, which improves the accuracy of fingerprint image gender classification.
The technical scheme adopted by the invention is that a method for using a self-coding network for fingerprint gender classification comprises the following specific steps:
step 1, data acquisition and data preprocessing of a fingerprint image;
step 2, inputting the fingerprint image collected in the step 1 into a feature extraction network to extract a feature vector;
step 3, performing up-sampling processing on the feature vector of the fingerprint image extracted in the step 2 to obtain an image of the predicted fingerprint;
and 4, performing gender classification on the image of the predicted fingerprint obtained in the step 3 by using a classifier.
The invention is also characterized in that:
wherein the step 1 specifically comprises the following steps:
step 1.1, 2400 fingerprint images with different genders are collected and the gender information of the fingerprint images is marked to be used as a data set of a supervised learning and self-coding network;
step 1.2, carrying out standardization operation on the fingerprint image acquired in the step 1.1 to standardize the fingerprint image data within a range of taking 0 as a mean value and 1 as a variance;
step 1.3, carrying out the operation of standardizing the size of the picture on the fingerprint image subjected to the standardization operation in the step 1.2 by utilizing a bilinear interpolation algorithm;
the processing process of the fingerprint image standardized operation in the step 1.2 comprises the following steps: firstly, calculating the mean value and the variance of a matrix corresponding to the fingerprint image, and then, subtracting the mean value from each number in the matrix to divide the square difference;
wherein the step 2 specifically comprises the following steps:
step 2.1, building a feature extraction network through a deep learning algorithm;
step 2.2, inputting the fingerprint image preprocessed in the step 1 into the feature extraction network built in the step 2.1 for feature extraction, and optimizing parameters of the feature extraction network;
step 2.3, saving the optimized feature extraction network model as an extraction model for outputting the fingerprint image features;
the method for constructing the feature extraction network in the step 2.1 adopts convolution integral, a cavity convolution algorithm and a residual error network module, and specifically comprises the following steps:
the feature extraction network, namely the coding network, consists of four parts, wherein the feature extraction network consists of 16 residual modules, the operation steps of each module are the same, and the modules sequentially comprise a cavity convolution, an excitation function, a cavity convolution integral and an excitation function, wherein the first part, the second part, the third part and the fourth part consist of 3 residual modules, 4 residual modules, 6 residual modules and 3 residual modules respectively;
wherein the step 3 specifically comprises the following steps:
step 3.1, building an up-sampling network through a deep learning algorithm;
step 3.2, inputting the feature vector of the fingerprint image extracted in the step 2 into the up-sampling network established in the step 3.1 to output and predict the fingerprint image, and adjusting and optimizing the parameters in the feature extraction network stored in the step 2.3;
step 3.3, saving the feature extraction network model optimized in the step 3.2;
in step 3.1, the up-sampling network, namely the decoding network, comprises 8 modules, wherein each module comprises convolution integral, an activation function and deconvolution operation;
the optimization process is to utilize an Adam optimizer to minimize a loss value obtained by a squared error loss function and adjust the parameters of the whole network in a back propagation mode, wherein the Adam optimizer comprises a method of gradient descent and momentum acceleration convergence, namely, a cost function and the gradient descent are utilized to optimize the parameters in an encoding network and a decoding network, and the formula of the cost function is as follows:
loss = ∑ (output-input) 2 /2 (1)；
Wherein the step 4 specifically comprises the following steps:
step 4.1, building a gender classifier through a deep learning algorithm;
step 4.2, inputting the fingerprint image into the optimized feature extraction network model obtained in the step 3.3 to obtain a feature vector of the fingerprint image, and learning the gender classifier, so that the accuracy of the classifier is improved;
step 4.3, inputting another group of feature vectors which do not have intersection with the fingerprint image training set used in the step 4.2 into a classifier to predict the fingerprint image;
wherein the step 4.2 specifically comprises: inputting the fingerprint image into the optimized feature extraction network model mentioned in the step 3.3 to obtain a feature vector of the fingerprint image, finally extracting the feature distribution condition of the biological information features under the biological information label through a loss function and an optimizer, feeding back the classification result of the biological information features, further adjusting the parameters of the network, and improving the accuracy of the fingerprint image gender classifier.
The invention has the beneficial effects that:
compared with the traditional machine learning methods such as dictionary learning, wavelet analysis and the like, the method for using the self-coding network for fingerprint gender classification has the advantages of smaller data volume requirement, easier data acquisition and the like.
Drawings
FIG. 1 is a flow chart of a method of using a self-encoding network for fingerprint gender classification in accordance with the present invention;
FIG. 2 is a diagram of a self-coding network structure in a method for using the self-coding network for fingerprint gender classification according to the present invention;
fig. 3 is a flow chart of the self-coding network in the method for using the self-coding network for fingerprint gender classification of the invention.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
The invention provides a method for using a self-coding network for fingerprint gender classification, which specifically comprises the following steps as shown in figure 1:
step 1, data acquisition and data preprocessing:
step 1.1, 2400 fingerprint images with different genders are collected and marked with the gender information to serve as a data set of a supervised learning and self-coding network.
Step 1.2, carrying out standardization operation on the fingerprint image acquired in the step 1.1, standardizing data in a range with 0 as a mean value and 1 as a variance, wherein the processing process comprises the steps of firstly calculating the mean value and the variance of a matrix corresponding to the fingerprint image, and then subtracting the mean value from each number to divide the variance, so that the operation can ensure that the convergence of a network is accelerated during training;
step 1.3, the fingerprint image after the standardization operation is subjected to the operation of standardizing the size of the picture by utilizing a bilinear interpolation algorithm, namely the shape of a matrix is changed into a standardized image under the principle that the characteristic quantity and the characteristic structure of effective information contained in the fingerprint image are not changed;
step 2, inputting the fingerprint image collected in the step 1 into a feature extraction network for extracting feature vectors:
step 2.1, a feature extraction network for extracting features of the fingerprint image is built through a deep learning algorithm, the existing several feature extraction networks are adopted for extracting the features, and a convolution integral and cavity convolution algorithm and a residual error network module are used:
the specific feature extraction network, i.e., the coding network, is as follows:
the coding network consists of four parts, wherein the coding network consists of 16 residual modules, and the operation steps of each module are identical, namely, a cavity convolution, an excitation function, a convolution integral and an excitation function. The first, second, third and fourth parts are respectively composed of 3, 4, 6 and 3 residual modules; the specific increase of the number of channels and the compression process of the fingerprint picture size are shown in fig. 3, wherein the increase of the number of channels to be described occurs at the second-time hole convolution operation of the first residual module of each part, the number of output channels of the hole convolution is the number marked by fig. 3 as the number of input channels, and the compression process of the fingerprint picture size is that the step length of the first-time hole convolution of the first residual module of each part is set to be 2;
the effect of the hole convolution is to enlarge the receptive field under the condition of losing information without pooling operation, so that each convolution output contains information in a larger range; the module of the residual error network avoids the phenomenon of gradient explosion or gradient disappearance caused by the network at a deeper layer;
2.2, inputting the fingerprint image preprocessed in the step 1 into a coding network for a feature extraction step (namely downsampling), extracting biological information features through convolution integral and void convolution, and optimizing parameters of the feature extraction network;
step 2.3, saving the optimized feature extraction network model as an extraction model for outputting the fingerprint image features;
step 3, utilizing the feature vector of the fingerprint image obtained in the step 2 to perform up-sampling to obtain a predicted fingerprint image:
step 3.1, performing up-sampling network operation on the fingerprint image features through deep learning algorithm construction, wherein the function of the up-sampling network operation is to predict the fingerprint image according to the obtained feature vector so as to calculate a loss function;
the specific up-sampling network, i.e. the decoding network, is expressed as follows:
the decoding network consists of 8 modules, wherein each module comprises convolution integral, activation function and deconvolution operation. The convolution integral input channel number is the same as the output channel number and the step length is 1, so the convolution integral operation does not change the size and the channel number of the picture. The step size of all deconvolution operations of 8 modules is 2, so the size of 1 × 1 can be changed to 256 × 256 through 8 operations; the number of input and output channels of the 8 times are shown in table 1 below.
TABLE 1 number of output channels
Thus, the length of 512 vectors can be changed into 256 × 256 output fingerprint images by the decoding network;
step 3.2, inputting the feature vector obtained by the feature extraction network into the up-sampling network created in the step 3.1 to output and predict the fingerprint image, and minimizing the obtained loss function to optimize and adjust the parameters in the feature extraction network stored in the step 2.3;
step 3.3, saving the optimized feature extraction network model as a fingerprint image feature extraction model;
the optimization method is specifically implemented in step 2.2 and step 3.2, and is a method for adjusting the whole network parameters in a back propagation mode by using a loss value obtained by minimizing a squared error loss function by using an Adam optimizer, wherein the Adam optimizer comprises a method for gradient descent and momentum accelerated convergence, the parameter optimization optimizes parameters in an encoding network and a decoding network by using a cost function and the gradient descent, so that a predicted fingerprint image is more fit with an input fingerprint image, and the obtained feature vector comprises effective features of the fingerprint image.
And 4, classifying the genders by using the feature vectors:
step 4.1, building a gender classifier through a deep learning algorithm, wherein the classifier has no special requirement for a classic classifier;
and 4.2, firstly inputting the fingerprint image into the optimized feature extraction network model mentioned in the step 3.3 to obtain a feature vector of the fingerprint image, finally extracting the feature distribution condition of the biological information features under the biological information label through a loss function and an optimizer, feeding back the classification result of the biological information features, further adjusting the parameters of the network, and achieving the purpose of improving the accuracy of gender classification of the fingerprint image. Wherein the cost function has the formula:
loss = ∑ (output-input) 2 /2 (1)
Then, the value of the cost function is reduced by utilizing an optimization function so as to achieve the purpose of optimizing network parameters and improve the accuracy of gender classification; a (c)
And 4.3, finally inputting another group of feature vectors without intersection with the fingerprint image training set used in the step 4.2 into a classifier to predict the fingerprint images to obtain a result.
As shown in fig. 2, the structure of the self-encoding network is shown, which includes two parts, namely an encoding network and a decoding network, and the functions of the two parts are respectively to extract the feature vector of the fingerprint image and output the predicted fingerprint image. The self-coding network has the advantages that the feature vector obtained by optimizing the network parameters contains the effective features of the fingerprint image, and the operation amount of the classifier is reduced when the feature vector is used for gender classification.
As shown in fig. 3, a specific process of the self-coding network is shown, wherein the top half is a coding network aiming at extracting feature vectors, the specific steps are that each hole convolution is followed by an activation function, and the specific diagram three that is different in the number of times of executing the hole convolution in each step is labeled; the obtained characteristic vector is subjected to deconvolution, the size is multiplied by two every time, the number of channels is halved, finally, the characteristic vector is converted into an output fingerprint image of 256X256, finally, the input fingerprint image and the output fingerprint image are compared by a training network, the output fingerprint image is enabled to be more fit with the input fingerprint image, and gender classification is carried out by utilizing the characteristic vector at the moment.
Claims (7)
1. A method for using a self-coding network for fingerprint gender classification is characterized by comprising the following specific steps:
step 1, data acquisition and data preprocessing of a fingerprint image;
step 2, inputting the fingerprint image collected in step 1 into a feature extraction network for extracting feature vectors, and specifically comprising the following steps:
step 2.1, building a feature extraction network through a deep learning algorithm;
step 2.2, inputting the fingerprint image preprocessed in the step 1 into the feature extraction network built in the step 2.1 for feature extraction, and optimizing parameters of the feature extraction network;
step 2.3, saving the optimized feature extraction network model as an extraction model for outputting the fingerprint image features;
step 3, performing up-sampling processing on the feature vector of the fingerprint image extracted in the step 2 to obtain an image of the predicted fingerprint, specifically comprising the following steps:
step 3.1, building an up-sampling network through a deep learning algorithm;
step 3.2, inputting the feature vector of the fingerprint image extracted in the step 2 into the up-sampling network established in the step 3.1 to output and predict the fingerprint image, and adjusting and optimizing the parameters in the feature extraction network stored in the step 2.3;
step 3.3, saving the feature extraction network model optimized in the step 3.2;
and 4, performing gender classification on the image of the predicted fingerprint obtained in the step 3 by using a classifier, specifically comprising the following steps of:
step 4.1, building a gender classifier through a deep learning algorithm;
step 4.2, inputting the fingerprint image into the optimized feature extraction network model obtained in the step 3.3 to obtain a feature vector of the fingerprint image, and learning the gender classifier, so that the accuracy of the classifier is improved;
and 4.3, inputting another group of feature vectors which do not have intersection with the fingerprint image training set used in the step 4.2 into a classifier to predict the fingerprint images.
2. The method for fingerprint gender classification using self-coding network as claimed in claim 1, wherein the step 1 comprises the following steps:
step 1.1, 2400 fingerprint images with different genders are collected and the gender information of the fingerprint images is marked to be used as a data set of a supervised learning and self-coding network;
step 1.2, carrying out standardization operation on the fingerprint image acquired in the step 1.1 to standardize the fingerprint image data within a range of taking 0 as a mean value and 1 as a variance;
and step 1.3, carrying out the operation of standardizing the size of the picture on the fingerprint image subjected to the standardization operation of the step 1.2 by utilizing a bilinear interpolation algorithm.
3. The method for fingerprint gender classification using self-coding network as claimed in claim 2, wherein the step 1.2 is to perform normalization operation on the fingerprint image by: the mean and variance of the matrix corresponding to the fingerprint image are calculated first, and then the mean is subtracted from each number in the matrix to divide the variance.
4. The method for classifying fingerprint gender by using the self-coding network according to claim 1, wherein the step 2.1 of constructing the feature extraction network adopts convolution integral, a hole convolution algorithm and a residual network module, and specifically comprises the following steps:
the feature extraction network, namely the coding network, consists of four parts, wherein the feature extraction network consists of 16 residual modules, the operation steps of each module are the same, and the module sequentially comprises a cavity convolution, an excitation function, a convolution integral and an excitation function, wherein the first part, the second part, the third part and the fourth part respectively consist of 3 residual modules, 4 residual modules, 6 residual modules and 3 residual modules.
5. A method of using self-coding network for fingerprint gender classification as claimed in claim 1 wherein the step 3.1 upsampling network (decoding network) consists of 8 modules, each of which contains convolution integral, activation function and deconvolution operation.
6. The method of claim 1, wherein the optimization process is to utilize an Adam optimizer to minimize a loss value obtained by a square error loss function and adjust the whole network parameters in a back propagation manner, wherein the Adam optimizer includes a gradient descent and momentum acceleration convergence method, that is, a cost function and a gradient descent are utilized to optimize the parameters in the encoding network and the decoding network, and the formula of the cost function is as follows:
loss = ∑ (output-input) 2/2 (1).
7. The method for fingerprint gender classification using self-coding network as claimed in claim 1, wherein the step 4.2 specifically comprises: inputting the fingerprint image into the optimized feature extraction network model mentioned in the step 3.3 to obtain a feature vector of the fingerprint image, finally extracting the feature distribution condition of the biological information features under the biological information label through a loss function and an optimizer, feeding back the classification result of the biological information features, further adjusting the parameters of the network, and improving the accuracy of the fingerprint image gender classifier.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911030724.XA CN110909614B (en) | 2019-10-28 | 2019-10-28 | Method for using self-coding network for fingerprint gender classification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911030724.XA CN110909614B (en) | 2019-10-28 | 2019-10-28 | Method for using self-coding network for fingerprint gender classification |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110909614A CN110909614A (en) | 2020-03-24 |
| CN110909614B true CN110909614B (en) | 2023-04-07 |
Family
ID=69815741
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911030724.XA Active CN110909614B (en) | 2019-10-28 | 2019-10-28 | Method for using self-coding network for fingerprint gender classification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110909614B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102902986A (en) * | 2012-06-13 | 2013-01-30 | 上海汇纳网络信息科技有限公司 | Automatic gender identification system and method |
| CN107657239A (en) * | 2017-09-30 | 2018-02-02 | 清华大学深圳研究生院 | Palmprint image gender classification method and device, computer installation and readable storage medium storing program for executing |
| WO2018028255A1 (en) * | 2016-08-11 | 2018-02-15 | 深圳市未来媒体技术研究院 | Image saliency detection method based on adversarial network |
| CN109508692A (en) * | 2018-11-30 | 2019-03-22 | 深圳大学 | A kind of gender identification method and system based on 3D fingerprint image |
| CN109657567A (en) * | 2018-11-30 | 2019-04-19 | 深圳大学 | A kind of Weakly supervised characteristic analysis method and system based on 3D fingerprint image |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9378406B2 (en) * | 2012-06-15 | 2016-06-28 | Seref Sagirouglu | System for estimating gender from fingerprints |
-
2019
- 2019-10-28 CN CN201911030724.XA patent/CN110909614B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102902986A (en) * | 2012-06-13 | 2013-01-30 | 上海汇纳网络信息科技有限公司 | Automatic gender identification system and method |
| WO2018028255A1 (en) * | 2016-08-11 | 2018-02-15 | 深圳市未来媒体技术研究院 | Image saliency detection method based on adversarial network |
| CN107657239A (en) * | 2017-09-30 | 2018-02-02 | 清华大学深圳研究生院 | Palmprint image gender classification method and device, computer installation and readable storage medium storing program for executing |
| CN109508692A (en) * | 2018-11-30 | 2019-03-22 | 深圳大学 | A kind of gender identification method and system based on 3D fingerprint image |
| CN109657567A (en) * | 2018-11-30 | 2019-04-19 | 深圳大学 | A kind of Weakly supervised characteristic analysis method and system based on 3D fingerprint image |
Non-Patent Citations (2)
| Title |
|---|
| 基于深度卷积稀疏自编码分层网络的人脸识别技术;王金平;《太原理工大学学报》(第05期);全文 * |
| 轻量化多特征融合的指纹分类算法研究;甘俊英等;《信号处理》(第05期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110909614A (en) | 2020-03-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108717568B (en) | A kind of image characteristics extraction and training method based on Three dimensional convolution neural network | |
| CN109522956B (en) | Low-rank discriminant feature subspace learning method | |
| CN101315663B (en) | Nature scene image classification method based on area dormant semantic characteristic | |
| CN108804397B (en) | Chinese character font conversion generation method based on small amount of target fonts | |
| CN110096968A (en) | A kind of ultrahigh speed static gesture identification method based on depth model optimization | |
| CN110348376A (en) | A kind of pedestrian's real-time detection method neural network based | |
| CN106650806A (en) | Cooperative type deep network model method for pedestrian detection | |
| CN110399821B (en) | Customer satisfaction acquisition method based on facial expression recognition | |
| CN110287777B (en) | Golden monkey body segmentation algorithm in natural scene | |
| CN103605972A (en) | Non-restricted environment face verification method based on block depth neural network | |
| CN108985177A (en) | A kind of facial image classification method of the quick low-rank dictionary learning of combination sparse constraint | |
| CN112800876A (en) | Method and system for embedding hypersphere features for re-identification | |
| CN106055653A (en) | Video synopsis object retrieval method based on image semantic annotation | |
| CN104008375A (en) | Integrated human face recognition mehtod based on feature fusion | |
| CN105956570B (en) | Smiling face's recognition methods based on lip feature and deep learning | |
| CN109800768A (en) | The hash character representation learning method of semi-supervised GAN | |
| CN114757864A (en) | Multi-level fine-grained image generation method based on multi-scale feature decoupling | |
| CN110991554B (en) | Improved PCA (principal component analysis) -based deep network image classification method | |
| CN112766283A (en) | Two-phase flow pattern identification method based on multi-scale convolution network | |
| CN110289004B (en) | Artificial synthesis voiceprint detection system and method based on deep learning | |
| CN109034192B (en) | Track-vehicle body vibration state prediction method based on deep learning | |
| CN109947960A (en) | The more attribute Combined estimator model building methods of face based on depth convolution | |
| CN110909614B (en) | Method for using self-coding network for fingerprint gender classification | |
| CN111414511A (en) | Automatic voiceprint modeling and warehousing method, device and equipment | |
| CN112990336B (en) | Deep three-dimensional point cloud classification network construction method based on competitive attention fusion |
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 |