CN111553277B - Chinese signature identification method and terminal introducing consistency constraint - Google Patents
Chinese signature identification method and terminal introducing consistency constraint Download PDFInfo
- Publication number
- CN111553277B CN111553277B CN202010350217.0A CN202010350217A CN111553277B CN 111553277 B CN111553277 B CN 111553277B CN 202010350217 A CN202010350217 A CN 202010350217A CN 111553277 B CN111553277 B CN 111553277B
- Authority
- CN
- China
- Prior art keywords
- signature
- chinese
- training
- loss function
- opt0
- 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
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000012549 training Methods 0.000 claims abstract description 59
- 230000006870 function Effects 0.000 claims abstract description 48
- 238000012795 verification Methods 0.000 claims abstract description 19
- 238000012360 testing method Methods 0.000 claims description 2
- 239000000284 extract Substances 0.000 abstract description 3
- 238000011160 research Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
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
- G06V40/30—Writer recognition; Reading and verifying signatures
- G06V40/33—Writer recognition; Reading and verifying signatures based only on signature image, e.g. static signature recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Collating Specific Patterns (AREA)
Abstract
The invention belongs to the field of supervised learning, and particularly discloses a Chinese signature identification method and a terminal introducing consistency constraint, wherein the method utilizes a loss function to train a Chinese signature identification full convolution network model, and the loss function L comprises a Focal loss function; the loss function L also includes a consistency constraint loss function used in conjunction with the Focal loss function. The invention creatively applies consistency loss in the full supervision field of Chinese offline signature authentication, firstly, a first group of training samples are relied to make correct prediction on verification signatures, and accordingly, verification result information is transmitted to corresponding amplified samples through the consistency loss, and more overturned samples can be correctly predicted along with the lengthening of model training time, which also shows that the model reduces the interference of background information and extracts more handwriting information characteristics, thereby improving the difficulty of handwriting information sparseness and improving the signature authentication rate.
Description
Technical Field
The invention relates to the field of supervised learning, in particular to a Chinese signature identification method and a terminal introducing consistency constraint.
Background
Since the last 60 s, foreign researchers began research related to the field of signature authentication. Compared with the foreign research state, the research on signature identification in China starts late, although the development is rapid, the identification rate still needs to be further improved, particularly in the field of Chinese signature identification, because a relatively authoritative public Chinese data set is lacked, the related research on Chinese signature identification is often carried out on a small data set made by individuals, and the research progress of China in the field of off-line signature identification is severely limited. It should be noted that, because the structure of chinese characters is unique and the signature characteristics have their particularity compared with the english language system, most of the effective methods for identifying chinese characters in english signatures cannot be directly transferred to the identification of chinese characters, which has been proved in some documents, but a series of foreign researches still have important reference and reference values.
In the prior art, the identification effect is good when the offline signature identification is carried out by utilizing the Focal loss function. However, in the experimental process, the fact that the strokes of the Chinese signature are relatively thin and long leads to that most of the area of the image is background information, so that the information of the off-line signature is relatively sparse, and the identification effect of the existing algorithm is greatly influenced.
Disclosure of Invention
The invention mainly solves the technical problem of providing a signature identification method which can effectively identify off-line signatures.
In order to solve the technical problems, the invention adopts a technical scheme that: a Chinese signature authentication method introducing consistency constraint is provided. The method comprises the following steps:
training a Chinese signature authentication full convolution network model by using a loss function, wherein the loss function L comprises a Focal loss function; the method is characterized in that:
the loss function L also includes a consistency constraint loss function used in conjunction with the Focal loss function.
Before the training of the full convolution network model for Chinese signature authentication by using the loss function, the method further comprises the following steps:
preparing a Chinese signature identification data set, constructing a Chinese signature identification full convolution network model, and selecting a training signature sample according to the Chinese signature identification data set; the chinese signature authentication dataset includes a reference signature and a verification signature, and the training signature samples include a plurality of training signature sample sets.
The training of the Chinese signature authentication full convolution network model by using the loss function L comprises the following steps:
inputting each training signature sample group into the Chinese signature identification full convolution network model respectively, and performing forward propagation to obtain output results, wherein the output results corresponding to each training signature sample group are marked as opt0, opt1,. and.optx respectively;
and performing one-time back propagation on all output results by using the loss function L to update the full convolution network model for the Chinese signature authentication.
The training signature sample comprises:
a first training signature sample set comprising the reference signature and a verification signature;
preferably, the training signature sample further comprises:
a second training signature sample set and/or a third training signature sample set;
the second training signature sample set comprises a reference signature and the verification signature for flipping foreground and background; the third training signature sample set includes a reference signature and a verification signature that flips the foreground and background.
The foreground refers to the handwriting part of the signature, and the background refers to the part without handwriting
Preferably, the loss function L is calculated by the formula:
L=FL(y,opt0)+m1|opt0-opt1|+m2|opt0-opt2|+...+mx|opt0-optx|;
where FL (y, opt0) represents the Focal loss function, m1|opt0-opt1|+m2|opt0-opt2|+...+mxI opt0-opt | represents the consistency constraint loss function, m1~mxRespectively representing weights of corresponding training signature sample groups; y represents the actual value of the sample specified.
After the training of the Chinese signature authentication full convolution network model by using the loss function, the method further comprises the following steps:
and authenticating the newly input Chinese signature and/or testing a Chinese signature authentication full convolution network model.
Preferably, when the chinese signature is authenticated, the chinese signature to be authenticated and N reference signatures are respectively combined to form N sets of inputs, the N sets of inputs are input into the chinese signature authentication full convolution network model, N determination results are output, and an average value of the N determination results is counted.
Preferably, the authenticating the chinese signature further includes setting a determination threshold, and when an average value of the N determination results is greater than the determination threshold, the chinese signature is a true signature.
Preferably, a terminal comprises a memory and a processor, the memory stores computer instructions executable on the processor, and the processor executes the computer instructions to perform the steps of the method for identifying chinese signatures with consistency constraints.
The invention has the beneficial effects that:
(1) the method introduces consistency constraint loss on the basis of the Focal loss function, reduces the interference of background information, extracts more handwriting information characteristics and further improves the difficulty of handwriting information sparseness.
(2) Data are augmented by creatively using a mode of turning over the foreground and the background to obtain a plurality of groups of training signature samples, and then the identification rate of signature identification is improved.
Drawings
FIG. 1 is a schematic diagram of an embodiment of the present invention;
Detailed Description
In order to make the technical scheme, the purpose and the beneficial effects of the invention clearer, the invention is further explained by combining the embodiment and the attached drawings.
In an exemplary embodiment, a method for Chinese signature authentication incorporating a consistency constraint includes the steps of:
training a Chinese signature authentication full convolution network model using loss functions, including Focal loss functions.
The loss function also includes a consistency constraint loss function used in conjunction with the Focal loss function.
In the prior art, when offline signature authentication is performed only by using a Focal loss function, because the strokes of the Chinese signature are relatively thin and long, most of the area of an image is background information, the offline signature information is relatively sparse, which greatly influences the identification effect of the conventional algorithm.
And the method creatively utilizes a mode of turning over the foreground and the background to perform data amplification to obtain three groups of training signature samples, and the model transmits the original sample information to the turned samples by minimizing consistency constraint loss.
The model firstly depends on the first group of training samples to make correct prediction on the verification signature, and accordingly, the verification result information is transmitted to the corresponding amplified samples through consistency loss, and more turned samples can be correctly predicted along with the lengthening of the model training time, which also shows that the model reduces the interference of background information and extracts more handwriting information characteristics, thereby improving the difficulty of handwriting information sparseness.
Further, before the training of the full convolution network model for Chinese signature authentication by using the loss function, the method further comprises the following steps:
preparing a Chinese signature identification data set, constructing a Chinese signature identification full convolution network model, and selecting a training signature sample according to the Chinese signature identification data set; the chinese signature authentication dataset includes a reference signature and a verification signature, and the training signature samples include a plurality of training signature sample sets.
The reference signature refers to a collected real signature, and the verification signature refers to a collected similar signature and is also a signature to be verified.
Further, as shown in fig. 1, the training of the full convolution network model for chinese signature authentication by using a loss function L includes:
inputting each training signature sample group into the Chinese signature identification full convolution network model respectively, and performing forward propagation to obtain output results, wherein the output results corresponding to each training signature sample group are marked as opt0, opt1,. and.optx respectively;
as shown in fig. 1, 3 training signature sample sets were selected in this example.
And performing one-time back propagation on all output results by using the loss function L to update the full convolution network model for the Chinese signature authentication.
Further, the selecting the training signature sample includes:
a first training signature sample set comprising the reference signature and a verification signature;
further, the training signature sample further comprises:
a second training signature sample set and/or a third training signature sample set;
the second training signature sample set comprises a reference signature and the verification signature for flipping foreground and background; the third training signature sample set includes a reference signature and a verification signature that flips the foreground and background.
The foreground refers to the handwriting portion of the signature and the background refers to the portion that does not contain the handwriting.
Further, the loss function L is calculated by the formula:
L=FL(y,opt0)+m1|opt0-opt1|+m2|opt0-opt2|+...+mx|opt0-optx|;
wherein FL (y, opt0) represents the Focal loss function,
FL(y,opt0)=-at(1-opt0)γlog(opt0),
atthe weight factor for adjusting the proportion of positive and negative samples in the Focal local, and gamma is the Focal LossAnd the adjustable parameters of the difficult and easy samples are distinguished in the lost function.
m1|opt0-opt1|+m2|opt0-opt2|+...+mxI opt0-opt | represents the consistency constraint loss function, m1~mxWeighting parameters respectively representing the corresponding training signature sample groups as balance consistency constraint loss and FocalLoss loss are obtained through network training; y represents the actual value of the sample specified.
Further, using Adam as an optimizer, the sample size of each batch is 128, the initial learning rate of the model is set to 0.001, the learning rate of five thousand batches per iteration is 0.1 times the current learning rate, and when the learning rate is reduced to 0.00001, the learning rate is not further reduced.
Further, the consistency constraint loss constrains the distance of opt0 from opt1 and opt2, respectively, and the model passes the original sample information to the flipped sample by minimizing the consistency constraint loss.
After the Chinese signature authentication full convolution network model is trained by using the loss function, the Chinese signature is authenticated.
Further, when the Chinese signature is authenticated, the Chinese signature to be authenticated is combined with N reference signatures to form N groups of inputs, the N groups of inputs are input into the Chinese signature authentication full convolution network, N judgment results are output, and the average value of the N judgment results is counted.
Further, when the Chinese signature is identified, firstly, the correct prediction is made on the Chinese signature by relying on the first group of training samples, and the verification result information is transmitted to the corresponding amplified sample.
Further, the formula for counting the average value of the N determination results is: wherein R isuFor the reference signature set of user u, P (u | R)uS) is the summary that the authentication signature S belongs to the user uThe ratio of the total weight of the particles,is the output of the convolutional neural network model.
Further, the authentication of the chinese signature further includes setting a determination threshold, where T is 0.5, if P (u | R)u,S)>And T, the Chinese signature is a real signature, otherwise, the Chinese signature is a fake signature.
Further, the Chinese signature authentication system introducing the consistency constraint comprises a memory and a processor, wherein the memory module stores the reference signature and the verification signature, and the processor realizes the steps of the Chinese signature authentication method introducing the consistency constraint.
The above description is only an embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes performed by the present specification and drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (8)
1. A Chinese signature identification method introducing consistency constraint comprises the following steps:
training a Chinese signature authentication full convolution network model by using a loss function, wherein the loss function L comprises a Focal loss function; the method is characterized in that:
the loss function L further comprises a consistency constraint loss function used in conjunction with the Focal loss function;
before the training of the Chinese signature authentication full convolution network model by using the loss function, the method further comprises the following steps:
preparing a Chinese signature identification data set, constructing a Chinese signature identification full convolution network model, and selecting a training signature sample according to the Chinese signature identification data set; the Chinese signature authentication dataset comprises a reference signature and a verification signature, and the training signature sample comprises a plurality of training signature sample groups;
the training of the Chinese signature authentication full convolution network model by using the loss function L comprises the following steps:
inputting each training signature sample group into the Chinese signature identification full convolution network model respectively, and performing forward propagation to obtain output results, wherein the output results corresponding to each training signature sample group are marked as opt0, opt1,. and.optx respectively;
and performing one-time back propagation on all output results by using the loss function L to update the full convolution network model for the Chinese signature authentication.
2. The method of claim 1, wherein the method comprises: the training signature sample comprises:
a first training signature sample set including the reference signature and a verification signature.
3. The method of claim 2, wherein the method comprises: the training signature samples further comprise:
a second training signature sample set and/or a third training signature sample set;
the second training signature sample set comprises a reference signature and the verification signature for flipping foreground and background; the third training signature sample group comprises a reference signature and a verification signature for turning over a foreground and a background;
the foreground refers to the handwriting portion of the signature and the background refers to the portion that does not contain the handwriting.
4. The method of claim 1, wherein the method comprises: the loss function L is calculated by the formula:
L=FL(y,opt0)+m1|opt0-opt1|+m2|opt0-opt2|+...+mx|opt0-optx|;
where FL (y, opt0) represents the Focal loss function, m1|opt0-opt1|+m2|opt0-opt2|+...+mxI opt0-opt | represents the consistency constraint loss function, m1~mxRespectively generation by generationThe table corresponds to the weights of the training signature sample set; y represents the actual value of the sample specified.
5. The method of claim 1, wherein the method comprises: after the training of the Chinese signature authentication full convolution network model by using the loss function, the method further comprises the following steps:
and authenticating the newly input Chinese signature and/or testing a Chinese signature authentication full convolution network model.
6. The method of claim 5, wherein the method comprises: and when the Chinese signature is identified, inputting the Chinese signature to be identified into the Chinese signature identification full convolution network model, outputting N judgment results, and counting the average value of the N judgment results.
7. The method of claim 6, wherein the method comprises: and when the Chinese signature is identified, setting a judgment threshold value, and when the average value of the N judgment results is greater than the judgment threshold value, the Chinese signature is a real signature.
8. A terminal comprising a memory and a processor, the memory having stored thereon computer instructions executable on the processor, the terminal comprising: the computer instructions when executed by the processor perform the steps of a method for chinese signature authentication incorporating a consistency constraint according to any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010350217.0A CN111553277B (en) | 2020-04-28 | 2020-04-28 | Chinese signature identification method and terminal introducing consistency constraint |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010350217.0A CN111553277B (en) | 2020-04-28 | 2020-04-28 | Chinese signature identification method and terminal introducing consistency constraint |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111553277A CN111553277A (en) | 2020-08-18 |
CN111553277B true CN111553277B (en) | 2022-04-26 |
Family
ID=72003282
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010350217.0A Active CN111553277B (en) | 2020-04-28 | 2020-04-28 | Chinese signature identification method and terminal introducing consistency constraint |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111553277B (en) |
Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108108351A (en) * | 2017-12-05 | 2018-06-01 | 华南理工大学 | A kind of text sentiment classification method based on deep learning built-up pattern |
CN108391092A (en) * | 2018-03-21 | 2018-08-10 | 四川弘和通讯有限公司 | Danger identifying system based on deep learning |
CN108764202A (en) * | 2018-06-06 | 2018-11-06 | 平安科技(深圳)有限公司 | Airport method for recognizing impurities, device, computer equipment and storage medium |
CN109190579A (en) * | 2018-09-14 | 2019-01-11 | 大连交通大学 | A kind of handwriting signature identification method of the production confrontation network SIGAN based on paired-associate learning |
CN109215085A (en) * | 2018-08-23 | 2019-01-15 | 上海小萌科技有限公司 | A kind of article statistic algorithm using computer vision and image recognition |
CN109409254A (en) * | 2018-10-10 | 2019-03-01 | 成都优易数据有限公司 | A kind of electronic contract handwritten signature identification method based on twin neural network |
CN110097107A (en) * | 2019-04-23 | 2019-08-06 | 安徽大学 | Alternaria mali roberts disease recognition and classification method based on convolutional neural networks |
CN110096977A (en) * | 2019-04-18 | 2019-08-06 | 中金金融认证中心有限公司 | The training method and handwriting verification method, equipment and medium of handwriting verification model |
CN110222728A (en) * | 2019-05-15 | 2019-09-10 | 图灵深视(南京)科技有限公司 | The training method of article discrimination model, system and article discrimination method, equipment |
CN110348384A (en) * | 2019-07-12 | 2019-10-18 | 沈阳理工大学 | A kind of Small object vehicle attribute recognition methods based on Fusion Features |
CN110556129A (en) * | 2019-09-09 | 2019-12-10 | 北京大学深圳研究生院 | Bimodal emotion recognition model training method and bimodal emotion recognition method |
CN110986949A (en) * | 2019-12-04 | 2020-04-10 | 日照职业技术学院 | Path identification method based on artificial intelligence platform |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10970564B2 (en) * | 2017-09-30 | 2021-04-06 | Tusimple, Inc. | System and method for instance-level lane detection for autonomous vehicle control |
US10565707B2 (en) * | 2017-11-02 | 2020-02-18 | Siemens Healthcare Gmbh | 3D anisotropic hybrid network: transferring convolutional features from 2D images to 3D anisotropic volumes |
-
2020
- 2020-04-28 CN CN202010350217.0A patent/CN111553277B/en active Active
Patent Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108108351A (en) * | 2017-12-05 | 2018-06-01 | 华南理工大学 | A kind of text sentiment classification method based on deep learning built-up pattern |
CN108391092A (en) * | 2018-03-21 | 2018-08-10 | 四川弘和通讯有限公司 | Danger identifying system based on deep learning |
CN108764202A (en) * | 2018-06-06 | 2018-11-06 | 平安科技(深圳)有限公司 | Airport method for recognizing impurities, device, computer equipment and storage medium |
CN109215085A (en) * | 2018-08-23 | 2019-01-15 | 上海小萌科技有限公司 | A kind of article statistic algorithm using computer vision and image recognition |
CN109190579A (en) * | 2018-09-14 | 2019-01-11 | 大连交通大学 | A kind of handwriting signature identification method of the production confrontation network SIGAN based on paired-associate learning |
CN109409254A (en) * | 2018-10-10 | 2019-03-01 | 成都优易数据有限公司 | A kind of electronic contract handwritten signature identification method based on twin neural network |
CN110096977A (en) * | 2019-04-18 | 2019-08-06 | 中金金融认证中心有限公司 | The training method and handwriting verification method, equipment and medium of handwriting verification model |
CN110097107A (en) * | 2019-04-23 | 2019-08-06 | 安徽大学 | Alternaria mali roberts disease recognition and classification method based on convolutional neural networks |
CN110222728A (en) * | 2019-05-15 | 2019-09-10 | 图灵深视(南京)科技有限公司 | The training method of article discrimination model, system and article discrimination method, equipment |
CN110348384A (en) * | 2019-07-12 | 2019-10-18 | 沈阳理工大学 | A kind of Small object vehicle attribute recognition methods based on Fusion Features |
CN110556129A (en) * | 2019-09-09 | 2019-12-10 | 北京大学深圳研究生院 | Bimodal emotion recognition model training method and bimodal emotion recognition method |
CN110986949A (en) * | 2019-12-04 | 2020-04-10 | 日照职业技术学院 | Path identification method based on artificial intelligence platform |
Non-Patent Citations (2)
Title |
---|
不均衡数据集下基于生成对抗网络的改进深度模型故障识别研究;包萍等;《电子测量与仪器学报》;20190315;第33卷(第3期);176-183 * |
基于尺度乘积的X射线焊缝区域提取算法研究;匡平等;《电子科技大学学报》;20150930;第44卷(第5期);737-742 * |
Also Published As
Publication number | Publication date |
---|---|
CN111553277A (en) | 2020-08-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2021135910A1 (en) | Machine reading comprehension-based information extraction method and related device | |
CN108564129B (en) | Trajectory data classification method based on generation countermeasure network | |
CN104966097B (en) | A kind of complex script recognition methods based on deep learning | |
CN110458187B (en) | Malicious code family clustering method and system | |
CN110021051A (en) | One kind passing through text Conrad object image generation method based on confrontation network is generated | |
CN109344759A (en) | A kind of relatives' recognition methods based on angle loss neural network | |
CN108320374B (en) | A kind of multinational paper money number character identifying method based on finger image | |
CN104849650B (en) | One kind is based on improved analog-circuit fault diagnosis method | |
CN102201236A (en) | Speaker recognition method combining Gaussian mixture model and quantum neural network | |
WO2019232861A1 (en) | Handwriting model training method and apparatus, text recognition method and apparatus, and device and medium | |
CN109255339B (en) | Classification method based on self-adaptive deep forest human gait energy map | |
CN110751183A (en) | Image data classification model generation method, image data classification method and device | |
CN109859222A (en) | Edge extracting method and system based on cascade neural network | |
CN114241564A (en) | Facial expression recognition method based on inter-class difference strengthening network | |
CN109842614B (en) | Network intrusion detection method based on data mining | |
CN115659823A (en) | Wing profile aerodynamic coefficient prediction method based on attribute reduction, electronic device and storage medium | |
CN111553277B (en) | Chinese signature identification method and terminal introducing consistency constraint | |
CN113850013B (en) | Ship radiation noise classification method | |
CN106909881A (en) | The method and system of corn breeding base ridge number are extracted based on unmanned aerial vehicle remote sensing images | |
WO2021227294A1 (en) | Identity recognition method based on expert feedback mechanism | |
CN113505692B (en) | Face recognition method based on partial area optimization under working characteristic curve of subject | |
CN110378414A (en) | The personal identification method of multi-modal biological characteristic fusion based on evolution strategy | |
Lu et al. | Cost-sensitive neural network classifiers for postcode recognition | |
CN111177370A (en) | Algorithm for natural language processing | |
CN113034473A (en) | Lung inflammation image target detection method based on Tiny-YOLOv3 |
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 |