CN110659599A - Scanning test paper-based offline handwriting authentication system and using method thereof - Google Patents
Scanning test paper-based offline handwriting authentication system and using method thereof Download PDFInfo
- Publication number
- CN110659599A CN110659599A CN201910884279.7A CN201910884279A CN110659599A CN 110659599 A CN110659599 A CN 110659599A CN 201910884279 A CN201910884279 A CN 201910884279A CN 110659599 A CN110659599 A CN 110659599A
- Authority
- CN
- China
- Prior art keywords
- image
- handwriting
- test paper
- input
- images
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/32—Digital ink
- G06V30/333—Preprocessing; Feature extraction
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/32—Digital ink
- G06V30/36—Matching; Classification
Abstract
The invention relates to the technical field of artificial intelligence image recognition, and discloses an off-line handwriting identification system based on a scanning test paper, which comprises multi-image tuple input and a two-channel convolution neural network, wherein the multi-image tuple input is used for sampling samples and collecting characteristics, the multi-image tuple is used for training, a plurality of image tuples are randomly extracted from a training set for training, then extracted local characteristics are clustered to form global characteristics, and the two-channel convolution neural network takes two images as a two-channel image input neural network to concurrently extract the characteristics of the two images; the invention can automatically extract the handwriting in the test paper in batch, identify the handwriting and finish the classification of the test paper, thereby saving the labor and time, and having reasonable algorithm, high accuracy and convenient use.
Description
Technical Field
The invention relates to the technical field of artificial intelligence image recognition, in particular to an offline handwriting identification system based on a scanning test paper and a using method thereof.
Background
A great defect of online paper reading at present is that teachers are often required to identify the handwriting of students, for example, when examinees have the same name and cannot rely on other unique identifiers (such as examination numbers), the attribution of test papers can be judged by the experienced teachers through the handwriting of the students. When the number of examinees increases and the number of students is too large, the work is a relatively boring work, and the teacher needs to repeatedly perform the identification, which takes a lot of time and effort. Therefore, the invention adopts the image recognition technology to automatically complete the handwriting authentication in the test paper, and can greatly save the energy and time of teachers.
In the traditional handwriting identification, a handwriting identification mechanism is used for artificially identifying and checking the character layout characteristics, the character writing characteristics and the like, judging the difference between the handwriting and the handwriting, and has long cycle time and large labor consumption. With the development of related technologies of computers in recent years, handwriting identification is more prominent, but most of handwriting identification at present requires writing of a plurality of groups of same characters to judge the difference of character handwriting, a handwriting sample of one person in an actual application scene is easy to obtain, handwriting samples of a plurality of persons writing the same characters are difficult to obtain, online examination is often performed in a large-scale examination at present, and in a scanned test paper, the handwriting of students in the test paper can be obtained in a large amount. Therefore, the invention provides a solution for identifying the handwriting of the student aiming at the scanning test paper.
Disclosure of Invention
Technical problem to be solved
Aiming at the defects of the prior art, the invention provides an offline handwriting identification system based on a scanning test paper and a using method thereof, and solves the problems that the traditional handwriting identification is implemented by manually identifying and checking the character layout characteristics, the character writing characteristics and the like through a handwriting identification mechanism, the difference between the handwriting is judged, the cycle time is long, and a large amount of manpower is consumed.
(II) technical scheme
In order to achieve the purpose, the invention provides the following technical scheme: an off-line handwriting identification system based on a scanning test paper comprises multi-image tuple input and a dual-channel convolution neural network, wherein the multi-image tuple input is used for sampling samples and collecting characteristics, the multi-image tuple is used for training, a plurality of image tuples are randomly extracted from a training set to participate in training, then extracted local characteristics are clustered to form global characteristics, and the dual-channel convolution neural network takes two images as a picture input neural network with dual channels to extract the characteristics of the two images concurrently.
Preferably, in the scheme of multi-component image input, each image is used as a component, 20 components are simultaneously input, features can be extracted from a plurality of handwritten character images, and specific features in characters are reduced, in the first step, 20 character image blocks are randomly extracted to form 20 component images, in the second step, each image of 20 components is input into a CNN local feature extractor, so that the CNN can extract specific features irrelevant to the texts, in the third step, the extracted local features are aggregated by a global feature aggregator in an average value mode, through global features extracted from a plurality of handwriting images, specific features independent of text authors, such as structure, balance, inclination and the like, in the fourth step, the clustered features are provided for a softmax classifier to predict, and the method is not only beneficial to finding local features but also beneficial to finding global features which are difficult to define, by combining the two characteristics, the style characteristics of the character level can be captured, and the direct psychomotor process of the strokes of the writer during writing can also be captured.
Preferably, the two-channel convolutional neural network is adopted, the two input images are taken as a picture input network with two channels, the characteristics of the two pictures can be learned concurrently, the robustness and the effectiveness are improved, the medium convolutional neural network structure of the invention adopts a residual block (residual block), 8 residual blocks are used in total, 21 convolutional layers are used in total, a relu nonlinear activation layer is used, and the sizes of convolutional cores are all 3.
A method for using an off-line handwriting authentication system based on a scanning test paper comprises the following steps:
s1: extracting handwriting of students in the scanned test paper: and taking out the handwriting characters written by the students by relevant methods such as projection segmentation, connected domain and the like.
S2: pretreatment: the characters are converted into gray single-channel characters, pixel values are subtracted from the characters by 255, so that the pixel values are reversed, image handwriting blocks with the length and the width being about 64 are spliced, the length and the width are all 64 by adjusting the size, then data pairs and labels are generated, the labels of the image handwriting blocks of the same person are 1, and the labels of the image handwriting blocks of different persons are 0.
S3: algorithm training: the algorithm adopts end-to-end network training, the network inputs data to images, the network outputs corresponding labels, a model with an excellent effect is obtained by training, and the hyper-parameters are set as follows:
1) learning rate: the initial learning rate was set to 0.0001.
2) An optimizer: an Adam optimizer was employed.
3) Batch size: the batch size is set to 50.
S4: and outputting a result: and loading the trained model, performing handwriting identification output on the scanned test paper image, and judging whether the test paper is the test paper of the same person.
(III) advantageous effects
The invention provides an off-line handwriting identification system based on a scanning test paper and a using method thereof, which have the following beneficial effects:
the invention can automatically extract the handwriting in the test paper in batch, and identify the handwriting, completes the classification of the test paper, saves the labor and the time, has reasonable algorithm, high accuracy and convenient use, and solves the problems that the traditional handwriting identification is to manually identify and check the character layout characteristics, the character writing characteristics and the like through a handwriting identification mechanism, judge the difference between the handwriting and the handwriting, has longer period and consumes a large amount of labor.
Drawings
FIG. 1 is an overall algorithm framework diagram of the present invention;
FIG. 2 is a flow chart of handwriting extraction and pre-processing in the present invention;
FIG. 3 is a diagram of a network architecture of the present invention;
fig. 4 is a structural diagram of a residual block of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the 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.
As shown in fig. 1 to 4, the present invention provides a technical solution: an off-line handwriting identification system based on a scanning test paper comprises multi-image tuple input and a dual-channel convolution neural network, wherein the multi-image tuple input is used for sampling samples and collecting characteristics, the multi-image tuple is used for training, a plurality of image tuples are randomly extracted from a training set to participate in training, then extracted local characteristics are clustered to form global characteristics, and the dual-channel convolution neural network takes two images as a picture input neural network with dual channels to extract the characteristics of the two images concurrently.
Specifically, a scheme of multi-image tuple input is adopted, each image is used as a tuple, 20 tuples are input simultaneously, characteristics can be extracted from a plurality of handwritten character images, and specific characteristics in characters are reduced, firstly, 20 character image blocks are randomly extracted to form a 20 tuple image, secondly, each image of the 20 tuples is input into a CNN local characteristic extractor, so that the CNN can extract specific characteristics irrelevant to the texts, thirdly, the obtained local characteristics are aggregated by a global characteristic aggregator in an average value mode, through global characteristics extracted from a plurality of handwriting images, specific characteristics independent of text authors, such as structure, balance, inclination and the like, can be obtained, fourthly, the clustered characteristics are provided for a softmax classifier to predict, and the method is not only beneficial to finding the local characteristics but also beneficial to finding global characteristics which are difficult to define, by combining the two characteristics, the style characteristic of a character level can be captured, the direct psychomotor process of the strokes of an author during writing can also be captured, the handwriting characters in the scanned test paper image can be extracted by the methods of projection segmentation, domain connection and the like, then the handwriting characters are converted into gray single-channel characters through preprocessing, the pixel values are subtracted by 255, the pixel values are inverted, and the handwriting block images with the length and the width of about 64 are spliced.
Specifically, a double-channel convolutional neural network is adopted, two input images are taken as a picture input network with two channels, the characteristics of the two pictures can be learned concurrently, the robustness and the effectiveness are improved, a medium convolutional neural network structure of the invention adopts a residual block (residual block), 8 residual blocks are used totally, 21 convolutional layers are used totally, a relu nonlinear activation layer is used, the size of a convolutional kernel is 3, two groups of 20 tuples are favorably input into a double-channel network CNN image pair at the same time, each group of image tuples is a type of handwriting, a 4 x 1024 characteristic matrix is obtained, then characteristic clustering is carried out to enable the extracted characteristic matrix mean value to be a 1 x 1024 matrix, then a softmax classifier is used for classification, 0 or 1 is output, whether the two types of handwriting are written by the same person is judged, after an image pair is input, a convolution is firstly carried out to change from 2 channels into 32 channels, then 8 residual modules (residual blocks) are input, the number of the channels changes once after every two residual modules (residual blocks) pass through, the channels change into 128, 256, 512 and 1024 respectively, the network finally outputs 4 x 1024, the residual modules (residual blocks) have specific structures referring to fig. 4, the results obtained by carrying out convolution twice on the input are added with the input to obtain the output, wherein the sizes of convolution kernels are all 3 x 3, and relu is adopted for carrying out nonlinear mapping.
When in use, the handwriting of students in a scanned test paper is extracted, handwriting characters written by the students are taken out by relevant methods such as projection segmentation, connected domain and the like, preprocessing is carried out, the characters are converted into gray single-channel characters, pixel values are subtracted from 255 to reverse the pixel values, image handwriting blocks with the length and width of about 64 are spliced, the length and the width are adjusted to be 64, then data pairs and labels are generated, the label of the image handwriting block of the same person is 1, the label of the image handwriting block of different persons is 0, algorithm training is carried out, end-to-end network training is adopted in the algorithm, the network inputs the data pairs and the images, the corresponding labels are output through the network, a model with a good effect is obtained through training, the super-parameter setting is as follows, the learning rate is set to be 0.0001, an optimizer is adopted, the batch processing size is set to be 50, the result is output, the trained model is loaded, and (4) performing handwriting identification output on the scanned test paper image, and judging whether the test paper is the test paper of the same person.
In conclusion, the invention can automatically extract the handwriting in the test paper in batch, identify the handwriting and complete the classification of the test paper, thereby saving the labor and time, having reasonable algorithm, high accuracy and convenient use, and solving the problems that the traditional handwriting identification is manually carried out by a handwriting identification mechanism to identify and check the character layout characteristics, the character writing characteristics and the like, the handwriting is judged to be different from the handwriting, the period is long and a large amount of labor is consumed.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (4)
1. An off-line handwriting authentication system based on a scanning test paper is characterized in that: the method comprises multi-image tuple input and a dual-channel convolution neural network, wherein the multi-image tuple input is used for sampling samples and collecting characteristics, the multi-image tuple is used for training, a plurality of image tuples are randomly extracted from a training set to participate in the training, then the extracted local characteristics are clustered to form global characteristics, and the dual-channel convolution neural network takes two images as a picture input neural network with two channels and extracts the characteristics of the two images simultaneously.
2. The scanning test paper-based offline handwriting authentication system according to claim 1, wherein: the scheme of adopting multi-image tuple input is characterized in that each image is used as a tuple, 20 tuples are input simultaneously, characteristics can be extracted from a plurality of handwritten character images, and specific characteristics in characters are reduced simultaneously, firstly, 20 character image blocks are randomly extracted to form 20 tuple images, secondly, each image of the 20 tuples is input into a CNN local characteristic extractor, so that the CNN can extract specific characteristics irrelevant to the texts, thirdly, the obtained local characteristics are aggregated by a global characteristic aggregator in an average value mode, through global characteristics extracted from a plurality of handwriting images, specific characteristics independent of text authors, such as structure, balance, inclination and the like, can be obtained, fourthly, the clustered characteristics are provided for a softmax classifier to predict, and the method is not only beneficial to finding the local characteristics but also beneficial to finding global characteristics which are difficult to define, by combining the two characteristics, the style characteristics of the character level can be captured, and the direct psychomotor process of the strokes of the writer during writing can also be captured.
3. The scanning test paper-based offline handwriting authentication system according to claim 1, wherein: the double-channel convolutional neural network is adopted, two input images are taken as a picture input network with two channels, the characteristics of the two pictures can be learned simultaneously, the robustness and the effectiveness are improved, a medium convolutional neural network structure of the invention adopts a residual block (residual block), 8 residual blocks are used totally, 21 convolutional layers are used totally, a relu nonlinear activation layer is used, and the sizes of convolutional cores are all 3.
4. A use method of an off-line handwriting identification system based on a scanning test paper is characterized in that: the method comprises the following steps:
s1: extracting handwriting of students in the scanned test paper: taking out handwriting characters written by students by relevant methods such as projection segmentation, connected domain and the like;
s2: pretreatment: converting the characters into gray single-channel characters, subtracting pixel values from 255 to invert the pixel values, splicing into image handwriting blocks with the length and the width of about 64, adjusting the sizes to ensure that the length and the width are 64, and then generating data pairs and labels, wherein the labels of the image handwriting blocks of the same person are 1, and the labels of the image handwriting blocks of different persons are 0;
s3: algorithm training: the algorithm adopts end-to-end network training, the network inputs data to images, the network outputs corresponding labels, a model with an excellent effect is obtained by training, and the hyper-parameters are set as follows:
1) learning rate: the initial learning rate was set to 0.0001;
2) an optimizer: an Adam optimizer is adopted;
3) batch size: batch size set to 50;
s4: and outputting a result: and loading the trained model, performing handwriting identification output on the scanned test paper image, and judging whether the test paper is the test paper of the same person.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910884279.7A CN110659599A (en) | 2019-09-19 | 2019-09-19 | Scanning test paper-based offline handwriting authentication system and using method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910884279.7A CN110659599A (en) | 2019-09-19 | 2019-09-19 | Scanning test paper-based offline handwriting authentication system and using method thereof |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110659599A true CN110659599A (en) | 2020-01-07 |
Family
ID=69038187
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910884279.7A Pending CN110659599A (en) | 2019-09-19 | 2019-09-19 | Scanning test paper-based offline handwriting authentication system and using method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110659599A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111291722A (en) * | 2020-03-10 | 2020-06-16 | 无锡物联网创新中心有限公司 | Vehicle weight recognition system based on V2I technology |
| CN112215076A (en) * | 2020-09-11 | 2021-01-12 | 西安深信科创信息技术有限公司 | Deep handwriting identification method and device based on double-tower network |
| CN112528799A (en) * | 2020-12-02 | 2021-03-19 | 广州宏途教育网络科技有限公司 | Teaching live broadcast method and device, computer equipment and storage medium |
| CN113673415A (en) * | 2021-08-18 | 2021-11-19 | 山东建筑大学 | Handwritten Chinese character identity authentication method and system |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1540571A (en) * | 2003-10-29 | 2004-10-27 | 中国科学院计算技术研究所 | Method of sicriminating handwriting by computer based on analyzing local feature |
| CN107316066A (en) * | 2017-07-28 | 2017-11-03 | 北京工商大学 | Image classification method and system based on multi-path convolutional neural networks |
| CN108875696A (en) * | 2018-07-05 | 2018-11-23 | 五邑大学 | The Off-line Handwritten Chinese Recognition method of convolutional neural networks is separated based on depth |
| CN109740605A (en) * | 2018-12-07 | 2019-05-10 | 天津大学 | A kind of handwritten Chinese text recognition method based on CNN |
| CN109753578A (en) * | 2019-01-25 | 2019-05-14 | 浙江理工大学 | A kind of image search method based on exposure mask selection convolution feature |
| CN109815809A (en) * | 2018-12-19 | 2019-05-28 | 天津大学 | A kind of English handwriting identification method based on CNN |
| US20190171899A1 (en) * | 2017-12-05 | 2019-06-06 | Bull Sas | Automatic extraction of attributes of an object within a set of digital images |
| 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 |
| CN110188653A (en) * | 2019-05-27 | 2019-08-30 | 东南大学 | Activity recognition method based on local feature polymerization coding and shot and long term memory network |
| WO2019169042A1 (en) * | 2018-02-27 | 2019-09-06 | Cornell University | Ultra-sensitive detection of circulating tumor dna through genome-wide integration |
| CN110222696A (en) * | 2019-04-16 | 2019-09-10 | 天津科技大学 | A method of plant leaf blade disease identification is carried out using GPCNNs and ELM |
| CN110232361A (en) * | 2019-06-18 | 2019-09-13 | 中国科学院合肥物质科学研究院 | Human body behavior intension recognizing method and system based on the dense network of three-dimensional residual error |
-
2019
- 2019-09-19 CN CN201910884279.7A patent/CN110659599A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1540571A (en) * | 2003-10-29 | 2004-10-27 | 中国科学院计算技术研究所 | Method of sicriminating handwriting by computer based on analyzing local feature |
| CN107316066A (en) * | 2017-07-28 | 2017-11-03 | 北京工商大学 | Image classification method and system based on multi-path convolutional neural networks |
| US20190171899A1 (en) * | 2017-12-05 | 2019-06-06 | Bull Sas | Automatic extraction of attributes of an object within a set of digital images |
| WO2019169042A1 (en) * | 2018-02-27 | 2019-09-06 | Cornell University | Ultra-sensitive detection of circulating tumor dna through genome-wide integration |
| CN108875696A (en) * | 2018-07-05 | 2018-11-23 | 五邑大学 | The Off-line Handwritten Chinese Recognition method of convolutional neural networks is separated based on depth |
| CN109740605A (en) * | 2018-12-07 | 2019-05-10 | 天津大学 | A kind of handwritten Chinese text recognition method based on CNN |
| CN109815809A (en) * | 2018-12-19 | 2019-05-28 | 天津大学 | A kind of English handwriting identification method based on CNN |
| CN109753578A (en) * | 2019-01-25 | 2019-05-14 | 浙江理工大学 | A kind of image search method based on exposure mask selection convolution feature |
| CN110222696A (en) * | 2019-04-16 | 2019-09-10 | 天津科技大学 | A method of plant leaf blade disease identification is carried out using GPCNNs and ELM |
| 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 |
| CN110188653A (en) * | 2019-05-27 | 2019-08-30 | 东南大学 | Activity recognition method based on local feature polymerization coding and shot and long term memory network |
| CN110232361A (en) * | 2019-06-18 | 2019-09-13 | 中国科学院合肥物质科学研究院 | Human body behavior intension recognizing method and system based on the dense network of three-dimensional residual error |
Non-Patent Citations (3)
| Title |
|---|
| DILARA GUMUSBAS 等: "Offline Signature Identification and Verification using Capsule Network", 《2019 IEEE INTERNATIONAL SYMPOSIUM ON INNOVATIONS IN INTELLIGENT SYSTEMS AND APPLICATIONS (INISTA)》 * |
| 彪临轩: "一种基于机器视觉的试卷分数智能识别系统设计", 《电子制作》 * |
| 高定国 等编著: "《藏文信息处理的原理与应用》", 31 December 2014, 成都：西南交通大学出版社 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111291722A (en) * | 2020-03-10 | 2020-06-16 | 无锡物联网创新中心有限公司 | Vehicle weight recognition system based on V2I technology |
| CN112215076A (en) * | 2020-09-11 | 2021-01-12 | 西安深信科创信息技术有限公司 | Deep handwriting identification method and device based on double-tower network |
| CN112215076B (en) * | 2020-09-11 | 2022-08-12 | 西安深信科创信息技术有限公司 | Deep handwriting identification method and device based on double-tower network |
| CN112528799A (en) * | 2020-12-02 | 2021-03-19 | 广州宏途教育网络科技有限公司 | Teaching live broadcast method and device, computer equipment and storage medium |
| CN113673415A (en) * | 2021-08-18 | 2021-11-19 | 山东建筑大学 | Handwritten Chinese character identity authentication method and system |
| CN113673415B (en) * | 2021-08-18 | 2022-03-04 | 山东建筑大学 | Handwritten Chinese character identity authentication method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110659599A (en) | Scanning test paper-based offline handwriting authentication system and using method thereof | |
| CN108121702B (en) | Method and system for evaluating and reading mathematical subjective questions | |
| CN111476307B (en) | Lithium battery surface defect detection method based on depth field adaptation | |
| CN111091007A (en) | Method for identifying relationships among multiple enterprises based on public sentiment and enterprise portrait | |
| CN111612759B (en) | Printed matter defect identification method based on deep convolution generation type countermeasure network | |
| CN114170411A (en) | Picture emotion recognition method integrating multi-scale information | |
| Sinha et al. | Answer evaluation using machine learning | |
| Hallale et al. | Offline handwritten digit recognition using neural network | |
| CN115731237A (en) | Hot rolled steel strip surface defect detection method based on semi-supervised transfer learning | |
| Zhu et al. | NAGNet: A novel framework for real‐time students' sentiment analysis in the wisdom classroom | |
| Bintoro et al. | Lampung script recognition using convolutional neural network | |
| Wu et al. | Handwriting text-line detection and recognition in answer sheet composition with few labeled data | |
| CN110941976A (en) | Student classroom behavior identification method based on convolutional neural network | |
| CN112364837A (en) | Bill information identification method based on target detection and text identification | |
| Wijerathna et al. | Recognition and translation of Ancient Brahmi Letters using deep learning and NLP | |
| CN107992482B (en) | Protocol method and system for solving steps of mathematic subjective questions | |
| CN114638988A (en) | Teaching video automatic classification method and system based on different presentation modes | |
| CN115984215A (en) | Fiber bundle defect detection method based on twin network | |
| Koushik et al. | Automated marks entry processing in Handwritten answer scripts using character recognition techniques | |
| CN113688789A (en) | Online learning investment recognition method and system based on deep learning | |
| CN114005054A (en) | AI intelligence system of grading | |
| Zhi et al. | A Hybrid Framework for Text Recognition Used in Commodity Futures Document Verification | |
| Zhang et al. | Optical character recognition of electrical equipment nameplate with contrast enhancement | |
| Kır et al. | The Application of optical character recognition for mobile device via artificial neural networks with negative correlation learning algorithm | |
| CN113344079A (en) | Semi-automatic image label labeling method, system, terminal and medium |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200107 |