CN110490267B - Bill sorting method based on deep learning - Google Patents
Bill sorting method based on deep learning Download PDFInfo
- Publication number
- CN110490267B CN110490267B CN201910787567.0A CN201910787567A CN110490267B CN 110490267 B CN110490267 B CN 110490267B CN 201910787567 A CN201910787567 A CN 201910787567A CN 110490267 B CN110490267 B CN 110490267B
- Authority
- CN
- China
- Prior art keywords
- bill
- training
- network
- pictures
- types
- 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
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
Abstract
The invention discloses a bill sorting method based on deep learning, which comprises the following steps: s1, collecting service data, filing and labeling: roughly classifying the collected business pictures by using a VGG16 network pre-trained by ImageNet and an unsupervised clustering algorithm K-Means, and then accurately classifying manually to obtain labeled bill data; s2, dividing a data set: the method comprises the steps of dividing TotalDatasct into a training set TrainDataset and a verification set ValidationDataset, selectively dividing the training and verification data sets based on the number of samples of each specific bill type, and considering the condition of sample imbalance in the aspect of division of training data and verification data; s3, training STNEfficientNet: training by utilizing TrainDataset and verifying the network structure STNEfficientNet by utilizing ValidationDataset; and S4, deploying the trained model, and automatically sorting new bills uploaded by the customer by using the trained network STNEfficientNet. The invention has the advantages of high training speed and high precision, and can effectively improve the efficiency of manually classifying bills.
Description
Technical Field
The invention relates to the technical field of pattern recognition image processing, in particular to a bill sorting method based on deep learning.
Background
The classification (sorting) of invoices plays an important role in the flow of financial accounting and the like. Only if the invoice is correctly classified, the subsequent processing of the accounting process cannot be influenced. The traditional invoice classification method is generally to manually classify and archive invoices, and is low in efficiency and prone to errors.
Disclosure of Invention
Aiming at the problems, the invention provides a bill sorting method based on deep learning, which is based on an STN network and an EfficientNet network, applies a general picture classification technology EfficientNet [2] to the field of bill classification, and improves by utilizing a Spatial Transformer Networks [1] (STN) network to improve the classification precision of quality-poor bills such as distortion, inclination and the like.
To better understand the related problems, the following terms are now explained.
VGG: a typical convolutional neural network, see reference [3] for details;
general category of bill sorting: which refers to how many types of documents in total need to be sorted, the total number of such types depending on the specific business requirements. For example, if the business only needs to distinguish the identity card and the train ticket, the total category number of the ticket sorting in the ticket sorting task is 2; if the business needs to distinguish the identity card, the train ticket and the quota invoice, the total classification number of the bill sorting in the bill sorting task is 3;
K-Means clustering: a simple and commonly used clustering algorithm [4 ];
the network structure of the invention is STNEfficientNet, namely a network structure formed by embedding STN into the head of the EfficientNet network.
Reference documents:
[1]https：//arxiv.org/pdf/1506.02025.pdf；
[2]https：//arxiv.org/pdf/1905.11946.pdf；
[3]https：//arxiv.org/pdf/1409.1556.pdf；
[4]https：//en.wikipedia.org/wiki/K-means_clustering；
the automatic sorting problem of the bills is essentially a sorting problem of the bill pictures, so that the problem can be solved by adopting an EfficientNet sorting network which is proved to have the advantages of high reasoning speed, high precision and small model. As a latest proposed picture classification algorithm, the EfficientNet network has the characteristics of high reasoning speed, small model and high precision as proved by experiments. The actual business bill usually has various degrees of interference such as deformation, distortion, inclination and the like, so the conventional classification model cannot achieve better results in the actual business bill sorting. The STN network is proved to have the functions of automatic rotation correction and automatic perspective stretching recovery, and the STN and EfficientNet can effectively improve the sorting accuracy of the bills under the interference of deformation, distortion, inclination and the like of the bills; the invention adopts a method of cascading STN modules and EfficientNet network to solve the problem of automatic sorting of bills.
The invention realizes the purpose through the following technical scheme:
a bill sorting method based on deep learning comprises the following steps:
s1, collecting service data, filing and labeling: roughly classifying the collected business pictures by using a VGG16 network pre-trained by ImageNet and an unsupervised clustering algorithm K-Means, and then accurately classifying manually to obtain labeled bill data;
s2, dividing a data set: the method comprises the steps of dividing TotalDataset into a training set TrainDataset and a verification set ValidationDataset, selectively dividing the training and verification data sets based on the number of samples of each specific bill type, and considering the condition of sample imbalance in the aspect of division of training data and verification data;
s3, training STNEfficientNet: training by utilizing TrainDataset and verifying the network structure STNEfficientNet by utilizing ValidationDataset;
and S4, deploying the trained model, and automatically sorting new bills uploaded by the customer by using the trained network STNEfficientNet.
Further, the specific steps of S1 are as follows:
step 1-1: collecting n bills uploaded by a client and archiving the bills, wherein the bills are marked as image _000001.png, image _0000002.png,. and image _ n.png;
step 1-2: obtaining a 4096-dimensional vector (image _ i.png, vec _ i _4096) output by a 2 nd layer from the VGG16 network pre-trained by ImageNet, wherein i represents the ith invoice;
step 1-3: acquiring a 4096-dimensional vector Invoint _ Vec (image _ i.png, Vec _ i _4096) of each bill by using a Step1-2, wherein i belongs to [1, N ], and clustering the Invoint _ Vec into N (200) classes by using a K-Means algorithm;
step1-4: clustering the N pictures into pictures with the N being 200 types, manually checking and filing the pictures into corresponding bill types, wherein the specific operation flow is as follows:
step 1-4-1: taking any one type N _ j of the N types, randomly taking 20 pictures from the N types, and taking all pictures if the total number of the pictures in the cluster type is less than 20;
step 1-4-2: manually judging the actual invoice type of each picture in the 20 pictures, and then taking the type with the largest number of pictures in the invoice types as a preset type of N _ j;
step 1-4-3: judging whether the preset type of each type N _ j in the N types of classification results is obtained, if so, turning to Step1-4-4, otherwise, turning to Step 1-4-1;
step 1-4-4: merging the N _ j1 and the N _ j2 with the same preset type into the same type M _ j1_ j2, for example, if the N _1 type and the N _20 type have the same preset type, merging the N _ j1 and the N _ j2 into the same preset type, namely, M _1_20, and if no type is the same as N _ j, additionally marking the merging result as M _ j;
step 1-4-5: note that all merged data is M ═ M _1, M _ 2.., M _ j.., M _ x ];
step 1-4-6: taking each data subset M _ j in M, and sequentially carrying out bill classification on each invoice to finish archiving;
step 1-5: all the archived pictures are marked as TotalDataset, which includes two parts: part 1 is all the collected n bill pictures, and part 2 is the bill category corresponding to each bill picture.
Further, the specific steps of S2 are as follows:
step 2-1: counting the sample number of each bill, setting the TrainDataset of a training set to be null, and setting the ValidationDataset of a verification set to be null;
step 2-2: taking one of the N types of marked bill types, and determining whether the number of samples is greater than 500? If yes, turning to Step2-3, otherwise, turning to Step 2-4;
step 2-3: randomly selecting 100 sheets from the bill types, directly placing the sheets into a verification set ValidationDataset, and placing the residual samples of the bill types into a training set TrainDataset;
step 2-4: randomly selecting 100 sheets from the bill types, copying the 100 sheets and putting the 100 sheets into a verification set ValidationDataset, and putting all samples of the bill types into a training set TrainDataset;
step 2-5: and whether all the samples of the bill types are divided into a training set TrainDataset and a verification set ValidationDataset, if so, turning to Step3, and otherwise, turning to Step 2-2.
Further, the specific steps of S3 are as follows:
step 3-1: setting algorithm parameters, specifically including an initial learning rate of 1e-4, a trained batch size of 16, a network parameter optimizer SGD (random gradient descent method) with a driving quantity of momentum of 0.9 and a weight attenuation of 1e-4, wherein the total training period of the algorithm is EPOACH of 50, and the initial EPOACH of 0;
step 3-2: establishing a network model (STNEfficientNet ()), and specifically establishing the network model according to the following steps:
step 3-1: establishing a space transformation network STN;
step 3-2: establishing an EfficientNet network;
step 3-3: the STN network is cascaded with an EfficientNet network to form the STNEfficientNet network;
step 3-3: judging whether the EPOACH is smaller than EPOACH, if so, turning to Step3-4, otherwise, turning to S4;
step 3-4: taking batch-size invoices from the TrainDataset, training the model, and updating the model algorithm by using the SGD;
step 3-5: judging whether all invoices in the TrainDataset are trained, if so, turning to the Step3-6, and otherwise, turning to the Step 3-4;
step 3-6: and verifying the precision of the model by using ValidationDataset, saving the trained model, and turning to Step 3-3.
The invention has the beneficial effects that:
the invention has the advantages of high training speed and high precision, and can effectively improve the efficiency of manually classifying bills.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the embodiments or the drawings needed to be practical in the prior art description, and obviously, the drawings in the following description are only some embodiments of the embodiments, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1: a flow chart of the overall implementation of the invention;
FIG. 2: the invention relates to a STNEfficientNet construction process.
FIG. 3: the invention relates to a business data collection process.
FIG. 4: the data set partitioning logic of the present invention.
FIG. 5: the invention relates to a STNEfficientNet training process.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be described in detail below. It is to be understood that the described embodiments are merely exemplary of the invention, and not restrictive of the full scope of the invention. All other embodiments, which can be derived by a person skilled in the art from the examples given herein without any inventive step, are within the scope of the present invention.
In any embodiment, as shown in fig. 1, the bill sorting method based on deep learning of the present invention includes the following steps:
s1, collecting service data, filing and labeling: roughly classifying the collected business pictures by using a VGG16 network pre-trained by ImageNet and an unsupervised clustering algorithm K-Means, and then accurately classifying manually to obtain labeled bill data;
s2, dividing a data set: the method comprises the steps of dividing TotalDataset into a training set TrainDataset and a verification set ValidationDataset, selectively dividing the training and verification data sets based on the number of samples of each specific bill type, and considering the condition of sample imbalance in the aspect of division of training data and verification data;
s3, training STNEfficientNet: training by utilizing TrainDataset and verifying the network structure STNEfficientNet by utilizing ValidationDataset;
and S4, deploying the trained model, and automatically sorting new bills uploaded by the customer by using the trained network STNEfficientNet.
In an embodiment, the collecting the business data according to the process shown in fig. 3 of the specification specifically includes the following steps:
step 1-1: collecting n bills uploaded by a client and archiving the bills, wherein the bills are marked as image _000001.png, image _0000002.png,. and image _ n.png;
step 1-2: obtaining a 4096-dimensional vector (image _ i.png, vec _ i _4096) output by a 2 nd layer from the VGG16 network pre-trained by ImageNet, wherein i represents the ith invoice;
step 1-3: acquiring a 4096-dimensional vector Invoint _ Vec (image _ i.png, Vec _ i _4096) of each bill by using a Step1-2, wherein i belongs to [1, N ], and clustering the Invoint _ Vec into N (200) classes by using a K-Means algorithm;
step1-4: clustering the N pictures into pictures with the N being 200 types, manually checking and filing the pictures into corresponding bill types, wherein the specific operation flow is as follows:
step 1-4-1: taking any one type N _ j of the N types, randomly taking 20 pictures from the N types, and taking all pictures if the total number of the pictures in the cluster type is less than 20;
step 1-4-2: manually judging the actual invoice type of each picture in the 20 pictures, and then taking the type with the largest number of pictures in the invoice types as a preset type of N _ j;
step 1-4-3: judging whether the preset type of each type N _ j in the N types of classification results is obtained, if so, turning to Step1-4-4, otherwise, turning to Step 1-4-1;
step 1-4-4: merging the N _ j1 and the N _ j2 with the same preset type into the same type M _ j1_ j2, for example, if the N _1 type and the N _20 type have the same preset type, merging the N _ j1 and the N _ j2 into the same preset type, namely, M _1_20, and if no type is the same as N _ j, additionally marking the merging result as M _ j;
step 1-4-5: note that all merged data is M ═ M _1, M _ 2.., M _ j.., M _ x ];
step 1-4-6: taking each data subset M _ j in M, and sequentially carrying out bill classification on each invoice to finish archiving;
step 1-5: all the archived pictures are marked as TotalDataset, which includes two parts: part 1 is all the collected n bill pictures, and part 2 is the bill category corresponding to each bill picture.
In a specific embodiment, the logic partitioning of TotalDataset into a training set TrainDataset and a verification set ValidationDataset according to the logic described in fig. 4 of the specification specifically includes the following steps:
step 2-1: counting the sample number of each bill, setting the TrainDataset of a training set to be null, and setting the ValidationDataset of a verification set to be null;
step 2-2: taking one of the N types of marked bill types, and determining whether the number of samples is greater than 500? If yes, turning to Step2-3, otherwise, turning to Step 2-4;
step 2-3: randomly selecting 100 sheets from the bill types, directly placing the sheets into a verification set ValidationDataset, and placing the residual samples of the bill types into a training set TrainDataset;
step 2-4: randomly selecting 100 sheets from the bill types, copying the 100 sheets and putting the 100 sheets into a verification set ValidationDataset, and putting all samples of the bill types into a training set TrainDataset;
step 2-5: and whether all the samples of the bill types are divided into a training set TrainDataset and a verification set ValidationDataset, if so, turning to Step3, and otherwise, turning to Step 2-2.
In a specific embodiment, the network structure STNEfficientNet constructed in the specification of fig. 2 is trained by using TrainDataset and validated by using validatabet, wherein the training mode is trained in the mode described in the specification of fig. 5, and the specific training steps are as follows:
step 3-1: setting algorithm parameters, specifically including an initial learning rate of 1e-4, a trained batch size of 16, a network parameter optimizer SGD (random gradient descent method) with a driving quantity of momentum of 0.9 and a weight attenuation of 1e-4, wherein the total training period of the algorithm is EPOACH of 50, and the initial EPOACH of 0;
step 3-2: establishing a network model (STNEfficientNet ()), and specifically establishing the network model according to the following steps:
step 3-1: establishing a space transformation network STN;
step 3-2: establishing an EfficientNet network;
step 3-3: the STN network is cascaded with an EfficientNet network to form the STNEfficientNet network;
step 3-3: judging whether the EPOACH is smaller than EPOACH, if so, turning to Step3-4, otherwise, turning to S4;
step 3-4: taking batch-size invoices from the TrainDataset, training the model, and updating the model algorithm by using the SGD;
step 3-5: judging whether all invoices in the TrainDataset are trained, if so, turning to the Step3-6, and otherwise, turning to the Step 3-4;
step 3-6: and verifying the precision of the model by using ValidationDataset, saving the trained model, and turning to Step 3-3.
The boxes with vertical lines in the figure indicate that this step is done by one sub-module.
Practice proves that the invention can realize efficient and accurate bill sorting.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. The various features described in the foregoing detailed description may be combined in any suitable manner without contradiction, and various combinations that are possible in the present invention will not be further described in order to avoid unnecessary repetition. Any combination of the different embodiments of the present invention is also possible, and the same should be considered as the disclosure of the present invention as long as it does not depart from the gist of the present invention.
Claims (3)
1. A bill sorting method based on deep learning is characterized by comprising the following steps:
s1, collecting service data, filing and labeling: roughly classifying the collected business pictures by using a VGG16 network pre-trained by ImageNet and an unsupervised clustering algorithm K-Means, and then accurately classifying manually to obtain labeled bill data;
s2, dividing a data set: the method comprises the steps of dividing TotalDataset into a training set TrainDataset and a verification set ValidationDataset, selectively dividing the training and verification data sets based on the number of samples of each specific bill type, and considering the condition of sample imbalance in the aspect of division of training data and verification data;
s3, training STNEfficientNet: training by utilizing TrainDataset and verifying the network structure STNEfficientNet by utilizing ValidationDataset; the S3 concrete steps are as follows:
step 3-1: setting algorithm parameters, specifically including an initial learning rate of 1e-4, a trained batch size of 16, a network parameter optimizer SGD (random gradient descent method) with a driving quantity of momentum of 0.9 and a weight attenuation of 1e-4, wherein the total training period of the algorithm is epoach of 50, and the initial epoach of 0;
step 3-2: establishing a network model (STNEfficientNet ()), and specifically establishing the network model according to the following steps:
step 3-2-1: establishing a space transformation network STN;
step 3-2-2: establishing an EfficientNet network;
step 3-2-3: the STN network is cascaded with an EfficientNet network to form the STNEfficientNet network;
step 3-3: judging whether the epoach is smaller than the epoach, if so, turning to Step3-4, otherwise, turning to S4;
step 3-4: taking batch-size invoices from the TrainDataset, training the model, and updating the model algorithm by using the SGD;
step 3-5: judging whether all invoices in the TrainDataset are trained, if so, turning to the Step3-6, and otherwise, turning to the Step 3-4;
step 3-6: verifying the precision of the model by using ValidationDataset, storing the trained model, and turning to Step 3-3;
and S4, deploying the trained model, and automatically sorting new bills uploaded by the customer by using the trained network STNEfficientNet.
2. The bill sorting method based on deep learning as claimed in claim 1, wherein the step S1 is as follows:
step 1-1: collecting n tickets uploaded by a client and archiving the n tickets, wherein the n tickets are marked as image _000001.png, image _0000002.png, … and image _ n.png;
step 1-2: obtaining a 4096-dimensional vector (image _ i.png, vec _ i _4096) output by a 2 nd layer from the VGG16 network pre-trained by ImageNet, wherein i represents the ith invoice;
step 1-3: acquiring a 4096-dimensional vector Invoint _ Vec (image _ i.png, Vec _ i _4096) of each bill by using a Step1-2, wherein i belongs to [1, N ], and clustering the Invoint _ Vec into N (200) classes by using a K-Means algorithm;
step1-4, clustering the N pictures into pictures with the N being 200 types, manually checking and filing the pictures into corresponding bill types, wherein the specific operation flow is as follows:
step 1-4-1: taking any one type N _ j of the N types, randomly taking 20 pictures from the N types, and taking all pictures if the total number of the pictures in the cluster type is less than 20;
step 1-4-2: manually judging the actual invoice type of each picture in the 20 pictures, and then taking the type with the largest number of pictures in the invoice types as a preset type of N _ j;
step 1-4-3: judging whether the preset type of each type N _ j in the N types of classification results is obtained, if so, turning to Step1-4-4, otherwise, turning to Step 1-4-1;
step 1-4-4: merging the N _ j1 and the N _ j2 with the same preset type into the same type M _ j1_ j2, for example, if the N _1 type and the N _20 type have the same preset type, merging the N _ j1 and the N _ j2 into the same preset type, namely, M _1_20, and if no type is the same as N _ j, additionally marking the merging result as M _ j;
step 1-4-5: note that all merged data is M ═ M _1, M _2, …, M _ j, …, M _ x ];
step 1-4-6: taking each data subset M _ j in M, and sequentially carrying out bill classification on each invoice to finish archiving;
step 1-5: all the archived pictures are marked as TotalDataset, which includes two parts: part 1 is all the collected n bill pictures, and part 2 is the bill category corresponding to each bill picture.
3. The bill sorting method based on deep learning as claimed in claim 1, wherein the step S2 is as follows:
step 2-1: counting the sample number of each bill, setting the TrainDataset of a training set to be null, and setting the ValidationDataset of a verification set to be null;
step 2-2: taking one of the N types of marked bill types, and determining whether the number of samples is greater than 500? If yes, turning to Step2-3, otherwise, turning to Step 2-4;
step 2-3: randomly selecting 100 sheets from the bill types, directly placing the sheets into a verification set ValidationDataset, and placing the residual samples of the bill types into a training set TrainDataset;
step 2-4: randomly selecting 100 sheets from the bill types, copying the 100 sheets and putting the 100 sheets into a verification set ValidationDataset, and putting all samples of the bill types into a training set TrainDataset;
step 2-5: and whether all the samples of the bill types are divided into a training set TrainDataset and a verification set ValidationDataset, if so, turning to Step3, and otherwise, turning to Step 2-2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910787567.0A CN110490267B (en) | 2019-08-23 | 2019-08-23 | Bill sorting method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910787567.0A CN110490267B (en) | 2019-08-23 | 2019-08-23 | Bill sorting method based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110490267A CN110490267A (en) | 2019-11-22 |
| CN110490267B true CN110490267B (en) | 2022-03-01 |
Family
ID=68553919
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910787567.0A Active CN110490267B (en) | 2019-08-23 | 2019-08-23 | Bill sorting method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110490267B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111523579B (en) * | 2020-04-14 | 2022-05-03 | 燕山大学 | Vehicle type recognition method and system based on improved deep learning |
| CN111666872B (en) * | 2020-06-04 | 2022-08-05 | 电子科技大学 | Efficient behavior identification method under data imbalance |
| CN112115943A (en) * | 2020-09-16 | 2020-12-22 | 四川长虹电器股份有限公司 | Bill rotation angle detection method based on deep learning |
| CN116778513B (en) * | 2023-08-24 | 2023-10-27 | 国网山西省电力公司太原供电公司 | Intelligent archiving control method for bills in power industry |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7165041B1 (en) * | 1999-05-27 | 2007-01-16 | Accenture, Llp | Web-based architecture sales tool |
| US20120002008A1 (en) * | 2010-07-04 | 2012-01-05 | David Valin | Apparatus for secure recording and transformation of images to light for identification, and audio visual projection to spatial point targeted area |
| US10600002B2 (en) * | 2016-08-04 | 2020-03-24 | Loom Systems LTD. | Machine learning techniques for providing enriched root causes based on machine-generated data |
| CN105045780B (en) * | 2015-07-15 | 2017-12-26 | 广州敦和信息技术有限公司 | A kind of recognition methods of invoice brief note semantic information and device |
| US20170124497A1 (en) * | 2015-10-28 | 2017-05-04 | Fractal Industries, Inc. | System for automated capture and analysis of business information for reliable business venture outcome prediction |
| CN106096667B (en) * | 2016-07-26 | 2017-11-10 | 江苏鸿信***集成有限公司 | Bill images sorting technique based on SVM |
| CN107292823A (en) * | 2017-08-20 | 2017-10-24 | 平安科技（深圳）有限公司 | Electronic installation, the method for invoice classification and computer-readable recording medium |
| CN108446621A (en) * | 2018-03-14 | 2018-08-24 | 平安科技（深圳）有限公司 | Bank slip recognition method, server and computer readable storage medium |
| CN110070665B (en) * | 2019-04-24 | 2021-05-28 | 武汉华创欣网科技有限公司 | Insurance bill classification method based on deep learning |
-
2019
- 2019-08-23 CN CN201910787567.0A patent/CN110490267B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN110490267A (en) | 2019-11-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110490267B (en) | Bill sorting method based on deep learning | |
| CN101329734B (en) | License plate character recognition method based on K-L transform and LS-SVM | |
| US8831361B2 (en) | Method and system for commercial document image classification | |
| CN102346847B (en) | License plate character recognizing method of support vector machine | |
| CN107067044A (en) | A kind of finance reimbursement unanimous vote is according to intelligent checks system | |
| CN107844559A (en) | A kind of file classifying method, device and electronic equipment | |
| US20070239638A1 (en) | Text classification by weighted proximal support vector machine | |
| CN108777021B (en) | Bill identification method and system based on scanner mixed scanning | |
| CN110705607B (en) | Industry multi-label noise reduction method based on cyclic re-labeling self-service method | |
| CN103617429A (en) | Sorting method and system for active learning | |
| CN108764302B (en) | Bill image classification method based on color features and bag-of-words features | |
| CN104834940A (en) | Medical image inspection disease classification method based on support vector machine (SVM) | |
| CN110287311B (en) | Text classification method and device, storage medium and computer equipment | |
| CN111597958B (en) | Highly automated bill classification method and system | |
| CN109344856B (en) | Offline signature identification method based on multilayer discriminant feature learning | |
| CN105930792A (en) | Human action classification method based on video local feature dictionary | |
| CN105975611A (en) | Self-adaptive combined downsampling reinforcing learning machine | |
| CN112733936A (en) | Recyclable garbage classification method based on image recognition | |
| CN107292225A (en) | A kind of face identification method | |
| CN110188714A (en) | A kind of method, system and storage medium for realizing financial management under chat scenario | |
| CN114663002A (en) | Method and equipment for automatically matching performance assessment indexes | |
| CN110991442B (en) | High-precision identification method for license plate cloud of expressway | |
| CN105512675A (en) | Memory multi-point crossover gravitational search-based feature selection method | |
| CN110197140A (en) | Material checking method and equipment based on Text region | |
| CN114511866A (en) | Data auditing method, device, system, processor and machine-readable storage 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |