CN112561480A - Intelligent workflow pushing method, equipment and computer storage medium - Google Patents
Intelligent workflow pushing method, equipment and computer storage medium Download PDFInfo
- Publication number
- CN112561480A CN112561480A CN202011492148.3A CN202011492148A CN112561480A CN 112561480 A CN112561480 A CN 112561480A CN 202011492148 A CN202011492148 A CN 202011492148A CN 112561480 A CN112561480 A CN 112561480A
- Authority
- CN
- China
- Prior art keywords
- workflow
- image data
- image
- training
- network
- 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
- 238000000034 method Methods 0.000 title claims abstract description 91
- 238000003860 storage Methods 0.000 title claims abstract description 13
- 238000012545 processing Methods 0.000 claims abstract description 22
- 238000012549 training Methods 0.000 claims description 67
- 238000013145 classification model Methods 0.000 claims description 33
- 230000008569 process Effects 0.000 claims description 32
- 230000007246 mechanism Effects 0.000 claims description 17
- 230000008014 freezing Effects 0.000 claims description 15
- 238000007710 freezing Methods 0.000 claims description 15
- 230000001737 promoting effect Effects 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 3
- 230000001902 propagating effect Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 15
- 238000010586 diagram Methods 0.000 description 11
- 230000006870 function Effects 0.000 description 10
- 238000004590 computer program Methods 0.000 description 7
- 238000012544 monitoring process Methods 0.000 description 6
- 238000012986 modification Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 5
- 230000001360 synchronised effect Effects 0.000 description 5
- 239000011159 matrix material Substances 0.000 description 4
- 238000003745 diagnosis Methods 0.000 description 3
- 238000011176 pooling Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 238000013526 transfer learning Methods 0.000 description 2
- 230000001960 triggered effect Effects 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002688 persistence Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
- G06Q10/103—Workflow collaboration or project management
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/08—Insurance
Abstract
The invention relates to the field of business process optimization, and discloses a method, equipment and a computer storage medium for intelligently pushing a workflow, wherein the method comprises the following steps: when the uploaded image classification meets one of preset classifications of images required by corresponding services, inserting image data corresponding to the uploaded image into a first operation table; when all the image data corresponding to the preset classification of the images required by the business in the first operation table are uploaded, generating a workflow pushing instruction by the first operation table; performing idempotent processing on the push workflow instruction, and writing the push workflow instruction into a message queue; and when the business system monitors the push workflow instruction in the message queue, executing the next operation of the business system. The method solves the problem of manual misoperation such as missed triggering or repeated triggering of the manually pushed workflow, and the like, and realizes the method of intelligently pushing the workflow without manual triggering, thereby improving the working efficiency.
Description
Technical Field
The present invention relates to the field of business process optimization, and in particular, to a method, an apparatus, and a computer storage medium for intelligently pushing a workflow.
Background
In recent years, with the rapid development of national economy and the rapid improvement of the living standard of people, the people have greatly improved insurance consciousness, and the insurance industry is greatly developed. With the rapid development of insurance industry, the business process of the insurance industry is more and more complex. The existing business process needs image scanning and then uploads the image data to an image system, after the image data is uploaded to the image system, whether the image data corresponding to the current business is uploaded completely or not is manually checked, and communication is needed with personnel who do not upload the image data part, wherein communication and coordination are needed, and if the image data corresponding to the current business is uploaded completely, manual triggering is needed to continuously push the workflow. In the prior art, excessive manual participation is needed in the process of uploading images and the process of pushing workflows, the problem of missed triggering or repeated triggering or manual misoperation is easy to occur, and meanwhile, the improvement of the working efficiency is not facilitated.
Disclosure of Invention
In view of this, a method for intelligently pushing a workflow is provided, which solves the problem of manual misoperation such as missed triggering or multiple triggering of a manually pushed workflow.
The embodiment of the application provides an intelligent workflow promoting method, which comprises the following steps:
when the uploaded image classification meets one of preset classifications of images required by corresponding services, inserting image data corresponding to the uploaded image into a first operation table;
when all the image data corresponding to the preset classification of the images required by the business in the first operation table are uploaded, generating a workflow pushing instruction by the first operation table;
performing idempotent processing on the push workflow instruction, and writing the push workflow instruction into a message queue;
and when the business system monitors the push workflow instruction in the message queue, executing the next operation of the business system.
In an embodiment, before the step of inserting the image data corresponding to the uploaded image into the first operation table when the uploaded image classification meets one of the preset classifications of images required by corresponding services, the method includes:
training part of the preset network of the pre-training model by using the image data labeled by the user-defined label to generate a corresponding image classification model;
inputting the images uploaded by image scanning into the image classification model for identification and classification; the image scanning and uploading mechanism adopts a multi-channel data acquisition mechanism.
In an embodiment, the training of the part of the predetermined network of the pre-training model using the image data labeled by the user-defined label includes:
fixing parameters of the frozen network;
and training the non-freezing network by using the image data labeled by the user.
In an embodiment, the training a part of the preset network of the pre-training model by using the image data labeled by the user, further includes:
training a freezing network and a non-freezing network by using the image data labeled by a user; wherein the parameters of the frozen network are dynamically adjusted according to a training process.
In an embodiment, the step of training a part of the preset network of the pre-training model by using the image data labeled by the user, wherein the step of updating the parameters of the image classification model by using a back propagation method includes:
comparing the recognition result generated by training with the image data labeled by the user-defined label;
and reversely propagating the comparison result, and updating the internal parameters of the image classification model.
In an embodiment, the process of constructing the custom labeled image data includes:
performing denoising preprocessing on the image data in the database;
and marking according to the service preset label.
In one embodiment, the idempotent processing includes at least one of:
globally unique identification, token mechanism, pessimistic lock, and optimistic lock.
In an embodiment, the preset training model employs a VGG-19 network, the frozen network is a first preset number of convolutional layers, and the non-frozen network is a second preset number of convolutional layers.
To achieve the above object, there is also provided a computer readable storage medium having stored thereon an intelligent push workflow method program, which when executed by a processor, implements the steps of any of the methods described above.
To achieve the above object, there is also provided an intelligent push workflow device, including a memory, a processor, and an intelligent push workflow method program stored on the memory and executable on the processor, where the processor executes the intelligent push workflow method program to implement any of the steps of the method described above.
One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: when the uploaded image classification meets one of preset classifications of images required by corresponding services, inserting image data corresponding to the uploaded image into a first operation table; the monitoring of the uploaded image data classification in the step does not need manual work, the problems of misoperation and the like caused by manual monitoring are avoided, and the correctness and timeliness of the image data inserted into the first operation table are guaranteed when the conditions are met. When all the image data corresponding to the preset classification of the images required by the business in the first operation table are uploaded, generating a workflow pushing instruction by the first operation table; in this step, the push workflow command may be generated only if the above conditions are satisfied. And the correctness of the next step of pushing the generation of the workflow instruction is ensured. Performing idempotent processing on the push workflow instruction, and writing the push workflow instruction into a message queue; the idempotent processing in the step ensures the uniqueness of the workflow pushing instruction written into the message queue and ensures that the business process is not triggered repeatedly. And when the business system monitors the push workflow instruction in the message queue, executing the next operation of the business system. In the step, when the business system monitors the instruction of pushing the workflow, the next operation of the business system is automatically executed, and in the whole process of pushing the business process, manual participation and operation are not needed, so that the problem of manual misoperation such as missed triggering or repeated triggering and the like of the manually pushed workflow is avoided, the method of intelligently pushing the workflow without manual triggering is realized, and the working efficiency is improved.
Drawings
FIG. 1 is a hardware architecture diagram of an intelligent push workflow method involved in an embodiment of the present application;
FIG. 2 is a schematic flow chart diagram illustrating a first embodiment of the intelligent push workflow method of the present application;
FIG. 3 is a schematic flow chart diagram illustrating a second embodiment of the intelligent push workflow method of the present application;
FIG. 4 is a flowchart illustrating the training of the image classification model for intelligently pushing workflow according to the present application
Fig. 5 is a flowchart illustrating a specific implementation step of step S210 in the second embodiment of the intelligent push workflow method of the present application;
fig. 6 is a flowchart illustrating another specific implementation step of step S210 in the second embodiment of the intelligent push workflow method of the present application.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The main solution of the embodiment of the invention is as follows: when the uploaded image classification meets one of preset classifications of images required by corresponding services, inserting image data corresponding to the uploaded image into a first operation table; when all the image data corresponding to the preset classification of the images required by the business in the first operation table are uploaded, generating a workflow pushing instruction by the first operation table; performing idempotent processing on the push workflow instruction, and writing the push workflow instruction into a message queue; and when the business system monitors the push workflow instruction in the message queue, executing the next operation of the business system. The method solves the problem of manual misoperation such as missed triggering or repeated triggering of the manually pushed workflow, and the like, and realizes the method of intelligently pushing the workflow without manual triggering, thereby improving the working efficiency.
In order to better understand the technical solution, the technical solution will be described in detail with reference to the drawings and the specific embodiments.
The present application relates to an intelligent push workflow device 010 comprising as shown in fig. 1: at least one processor 012, memory 011.
The processor 012 may be an integrated circuit chip having signal processing capability. In implementation, the steps of the method may be performed by hardware integrated logic circuits or instructions in the form of software in the processor 012. The processor 012 may be a general-purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic device, or discrete hardware components. The various methods, steps and logic blocks disclosed in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The software module may be located in ram, flash memory, rom, prom, or eprom, registers, etc. storage media as is well known in the art. The storage medium is located in the memory 011, and the processor 012 reads the information in the memory 011 and completes the steps of the method in combination with the hardware.
It is to be understood that the memory 011 in embodiments of the present invention can be either volatile memory or nonvolatile memory, or can include both volatile and nonvolatile memory. The non-volatile Memory may be a Read Only Memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically Erasable PROM (EEPROM), or a flash Memory. Volatile Memory can be Random Access Memory (RAM), which acts as external cache Memory. By way of illustration and not limitation, many forms of RAM are available, such as Static random access memory (Static RAM, SRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic random access memory (Synchronous DRAM, SDRAM), Double data rate Synchronous Dynamic random access memory (ddr DRAM), Enhanced Synchronous SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), and Direct Rambus RAM (DRRAM). The memory 011 of the systems and methods described in connection with the embodiments of the invention is intended to comprise, without being limited to, these and any other suitable types of memory.
Referring to fig. 2, fig. 2 is a first embodiment of the intelligent workflow-driving method of the present application, which includes:
step S110: and when the uploaded image classification meets one of the preset classifications of the images required by the corresponding service, inserting the image data corresponding to the uploaded image into a first operation table.
The image data required in the business process is a preset number, and the preset number is not limited.
When the image classification obtained by classifying the input image through the image classification model meets one of the preset classifications of the images required by the corresponding business, the currently uploaded image data can be inserted into the first operation table.
The image data may include at least one of: images, business affiliated departments, uploads persons, uploading time and uploading models. The attributes in the first operation table include: images, business affiliated departments, uploads persons, uploading time and uploading models. If the image data does not obtain the data of the attribute corresponding to the first operation table, the attribute of the image data is set to be null in the corresponding attribute of the first operation table.
In the process of inserting the currently uploaded image data into the first operation table, the insertion may be sequentially performed using a first come first serve (FIFO) manner.
Wherein one of the first operation tables corresponds to one corresponding service.
Step S120: and when all the image data corresponding to the preset classification of the images required by the service in the first operation table are uploaded, generating a workflow pushing instruction by the first operation table.
The first operation table monitors the data of the first operation table, and the monitoring method can be that a set of preset classifications of images required by the service is constructed in advance, and elements in the set are the preset classifications of the images required by the service. When all elements in the set are uploaded, generating a push workflow instruction by the first operation table.
Step S130: and performing idempotent processing on the push workflow instruction, and writing the push workflow instruction into a message queue.
Idempotent is a mathematical and computer concept, which is commonly found in abstract algebra. An idempotent operation in programming is characterized by the same effect of any number of executions as one. An idempotent function, or idempotent method, refers to a function that can be repeatedly performed using the same parameters and achieve the same result. These functions do not affect the system state and there is no concern that repeated execution will cause changes to the system.
In internet application, especially under the condition that a network is unstable, messages of a message queue rockmq can repeat unpredictably, when the messages of the message queue rockmq repeat, in the embodiment, subsequent business workflow can be repeatedly advanced, which can cause propulsion error of business work, so that performing idempotent operation on the messages can ensure correctness of business work advancement, avoid repeated advancement of business work, and ensure that the transmission of the workflow-advancing instruction is correct.
Two cases of message repetition:
first, message repetition at transmission
When a message is successfully sent to the server and persistence is completed, network flash or a client is down occurs, so that the server fails to respond to the client. If at this point the producer realizes that the message was sent unsuccessfully and attempts to send the message again, the consumer will subsequently receive two messages with the same content and the same message sequence number.
Second, message repetition at delivery
Under the scene of message consumption, the message is delivered to a consumer and service processing is completed, and when the client feeds back a response to the server, the network is flashed. In order to ensure that the message is consumed at least once, the server side of the message queue rockmq will send the processed message before trying to deliver the message again after the network recovers, and the consumer will subsequently receive two messages with the same content and the same message sequence number.
Step S140: and when the business system monitors the push workflow instruction in the message queue, executing the next operation of the business system.
The business system does not need manual participation in the process of monitoring the push workflow command, realizes the intellectualization of the push workflow, reduces the manual input and improves the working efficiency.
In the above embodiment, there are advantageous effects: when the uploaded image classification meets one of preset classifications of images required by corresponding services, inserting image data corresponding to the uploaded image into a first operation table; the monitoring of the uploaded image data classification in the step does not need manual work, the problems of misoperation and the like caused by manual monitoring are avoided, and the correctness and timeliness of the image data inserted into the first operation table are guaranteed when the conditions are met. When all the image data corresponding to the preset classification of the images required by the business in the first operation table are uploaded, generating a workflow pushing instruction by the first operation table; in this step, the push workflow command may be generated only if the above conditions are satisfied. And the correctness of the next step of pushing the generation of the workflow instruction is ensured. Performing idempotent processing on the push workflow instruction, and writing the push workflow instruction into a message queue; the idempotent processing in the step ensures the uniqueness of the workflow pushing instruction written into the message queue and ensures that the business process is not triggered repeatedly. And when the business system monitors the push workflow instruction in the message queue, executing the next operation of the business system. In the step, when the business system monitors the instruction of pushing the workflow, the next operation of the business system is automatically executed, and in the whole process of pushing the business process, manual participation and operation are not needed, so that the problem of manual misoperation such as missed triggering or repeated triggering and the like of the manually pushed workflow is avoided, the method of intelligently pushing the workflow without manual triggering is realized, and the working efficiency is improved.
Referring to fig. 3, fig. 3 is a second embodiment of the method for intelligently pushing a workflow according to the present application, where before the step of inserting image data corresponding to an uploaded image into a first operation table when the uploaded image classification meets one of preset classifications of images required by corresponding services, the method includes:
step S210: and training part of the preset network of the pre-training model by using the image data labeled by the user-defined label to generate a corresponding image classification model.
The pre-training model can be obtained by large-scale data set training, and the pre-training model can be used as a fixed feature extractor or the initialization weight of a network to be applied to other tasks.
The pre-training model is trained by using the image data set labeled by the user, namely, the parameter experience value of the pre-training model is ensured, and the characteristics of the image data set labeled by the user are added, so that the accuracy of image data classification can be effectively improved.
Transfer Learning (Transfer Learning) is, as its name implies, to Transfer the trained model (pre-trained model) parameters to a new model (image classification model) to optimize the new model training. Because most of data and tasks are relevant, parameters of the pre-trained model (which can also be understood as knowledge learned by the pre-trained model) can be migrated to a new model in some way through migration learning, and therefore the learning efficiency of the model is accelerated and optimized.
Step S220: inputting the images uploaded by image scanning into the image classification model for identification and classification; the image scanning and uploading mechanism adopts a multi-channel data acquisition mechanism.
The data acquisition mechanism of the multi-channel can be used for simultaneously acquiring data uploaded by a plurality of terminals, and the terminals can be located in different regions, so that the work of uploading image data is more convenient. The terminal can be a printer with a scanning function, a scanner or an intelligent terminal with a scanning function, such as a mobile phone and a tablet computer.
The multi-channel data acquisition mechanism can facilitate uploading of image data, and particularly, the multi-channel data acquisition mechanism is applied to a mobile phone and a tablet personal computer commonly used by a user, so that the image data uploaded in a business process is more timely, and the working efficiency is improved.
In this embodiment, the data acquisition mechanism of multichannel can deal with the condition that needs a large amount of data to upload, when needing a large amount of data to upload, can use many terminals to upload simultaneously to utilize image classification model to classify the data and upload, the material resources of using manpower sparingly have also accelerated the job schedule of business process, improve work efficiency simultaneously.
In this embodiment, the image data acquired by using the multi-channel data acquisition mechanism includes: images, business affiliated departments, uploads persons, uploading time and uploading models.
The image can be the certificate and document information needed by each business; the certificate and document information required by each service is different. For example, vehicle insurance needs driving license, identity card, etc., and medical insurance needs admission card, medical bill, medical list, identity card, etc.
The uploading machine type refers to a machine type used by an uploader for uploading images, such as a printer, a scanner, a mobile phone, a tablet computer and the like.
The image classification model classifies the input images, the features of the images are matched with the features of the image classification learned by the image classification model, and when the matching result is greater than a preset threshold value, the input images are judged to belong to the corresponding image classification.
Step S230: and when the uploaded image data classification meets one of the preset classifications of the images required by the corresponding service, inserting the uploaded image data into a first operation table.
Step S240: and when all the image data corresponding to the preset classification of the images required by the service in the first operation table are uploaded, generating a workflow pushing instruction by the first operation table.
Step S250: and performing idempotent processing on the push workflow instruction, and writing the push workflow instruction into a message queue.
Step S260: and when the business system monitors the push workflow instruction in the message queue, executing the next operation of the business system.
Compared with the first embodiment, the second embodiment further includes step S210 and step S220, and other steps are the same as those of the first embodiment and are not repeated.
Referring to fig. 4, fig. 4 is a flowchart illustrating an image classification model training process of the intelligent push workflow of the present application. The pre-training model is a trained model, experience parameters or learned knowledge of the model can be transferred to the image classification model, the user-defined marked image data is added for fine tuning, the characteristics of the user-defined marked image data are added to the pre-training model, and therefore a new image classification model is generated, wherein fine tuning is achieved by retraining the preset part of the pre-training model by the characteristics of the user-defined marked image data, and the effect of adjusting the parameters of the pre-training model after the characteristics of the user-defined marked image data are added is achieved.
In the above embodiment, there are advantageous effects: the method for training the image classification model is provided specifically, the classification effect of the image classification model is guaranteed, and therefore the correctness of generation of the push workflow instruction is guaranteed.
Referring to fig. 5, fig. 5 is a specific implementation step of step S210 in the second embodiment of the intelligent workflow promoting method of the present application, where the training of part of the preset network of the pre-training model using the image data labeled by the user-defined method includes:
step S211: the parameters of the frozen network are fixed.
The parameters include at least one of: class _ num, image _ size, img _ channels, batch _ size, iterations, total _ epoch, weight _ decay, drop _ rate, momentum _ rate.
The parameters of the freeze network may be class _ num ═ 10, image _ size ═ 32, img _ channels ═ 3, batch _ size ═ 250, iterations ═ 200, total _ epoch ═ 200, weight _ decay ═ 0.0003, and drop _ rate ═ 0.5.
Step S212: and training the non-freezing network by using the image data labeled by the user.
In this embodiment, the pre-training model can be divided into a frozen network and a non-frozen network, wherein the learned parameters in the frozen network are fixed, i.e. the operation of updating the parameters is not performed in the back propagation. And the image data marked by the user-defined is used for training the non-freezing network, and the operation of updating the parameters is executed during the back propagation in the training process.
In the above embodiment, there are advantageous effects: the implementation step of training part of the preset network of the pre-training model by using the image data labeled by the user is provided, so that the utilization of the experience value of the pre-training model is ensured, meanwhile, the image data labeled by the user is used for training the non-freezing network, the characteristic of the image data labeled by the user is added, and the classification effect of the image classification model is ensured.
Referring to fig. 6, fig. 6 is another specific implementation step of step S210 in the second embodiment of the intelligent workflow promoting method of the present application, where the training of the part of the preset network of the pre-training model using the image data labeled by the user-defined method further includes:
step S211': training a freezing network and a non-freezing network by using the image data labeled by a user; wherein the parameters of the frozen network are dynamically adjusted according to a training process.
In this embodiment, the image data labeled by the user-defined method can be used to perform fine adjustment on the freezing network and the non-freezing network, and the parameters of the freezing layer and the non-freezing layer are updated through back propagation in the training process.
In the above embodiment, there are advantageous effects: and another method for training part of the preset network of the pre-training model is provided, so that the classification effect of the image classification model is ensured.
In one embodiment, the step of training a part of the preset network of the pre-training model by using the image data labeled by the user, wherein the step of updating the parameters of the image classification model by using a back propagation method includes:
step S211 a: comparing the recognition result generated by training with the image data labeled by the user-defined label;
and the recognition result has an error with the image data of the user-defined label, the error between the recognition result and the user-defined label result is calculated through comparison, and the error is reversely propagated from the output layer to the hidden layer until the error is propagated to the input layer.
Step S212 a: and reversely propagating the comparison result, and updating the internal parameters of the image classification model.
And in the back propagation process, adjusting the value of the parameter in the image classification model according to the error, and continuously iterating the process until convergence.
In the above embodiment, there are advantageous effects: specifically, an implementation step of training and updating the image classification model in a back propagation mode in the training process is provided, so that the training effect of the image classification model is ensured, and the recognition effect of the image classification model is ensured.
In one embodiment, the process of constructing the custom labeled image data includes:
step S211 b: performing denoising preprocessing on the image data in the database;
data noise refers to data variations that cannot be accounted for in a set of data, i.e., some data that is not consistent with other data. Data noise exists in the database, and the training effect can be influenced in the training process.
In this embodiment, there may be unclear influence data collected by the service in the database, or pictures unrelated to the current service, and a manual deletion operation is performed. The data noise affects the quality of the custom training set, and the accuracy of the image classification model is affected.
Step S212 b: and marking according to the service preset label.
And marking the existing image data in the database according to a preset service tag, wherein the tag is a corresponding document in the services of an identity card, a driving license, a hospitalization certificate, a diagnosis and treatment sheet, a medical list, a birth certificate and the like.
The number of tags and the names of the tags are not limited herein, and are dynamically adjusted according to the related services.
In one embodiment, the idempotent processing includes at least one of:
globally unique identification, token mechanism, pessimistic lock, and optimistic lock.
In this implementation, a globally unique identifier may be added to avoid repeated sending or receiving of messages.
And after the push workflow instruction is subjected to idempotent processing, the uniqueness of the push workflow instruction written in the MQ message queue is ensured.
The Token mechanism may be Token mechanism: duplicate submissions of pages, i.e., duplicate generation of messages, can be prevented such that duplicate messages appear in the message queue.
Pessimistic locks: when the message data is sent, locking is carried out, pessimistic locks are generally used along with transactions, the locking time of the message data is possibly long, and the locking time is selected according to actual conditions.
An optimistic lock: when receiving message data, the locking receiving is carried out, the message queue is locked at the moment when receiving a new message needing to be updated, the message queue is not locked at other moments, and the efficiency is higher.
The processing method is not limited to the above-described processing method, and other methods capable of performing idempotent processing may be used.
In the above embodiment, there are advantageous effects: a specific method for idempotent processing is provided, and uniqueness of the work instruction is guaranteed, so that business process errors caused by repeated promotion of business work are avoided.
In one embodiment, the preset training model uses a VGG-19 network, the frozen network is a first preset number of convolutional layers, and the non-frozen network is a second preset number of convolutional layers.
The pre-training model may employ a VGG19 network, where the first twelve convolutional layers are frozen networks and the thirteenth through sixteenth convolutional layers are non-frozen networks.
In another embodiment, the pre-training model may employ a VGG-19 network, with the first eight layers being frozen networks and the ninth through sixteenth layers being non-frozen networks.
Wherein as shown in table 1: conv3-64 shows that the size of a convolution kernel is 3 x 3, the number of the convolution kernels is 64, after one convolution kernel sweeps image matrix data, a new matrix is generated, and 64 layers of new matrices can be generated by 64 convolution kernels; the Conv3-128 indicates that the size of a convolution kernel is 3 x 3, the number of the convolution kernels is 128, the Conv3-256 indicates that the size of the convolution kernel is 3 x 3, the number of the convolution kernels is 256, the Conv-518 indicates that the size of the convolution kernel is 3 x 3, and the number of the convolution kernels is 518, wherein the activation functions of the convolution layer and the full connection layer both use relu, the Max Pool layer uses maximum pooling, the pooled small matrix is 2 x 2, and after the pooling, the length and the width of the matrix are reduced by half. After the corresponding convolution and corresponding pooling operations are performed, flattening the data into one-dimensional vectors using Flatten (); in two full connection layers FC-4096, the number of neurons in the full connection layer is 4096, where 4096 is an empirical value given by VGG16, and can also be dynamically adjusted and modified; the number of neurons of the last full connection layer FC-100 is 100, which is designed according to a task, in this embodiment, the number of image classifications required by a service is preset to 100, that is, the number of all image classifications is 100, for example, vehicle insurance requires a driving license, an identity card, etc., medical insurance requires a hospitalization license, a diagnosis and treatment sheet, a medical list, an identity card, etc., and the number of classifications of all uploaded image data is 100.
TABLE 1
The preset number of the classified images is 100, which is not limited herein, and is dynamically adjusted according to the corresponding service.
To achieve the above object, there is also provided a computer readable storage medium having stored thereon an intelligent push workflow method program, which when executed by a processor, implements the steps of any of the methods described above.
To achieve the above object, there is also provided an intelligent push workflow device, including a memory, a processor, and an intelligent push workflow method program stored on the memory and executable on the processor, where the processor executes the intelligent push workflow method program to implement any of the steps of the method described above.
There is one specific embodiment that can be implemented as follows: the uploading person A uploads the images 1, 2 and 3 on the mobile phone terminal, and after the images are identified by the image classification model, the image 1 belongs to the preset classification of the identity card, the image 2 belongs to the preset classification of the driving license, and the image 3 belongs to the preset classification of the driving license; the method comprises the steps that image 1 meets the preset classification needed by a corresponding service, corresponding image data is added into a first operation table, images 2 and 3 also meet the preset classification needed by the corresponding service, the corresponding image data is added into the first operation table, when the first operation table monitors that all the image data needed by the corresponding service are uploaded, a workflow pushing instruction is generated, then power-equal operation is conducted on the workflow pushing instruction and added into an MQ message queue, and the uniqueness of messages in the MQ message queue is guaranteed through the power-equal operation, so that the correctness in the process of pushing the corresponding service flow is guaranteed.
There is another specific example that can be implemented as follows: an uploader A uploads images 1 and 2 at a mobile phone end of a region A, an uploader B uploads images 3 and 4 at a scanner end of a region B (a multi-channel data acquisition mechanism can ensure that different terminals in different regions can work simultaneously, the working difficulty is reduced, and the working efficiency is improved), an image classification model can be transplanted to a corresponding uploading terminal, after the image classification model is identified, the image 1 belongs to a hospitalization certificate preset classification, the image 2 is a diagnosis and treatment sheet preset classification, the image 3 belongs to an identity card preset classification, the image 4 belongs to a medical list preset classification, the image data 1, 2, 3 and 4 meet the preset classification required by corresponding insurance business, corresponding image data is added into a first operation table, when the first operation table monitors that all the image data required by the corresponding business are uploaded, a workflow pushing instruction is generated, and then the workflow pushing instruction is subjected to idempotent operation, the method is added into the MQ message queue, and the uniqueness of the messages in the MQ message queue is ensured through idempotent operation, so that the correctness is ensured when the corresponding business process is promoted.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It should be noted that in the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The usage of the words first, second and third, etcetera do not indicate any ordering. These words may be interpreted as names.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the invention.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
Claims (10)
1. An intelligent push workflow method, the method comprising:
when the uploaded image classification meets one of preset classifications of images required by corresponding services, inserting image data corresponding to the uploaded image into a first operation table;
when all the image data corresponding to the preset classification of the images required by the business in the first operation table are uploaded, generating a workflow pushing instruction by the first operation table;
performing idempotent processing on the push workflow instruction, and writing the push workflow instruction into a message queue;
and when the business system monitors the push workflow instruction in the message queue, executing the next operation of the business system.
2. The intelligent workflow promoting method according to claim 1, wherein before the step of inserting the image data corresponding to the uploaded image into the first operation table when the uploaded image classification satisfies one of the preset classifications of images required by the corresponding business, the method comprises:
training part of the preset network of the pre-training model by using the image data labeled by the user-defined label to generate a corresponding image classification model;
inputting the images uploaded by image scanning into the image classification model for identification and classification; the image scanning and uploading mechanism adopts a multi-channel data acquisition mechanism.
3. The intelligent workflow-driven method of claim 2, wherein the training of the pre-set network of pre-trained models using custom labeled image data comprises:
fixing parameters of the frozen network;
and training the non-freezing network by using the image data labeled by the user.
4. The intelligent workflow-driven method of claim 2, wherein the training of the portion of the pre-defined network of pre-trained models using custom labeled image data, further comprises:
training a freezing network and a non-freezing network by using the image data labeled by a user; wherein the parameters of the frozen network are dynamically adjusted according to a training process.
5. The intelligent workflow-driven method of claim 4, wherein the step of training the partial predetermined network of pre-trained models with custom labeled image data comprises updating parameters of the image classification models in a back propagation manner, and comprises:
comparing the recognition result generated by training with the image data labeled by the user-defined label;
and reversely propagating the comparison result, and updating the internal parameters of the image classification model.
6. The intelligent workflow-driven method of claim 2, wherein the process of constructing the custom labeled image data comprises:
performing denoising preprocessing on the image data in the database;
and marking according to the service preset label.
7. The intelligent pushed workflow method of claim 1 wherein the idempotent processing comprises at least one of:
globally unique identification, token mechanism, pessimistic lock, and optimistic lock.
8. The intelligent pushed workflow method of claim 2 wherein the pre-defined training model employs a VGG-19 network, the frozen network being a first pre-defined number of convolutional layers and the non-frozen network being a second pre-defined number of convolutional layers.
9. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon an intelligent push workflow method program which, when executed by a processor, implements the steps of the method of any one of claims 1 to 8.
10. An intelligent push workflow device comprising a memory, a processor and an intelligent push workflow method program stored on said memory and executable on said processor, said processor implementing the steps of the method of any one of claims 1 to 8 when executing said intelligent push workflow method program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011492148.3A CN112561480A (en) | 2020-12-16 | 2020-12-16 | Intelligent workflow pushing method, equipment and computer storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011492148.3A CN112561480A (en) | 2020-12-16 | 2020-12-16 | Intelligent workflow pushing method, equipment and computer storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112561480A true CN112561480A (en) | 2021-03-26 |
Family
ID=75064383
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011492148.3A Pending CN112561480A (en) | 2020-12-16 | 2020-12-16 | Intelligent workflow pushing method, equipment and computer storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112561480A (en) |
Citations (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20110102777A (en) * | 2010-03-11 | 2011-09-19 | 악사손해보험 주식회사 | Insurance work system through ordered workflow |
| US20120221728A1 (en) * | 2011-02-25 | 2012-08-30 | International Business Machines Corporation | Administering Medical Digital Images With Intelligent Analytic Execution Of Workflows |
| CN102970353A (en) * | 2012-11-08 | 2013-03-13 | 大唐软件技术股份有限公司 | Method and system for business data processing |
| US20130253976A1 (en) * | 2012-03-20 | 2013-09-26 | Jagdeesh Shukla | System and method for processing electronic mails in a high volume shared services environment for initiating and processing transactions |
| US9058580B1 (en) * | 2012-01-12 | 2015-06-16 | Kofax, Inc. | Systems and methods for identification document processing and business workflow integration |
| CN107153822A (en) * | 2017-05-19 | 2017-09-12 | 北京航空航天大学 | A kind of smart mask method of the semi-automatic image based on deep learning |
| CN108805569A (en) * | 2018-05-29 | 2018-11-13 | 阿里巴巴集团控股有限公司 | Transaction processing method and device, electronic equipment based on block chain |
| CN108805697A (en) * | 2018-06-12 | 2018-11-13 | 平安科技（深圳）有限公司 | Loan transaction processing method, device, computer equipment and readable storage medium storing program for executing |
| CN109034280A (en) * | 2018-06-04 | 2018-12-18 | 平安科技（深圳）有限公司 | Handwriting model training method, hand-written character recognizing method, device, equipment and medium |
| CN109189389A (en) * | 2018-08-16 | 2019-01-11 | 阿里巴巴集团控股有限公司 | Data processing method and device, server |
| CN109359809A (en) * | 2018-09-03 | 2019-02-19 | 中国平安人寿保险股份有限公司 | Task distribution method, system, computer equipment and storage medium |
| US20190073569A1 (en) * | 2017-09-07 | 2019-03-07 | International Business Machines Corporation | Classifying medical images using deep convolution neural network (cnn) architecture |
| WO2019051797A1 (en) * | 2017-09-15 | 2019-03-21 | 广东欧珀移动通信有限公司 | Image processing method and apparatus, computer device, and computer-readable storage medium |
| CN109543502A (en) * | 2018-09-27 | 2019-03-29 | 天津大学 | A kind of semantic segmentation method based on the multiple dimensioned neural network of depth |
| CN109544096A (en) * | 2018-10-18 | 2019-03-29 | 国网福建省电力有限公司 | Low pressure industry based on artificial intelligence expands business handling method and device on line |
| CN109671068A (en) * | 2018-12-14 | 2019-04-23 | 复旦大学附属中山医院 | A kind of abdominal muscles mask method and device based on deep learning |
| CN109685690A (en) * | 2018-12-12 | 2019-04-26 | 南宁市国土测绘地理信息中心 | A kind of real estate self-service declaration system |
| CN109726074A (en) * | 2018-08-31 | 2019-05-07 | 网联清算有限公司 | Log processing method, device, computer equipment and storage medium |
| CN110543815A (en) * | 2019-07-22 | 2019-12-06 | 平安科技（深圳）有限公司 | Training method of face recognition model, face recognition method, device, equipment and storage medium |
| CN110674091A (en) * | 2019-09-30 | 2020-01-10 | 深圳前海环融联易信息科技服务有限公司 | File uploading method and system based on artificial intelligence and storage medium |
| CN111563897A (en) * | 2020-04-13 | 2020-08-21 | 北京理工大学 | Breast nuclear magnetic image tumor segmentation method and device based on weak supervised learning |
| CN111563481A (en) * | 2020-06-09 | 2020-08-21 | 中国农业银行股份有限公司山东省分行 | Intelligent authorization method and device for bank counter business |
| US20200315587A1 (en) * | 2019-04-04 | 2020-10-08 | Koninklijke Philips N.V. | Data augmentation for training deep learning models with ultrasound images |
| CN111897825A (en) * | 2020-06-01 | 2020-11-06 | 中国人民财产保险股份有限公司 | Distributed transaction processing method and device |
| CN111950340A (en) * | 2020-06-05 | 2020-11-17 | 浙江工业大学 | Face convolution neural network feature expression learning and extracting method suitable for mask wearing |
| CN112036342A (en) * | 2020-09-03 | 2020-12-04 | 中国平安财产保险股份有限公司 | Document snapshot method, device and computer storage medium |
-
2020
- 2020-12-16 CN CN202011492148.3A patent/CN112561480A/en active Pending
Patent Citations (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20110102777A (en) * | 2010-03-11 | 2011-09-19 | 악사손해보험 주식회사 | Insurance work system through ordered workflow |
| US20120221728A1 (en) * | 2011-02-25 | 2012-08-30 | International Business Machines Corporation | Administering Medical Digital Images With Intelligent Analytic Execution Of Workflows |
| US9058580B1 (en) * | 2012-01-12 | 2015-06-16 | Kofax, Inc. | Systems and methods for identification document processing and business workflow integration |
| US20130253976A1 (en) * | 2012-03-20 | 2013-09-26 | Jagdeesh Shukla | System and method for processing electronic mails in a high volume shared services environment for initiating and processing transactions |
| CN102970353A (en) * | 2012-11-08 | 2013-03-13 | 大唐软件技术股份有限公司 | Method and system for business data processing |
| CN107153822A (en) * | 2017-05-19 | 2017-09-12 | 北京航空航天大学 | A kind of smart mask method of the semi-automatic image based on deep learning |
| US20190073569A1 (en) * | 2017-09-07 | 2019-03-07 | International Business Machines Corporation | Classifying medical images using deep convolution neural network (cnn) architecture |
| WO2019051797A1 (en) * | 2017-09-15 | 2019-03-21 | 广东欧珀移动通信有限公司 | Image processing method and apparatus, computer device, and computer-readable storage medium |
| CN108805569A (en) * | 2018-05-29 | 2018-11-13 | 阿里巴巴集团控股有限公司 | Transaction processing method and device, electronic equipment based on block chain |
| CN109034280A (en) * | 2018-06-04 | 2018-12-18 | 平安科技（深圳）有限公司 | Handwriting model training method, hand-written character recognizing method, device, equipment and medium |
| CN108805697A (en) * | 2018-06-12 | 2018-11-13 | 平安科技（深圳）有限公司 | Loan transaction processing method, device, computer equipment and readable storage medium storing program for executing |
| CN109189389A (en) * | 2018-08-16 | 2019-01-11 | 阿里巴巴集团控股有限公司 | Data processing method and device, server |
| CN109726074A (en) * | 2018-08-31 | 2019-05-07 | 网联清算有限公司 | Log processing method, device, computer equipment and storage medium |
| CN109359809A (en) * | 2018-09-03 | 2019-02-19 | 中国平安人寿保险股份有限公司 | Task distribution method, system, computer equipment and storage medium |
| CN109543502A (en) * | 2018-09-27 | 2019-03-29 | 天津大学 | A kind of semantic segmentation method based on the multiple dimensioned neural network of depth |
| CN109544096A (en) * | 2018-10-18 | 2019-03-29 | 国网福建省电力有限公司 | Low pressure industry based on artificial intelligence expands business handling method and device on line |
| CN109685690A (en) * | 2018-12-12 | 2019-04-26 | 南宁市国土测绘地理信息中心 | A kind of real estate self-service declaration system |
| CN109671068A (en) * | 2018-12-14 | 2019-04-23 | 复旦大学附属中山医院 | A kind of abdominal muscles mask method and device based on deep learning |
| US20200315587A1 (en) * | 2019-04-04 | 2020-10-08 | Koninklijke Philips N.V. | Data augmentation for training deep learning models with ultrasound images |
| CN110543815A (en) * | 2019-07-22 | 2019-12-06 | 平安科技（深圳）有限公司 | Training method of face recognition model, face recognition method, device, equipment and storage medium |
| CN110674091A (en) * | 2019-09-30 | 2020-01-10 | 深圳前海环融联易信息科技服务有限公司 | File uploading method and system based on artificial intelligence and storage medium |
| CN111563897A (en) * | 2020-04-13 | 2020-08-21 | 北京理工大学 | Breast nuclear magnetic image tumor segmentation method and device based on weak supervised learning |
| CN111897825A (en) * | 2020-06-01 | 2020-11-06 | 中国人民财产保险股份有限公司 | Distributed transaction processing method and device |
| CN111950340A (en) * | 2020-06-05 | 2020-11-17 | 浙江工业大学 | Face convolution neural network feature expression learning and extracting method suitable for mask wearing |
| CN111563481A (en) * | 2020-06-09 | 2020-08-21 | 中国农业银行股份有限公司山东省分行 | Intelligent authorization method and device for bank counter business |
| CN112036342A (en) * | 2020-09-03 | 2020-12-04 | 中国平安财产保险股份有限公司 | Document snapshot method, device and computer storage medium |
Non-Patent Citations (2)
| Title |
|---|
| 刘丰威;潘炜;: "基于场景的图像文字稽核系统设计与实现", 电子技术与软件工程, no. 09, 1 May 2020 (2020-05-01) * |
| 陶源;: "浅谈任务管理系统中任务调度的实现", 科技广场, no. 05, 18 May 2006 (2006-05-18) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11922308B2 (en) | Generating neighborhood convolutions within a large network | |
| US11669744B2 (en) | Regularized neural network architecture search | |
| DE112017006166T5 (en) | METHOD AND SYSTEM FOR GENERATING A MULTI-RELEVANT LABEL | |
| DE202020101701U1 (en) | Training neural networks using data augmentation strategies | |
| WO2023040506A1 (en) | Model-based data processing method and apparatus, electronic device, computer-readable storage medium, and computer program product | |
| US20220108478A1 (en) | Processing images using self-attention based neural networks | |
| DE112015005985T5 (en) | CLASSIFICATION AND STORAGE OF DOCUMENTS | |
| US20220375211A1 (en) | Multi-layer perceptron-based computer vision neural networks | |
| CN110264407B (en) | Image super-resolution model training and reconstruction method, device, equipment and storage medium | |
| WO2019223145A1 (en) | Electronic device, promotion list recommendation method and system, and computer-readable storage medium | |
| EP4095758A1 (en) | Training large-scale vision transformer neural networks | |
| CN106874922B (en) | Method and device for determining service parameters | |
| CN113159195A (en) | Ultrasonic image classification method, system, electronic device and storage medium | |
| CN112995414B (en) | Behavior quality inspection method, device, equipment and storage medium based on voice call | |
| US11361032B2 (en) | Computer driven question identification and understanding within a commercial tender document for automated bid processing for rapid bid submission and win rate enhancement | |
| DE102021107510A1 (en) | TRAINING OF A NEURAL NETWORK UNDER MEMORY RESTRICTION | |
| CN112561480A (en) | Intelligent workflow pushing method, equipment and computer storage medium | |
| CN111177282A (en) | Prediction method for integrating attention mechanism | |
| Yang et al. | Finding badly drawn bunnies | |
| CN113159202B (en) | Image classification method, device, electronic equipment and storage medium | |
| WO2021174869A1 (en) | User image data processing method, apparatus, computer device, and storage medium | |
| EP2846499B1 (en) | Method And Device For Classifying A Message | |
| US20210318840A1 (en) | Printing relevant content | |
| US20220172066A1 (en) | End-to-end training of neural networks for image processing | |
| US20240062051A1 (en) | Hierarchical data labeling for machine learning using semi-supervised multi-level labeling framework |
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 |