CN112906554A - Model training optimization method and device based on visual image and related equipment - Google Patents
Model training optimization method and device based on visual image and related equipment Download PDFInfo
- Publication number
- CN112906554A CN112906554A CN202110182525.1A CN202110182525A CN112906554A CN 112906554 A CN112906554 A CN 112906554A CN 202110182525 A CN202110182525 A CN 202110182525A CN 112906554 A CN112906554 A CN 112906554A
- Authority
- CN
- China
- Prior art keywords
- model
- training
- visual
- target
- image data
- 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.)
- Granted
Links
- 230000000007 visual effect Effects 0.000 title claims abstract description 189
- 238000012549 training Methods 0.000 title claims abstract description 145
- 238000005457 optimization Methods 0.000 title claims abstract description 80
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000012545 processing Methods 0.000 claims abstract description 8
- 238000004590 computer program Methods 0.000 claims description 17
- 238000004364 calculation method Methods 0.000 claims description 13
- 238000004458 analytical method Methods 0.000 claims description 8
- 230000006870 function Effects 0.000 description 13
- 230000011218 segmentation Effects 0.000 description 13
- 238000001514 detection method Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000007405 data analysis Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 230000005477 standard model Effects 0.000 description 1
- 238000002054 transplantation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the field of computer vision processing, and discloses a model training optimization method and device based on a visual image, computer equipment and a storage medium, wherein the method comprises the following steps: acquiring visual image data and user configuration parameters; determining a training target for the visual image data according to the user configuration parameters; selecting a visual piece corresponding to the training target from a preset visual piece set as a target visual piece; determining a preset model according to the target visual piece; the visual image data is adopted to train the preset model to obtain a training optimization model, and the method improves the generation efficiency of the training task model and reduces the complexity.
Description
Technical Field
The invention relates to the field of computer vision processing, in particular to a model training optimization method and device based on visual images, computer equipment and a storage medium.
Background
In the prior art, a common deep learning method is TensorFlow (symbolic mathematical system based on data flow programming), PyTorch (open source Python machine learning library).
However, for computer vision problems, a TensorFlow and PyTorch deep learning method is used, a model needs to be trained independently according to the visual task requirement of each user, and with the continuous improvement of task complexity, a mainstream framework system becomes complex due to the reasons of structural design, continuous updating and the like, and the structural optimization and transplantation become more difficult, so that the generation efficiency of the training task model becomes low and the complexity becomes high.
Therefore, the existing method has the problems of low generation efficiency and high complexity of the training task model caused by diversified visual task requirements.
Disclosure of Invention
The embodiment of the invention provides a model training optimization method and device based on visual images, computer equipment and a storage medium, so that the model training optimization generation efficiency based on the visual images is improved, and the complexity is reduced.
In order to solve the above technical problem, an embodiment of the present application provides a model training optimization method based on visual images, and the model training optimization method based on visual images includes:
acquiring visual image data and user configuration parameters;
determining a training target for the visual image data according to the user configuration parameters;
selecting a visual piece corresponding to the training target from a preset visual piece set as a target visual piece;
determining a preset model according to the target visual piece;
and training the preset model by adopting the visual image data to obtain a training optimization model.
In order to solve the above technical problem, an embodiment of the present application further provides a model training optimization apparatus based on visual images, including:
the data acquisition module is used for acquiring visual image data and user configuration parameters;
the training target determining module is used for determining a training target for the visual image data according to the user configuration parameters;
the target visual piece determining module is used for selecting a visual piece corresponding to the training target from a preset visual piece set to serve as a target visual piece;
the preset model determining module is used for determining a preset model according to the target visual piece;
and the training optimization module is used for training the preset model by adopting the visual image data to obtain a training optimization model.
In order to solve the above technical problem, an embodiment of the present application further provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor, when executing the computer program, implements the steps of the above method for training optimization based on visual images.
In order to solve the above technical problem, an embodiment of the present application further provides a computer-readable storage medium, where a computer program is stored, and the computer program, when executed by a processor, implements the steps of the method for optimizing model training optimization based on visual images for model training based on visual images.
The model training optimization method, the model training optimization device, the computer equipment and the storage medium based on the visual images provided by the embodiment of the invention acquire visual image data and user configuration parameters; determining a training target for the visual image data according to the user configuration parameters; selecting a visual piece corresponding to a training target from a preset visual piece set as a target visual piece; determining a preset model according to the target visual piece; the preset model is trained by adopting visual image data to obtain a training optimization model, so that the generation efficiency of a training task model is improved, and the complexity is reduced.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments of the present invention will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without inventive labor.
FIG. 1 is a schematic diagram of an application environment of a method for optimizing model training based on visual images according to an embodiment of the present application;
FIG. 2 is a flow diagram of one embodiment of a visual image-based model training optimization method of the present application;
FIG. 3 is a schematic diagram of an embodiment of a visual image-based model training optimization apparatus according to the present application;
FIG. 4 is a schematic block diagram of one embodiment of a computer device according to the present application.
Detailed Description
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "including" and "having," and any variations thereof, in the description and claims of this application and the description of the above figures are intended to cover non-exclusive inclusions. The terms "first," "second," and the like in the description and claims of this application or in the above-described drawings are used for distinguishing between different objects and not for describing a particular order.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The model training optimization method based on the visual images can be applied to the application environment as shown in fig. 1, wherein a client communicates with a server through a network. The client may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices. The server may be implemented as a stand-alone server or as a server cluster consisting of a plurality of servers.
The method for optimizing model training based on visual images provided by the embodiment of the present application is executed by a server, and accordingly, a device for optimizing model training based on visual images is disposed in the server.
Referring to fig. 2, fig. 2 shows a model training optimization method based on visual images according to an embodiment of the present invention, which is described by taking the method applied to the server in fig. 1 as an example, and includes the following steps S101 to S105:
s101, acquiring visual image data and user configuration parameters.
In step S101, it is specifically:
visual image data and user configuration parameters are acquired by the training platform.
The visual image data is training data with label information for training a visual model, and the visual model is a model established for different visual tasks in the visual field.
The training platform is a training platform which trains the visual model based on a component mode and can automatically select a specific visual model. The training platform includes, but is not limited to, a deep learning training platform, a machine learning based training platform. Here, the component-based method means that model training is divided into a visual element, a built-in base model, and a meta-operator. The user can configure parameters for the visual piece and the built-in base model through the training platform.
The user configuration parameters include, but are not limited to, a visual model network structure, an objective function, an optimization method, and the like.
The Visual model Network structure refers to a model used for processing a Visual image task, and includes, but is not limited to, a ResNet (residual Network structure), a VGG (Visual Geometry Group Network, a convolutional neural Network structure based on VGG), a YOLO (a deep neural Network structure based on YOLO), an SSD (Single Shot multi box Detector, a target detection Network structure based on SSD), and the like.
The objective function refers to a function for optimizing random gradient descent in a model training process, and includes but is not limited to softmax maximum value classification, binary cross entropy classification, classification cross entropy, mean square error regression, and join-sense time sequence classification.
The optimization method includes, but is not limited to, SGD (stochastic gradient optimization) and Adam (AdaptiVe Moment Estimation optimization).
It should be noted that, in this embodiment, multiple task models (picture classification, picture detection, and scene segmentation) are preset for tasks in different visual fields, and when model training is required, only a small amount of visual image data with labeling information needs to be provided by a client as sample data, and a fine-tuning technique of migration learning is adopted to increase the capacity of the sample data, so that better model performance is obtained on the basis of a small amount of visual image data, the data amount of network transmission between the client and the server is reduced, and the efficiency of network transmission of the visual image data is improved.
By acquiring the user configuration parameters, various parameter information can be flexibly acquired according to specific requirements of users, for example, a visual model network structure, an objective function and an optimization method, and the visual model is trained through various parameter information, so that the complexity of the training model is reduced.
And S102, determining a training target for the visual image data according to the user configuration parameters.
In step S102, the training targets include, but are not limited to, classification, regression, detection, segmentation, and the like. In step S102, it is specifically:
and the training platform reads the training target in the user configuration parameters and determines a training task for the visual image data according to the training target.
It should be noted here that the training target may be one or more.
For example, when the training target in the user configuration parameter is [ class, segment ], then two training targets may be determined for the visual image data. The visual image data is classified first, and the classified visual image data is divided.
S103, selecting a visual piece corresponding to the training target from a preset visual piece set as a target visual piece.
In step S103, the preset visual element set includes, but is not limited to, a classification visual element, a regression visual element, a detection visual element, and a segmentation visual element.
Classified visuals include, but are not limited to, ResNet-based classified visuals and VGG-based classified visuals, inspection visuals include, but are not limited to, SSD-based inspection visuals and target detection-based visuals (e.g., fast-RCNN inspection visuals), segmentation visuals include, but are not limited to, semantic segmentation-based visuals (e.g., deep Lab segmentation visuals), target segmentation-based visuals (e.g., Mask-RCNN segmentation visuals).
And S104, determining a preset model according to the target visual piece.
In step S104, it should be noted that the preset model is a special task model preset by the training platform according to different training tasks in the visual field, for example, a classification preset model, a regression preset model, a detection preset model, and a segmentation preset model.
In step S104, it specifically includes the following steps a to B:
A. and selecting a built-in base model corresponding to the target visual part from a preset built-in base model set as a target built-in base model.
B. And determining a preset model according to the target visual piece and the target built-in base model.
For the step a, specifically, a built-in base model corresponding to the target visual part is selected from a preset built-in base model set according to the user configuration parameters to serve as the target built-in base model.
It should be noted here that the relationship between the target visual element and the built-in base model is a one-to-many relationship. For example, the classification target visual corresponds to a ResNet-based built-in base model, a DenseNet-based built-in base model.
In step a, the preset built-in base model set includes, but is not limited to, a feature-end built-in base model and a target-end built-in base model.
The feature-end built-in base model includes, but is not limited to, a color feature-based built-in base model, a shape feature-based built-in base model, a texture feature-based built-in base model, and a depth feature-based built-in base model. The target-side built-in base model includes, but is not limited to, a built-in base model based on binary cross entropy classification, a built-in base model based on classification cross entropy, a built-in base model based on mean square error regression, a built-in base model based on binding-sense time-series classification, a built-in base model based on ResNet, and a DenseNet built-in base model.
For the step B, it should be noted here that the relationship between the preset model and the target built-in base model is a one-to-many relationship. For example, the classification preset model corresponds to a ResNet-based built-in basis model and a DenseNet-based built-in basis model.
In a specific example, the target visual element is a classified visual element, and the preset selectable set of built-in base models corresponding to the classified visual element is a built-in base model based on ResNet and a DenseNet built-in base model. In the user configuration parameters, a built-in basis model based on ResNet is selected.
In the step A, according to the user configuration parameters, the target built-in base model corresponding to the target visual element is determined to be a ResNet-based built-in base model.
In the step B, according to the fact that the target visual part is the classified visual part and the target built-in base model is a built-in base model based on ResNet, the preset model is determined to be a classified preset model based on ResNet.
By carrying out layered design on the visual part and the built-in base model, a user can diversify and flexibly configure the model according to respective requirements, so that specific requirements are met, in the parameter configuration process, the user can select a proper tuning method, even if the complexity of the user requirements is continuously improved, the user can be hierarchically selected through the layered design of the visual part and the built-in base model, and the complexity of training the model is reduced.
And S105, training the preset model by adopting visual image data to obtain a training optimization model.
As for the above step S105, it specifically includes the following steps C to D:
C. and determining a meta-operator according to the target built-in base model, wherein the meta-operator carries out calculation processing on the visual image data.
D. And training the visual image data according to all the meta-operators to obtain a training optimization model.
For the step C, the target built-in base model and the meta-operator are in one-to-many relationship, and each meta-operator for realizing the target built-in base model can be determined according to one target built-in base model. The calculation order of each meta-operator corresponds to the method used by the target built-in basis model.
In step C, the above-mentioned meta-operators include, but are not limited to, a compute meta-operator and an index meta-operator.
The compute meta-operators include, but are not limited to, dense compute meta-operators and sparse compute meta-operators. The dense compute meta-operators include, but are not limited to, vector optimization dense compute meta-operators and neural network dense compute meta-operators. Vector optimization dense compute operators include, but are not limited to, matrix-to-matrix multiplication, matrix-to-vector multiplication, and vector-to-vector multiplication. Neural network intensive computational operators include, but are not limited to convolution, pooling, full join.
Index element operators include, but are not limited to, a split index element operator and an accumulate index element operator.
For the step D, specifically, the visual image data is computationally trained according to the computation sequence of the meta-operator, so as to obtain a training optimization model.
The network structure (such as a neural network structure and a residual network structure) formed by combining the computational element operators of different methods enables training of visual image data to be simple and uniform, and is convenient for fusing various computational graphs. Meanwhile, according to the optimization method in the user configuration parameters, the precision of the training optimization model can be automatically verified, and the accuracy of the training optimization model is improved.
In a specific example, if the target built-in basis model is a built-in basis model for extracting texture features based on LBP (local binary feature extraction), the meta-operators are determined to be a segmentation meta-operator, a frequency computation meta-operator, and a full join meta-operator. According to the LBP algorithm used by the target built-in base model, the calculation sequence of the meta-operator can be determined to be a segmentation meta-operator, a frequency calculation meta-operator and a full connection meta-operator.
According to the segmentation meta-operator, the comparison meta-operator, the frequency calculation meta-operator and the full-connection meta-operator, segmenting visual image data, performing frequency calculation on an image obtained through segmentation, and finally performing full-connection operation, so that a training optimization model is obtained.
In this example, step S105 is followed by steps E to G:
E. and acquiring parameter information of the network equipment.
F. And determining a deployment model according to the training optimization model and the parameter information of the network equipment.
G. And applying the deployment model to the network equipment so that the network equipment performs operation corresponding to the training target on the image data according to the deployment model.
For step E, the network devices include, but are not limited to, network devices supporting mainstream hardware and network devices supporting home chips. The network devices supporting the mainstream hardware include but are not limited to network devices based on a graphic processor, network devices based on a central processing unit and network devices based on an instruction set processor, and the network devices supporting the domestic chips include but are not limited to network devices supporting Huashi chips and network devices supporting Bitsuan and Membranan custom chips.
For the step F, it is specifically:
and analyzing the training optimization model to obtain an analysis model.
And according to the parameter information of the network equipment, optimizing and compiling a calculation graph of the analysis model to generate a deployment model.
The analytical model is a network model with inputs and outputs. The network device includes, but is not limited to, a cloud, a server, and various terminal devices.
The optimization of the computation graph on the analytical model includes, but is not limited to, feature fusion of different network layers in the analytical model, and superposition of features of different network layers in the analytical model. And compiling the analysis model after the calculation graph is optimized by adopting a pre-compiling technology according to parameter information of different network equipment so as to obtain a deployment model corresponding to the network equipment.
For the step G, it is specifically:
and transmitting the deployment model to the corresponding network equipment by calling the uniform interface.
And applying the deployment model to the network equipment so that the network equipment performs operation corresponding to the training target on the image data according to the deployment model.
In a specific example, the network device is a mobile phone terminal of a HUAWEI P40 model, and parameter information of the mobile phone terminal includes system Android and a hardware chip ARM. The training target is classified visual image data, and a training optimization model of the classified visual image data is obtained. And analyzing the classified visual image data training optimization model to obtain a classified visual image data analysis model. And transmitting the classified visual image data analysis model to a HUAWEI P40 model mobile phone terminal by calling a uniform interface. And optimizing and compiling a calculation chart of the classified visual image data analysis model according to Android of a parameter information system of a HUAWEI P40 model mobile phone terminal and an ARM (hardware-in-a-ARM) chip, and generating a classified visual image data deployment model. And applying the classified visual image data deployment model to a HUAWEI P40 model mobile phone terminal, wherein the mobile phone terminal can classify the visual image data according to the classified visual image data deployment model.
And acquiring a training optimization model corresponding to a training target through user configuration parameters, wherein the training optimization model is a unified standard model and supports third-party platform conversion. By calling the unified interface, the training optimization model can be applied to various platforms, a cross-platform deployment model is realized, algorithm personnel do not need to call artificial intelligence libraries under different platforms for deployment when the same type of model is deployed on different platforms, and therefore the efficiency of applying the training optimization model across platforms is improved.
In the embodiment, visual image data and user configuration parameters are acquired; determining a training target for the visual image data according to the user configuration parameters; selecting a visual piece corresponding to a training target from a preset visual piece set as a target visual piece; determining a preset model according to the target visual piece; the preset model is trained by adopting visual image data to obtain a training optimization model, so that the generation efficiency of a training task model is improved, and the complexity is reduced.
It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.
Fig. 3 is a schematic block diagram of a visual image-based model training optimization device in one-to-one correspondence with the visual image-based model training optimization method according to the foregoing embodiment. As shown in fig. 3, the visual image-based model training optimization device includes a data acquisition module 11, a training target determination module 12, a target visual element determination module 13, a preset model determination module 14, and a training optimization module 15. The functional modules are explained in detail as follows:
a data acquisition module 11, configured to acquire visual image data and user configuration parameters;
a training target determining module 12, configured to determine a training target for the visual image data according to the user configuration parameter;
a target visual piece determining module 13, configured to select a visual piece corresponding to the training target from a preset visual piece set as a target visual piece;
a preset model determining module 14, configured to determine a preset model according to the target visual element;
and the training optimization module 15 is configured to train the preset model by using the visual image data to obtain a training optimization model.
In one embodiment, the preset model determining module 14 further includes:
the built-in base model determining unit is used for selecting a built-in base model corresponding to the target visual part from a preset built-in base model set to serve as a target built-in base model;
and the preset model determining unit is used for determining a preset model according to the target visual piece and the target built-in base model.
In one embodiment, the training optimization module 15 further comprises:
the meta-operator determining unit is used for determining a meta-operator according to the target built-in base model, wherein the meta-operator performs calculation processing on the visual image data;
and training the visual image data according to all the meta-operators to obtain a training optimization model.
In this example, training the optimization module 15 then comprises:
the parameter information acquisition module is used for acquiring the parameter information of the network equipment;
the deployment model determining module is used for determining a deployment model according to the training optimization model and the parameter information of the network equipment;
and the application module is used for applying the deployment model to the network equipment so that the network equipment performs operation corresponding to the training target on the image data according to the deployment model.
In one embodiment, the deployment model determining module further comprises:
the analysis model determining unit is used for analyzing the training optimization model to obtain an analysis model;
and the deployment model generation unit is used for optimizing and compiling the calculation graph of the analysis model according to the parameter information of the network equipment to generate a deployment model.
For specific limitations of the apparatus for optimizing model training based on visual images, reference may be made to the above limitations of the method for optimizing model training based on visual images, and details thereof are not repeated here. The modules in the model training and optimizing device based on visual images can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 4. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing data involved in the model training optimization method based on the visual images. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of model training optimization based on visual images.
In one embodiment, a computer device is provided, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor when executing the computer program implements the steps of the method for optimizing model training based on visual images in the above embodiments, such as the steps S101 to S105 shown in fig. 2 and other extensions of the method and related steps. Alternatively, the processor, when executing the computer program, implements the functions of the modules/units of the visual image-based model training optimization apparatus in the above embodiments, such as the functions of the modules 11 to 15 shown in fig. 3. To avoid repetition, further description is omitted here.
The Processor may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component, etc. The general purpose processor may be a microprocessor or the processor may be any conventional processor or the like which is the control center for the computer device and which connects the various parts of the overall computer device using various interfaces and lines.
The memory may be used to store the computer programs and/or modules, and the processor may implement various functions of the computer device by running or executing the computer programs and/or modules stored in the memory and invoking data stored in the memory. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data (such as audio data, video data, etc.) created according to the use of the cellular phone, etc.
The memory may be integrated in the processor or may be provided separately from the processor.
In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which, when being executed by a processor, implements the steps of the visual image-based model training optimization method in the above-described embodiments, such as the steps S101 to S105 and other extensions of the method and extensions of related steps shown in fig. 2. Alternatively, the computer program is executed by a processor to implement the functions of the modules/units of the visual image-based model training optimization apparatus in the above-described embodiments, such as the functions of the modules 11 to 15 shown in fig. 3. To avoid repetition, further description is omitted here.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present application may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (such as a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present application.
It is to be understood that the above-described embodiments are merely illustrative of some, but not restrictive, of the broad invention, and that the appended drawings illustrate preferred embodiments of the invention and do not limit the scope of the invention. This application is capable of embodiments in many different forms and is provided for the purpose of enabling a thorough understanding of the disclosure of the application. Although the present application has been described in detail with reference to the foregoing embodiments, it will be apparent to one skilled in the art that the present application may be practiced without modification or with equivalents of some of the features described in the foregoing embodiments. All equivalent structures made by using the contents of the specification and the drawings of the present application are directly or indirectly applied to other related technical fields and are within the protection scope of the present application.
Claims (10)
1. A model training optimization method based on visual images is characterized by comprising the following steps:
acquiring visual image data and user configuration parameters;
determining a training target for the visual image data according to the user configuration parameters;
selecting a visual piece corresponding to the training target from a preset visual piece set as a target visual piece;
determining a preset model according to the target visual piece;
and training the preset model by adopting the visual image data to obtain a training optimization model.
2. The method of claim 1, wherein determining a preset model based on the target visual comprises:
selecting a built-in base model corresponding to the target visual part from a preset built-in base model set as a target built-in base model;
and determining a preset model according to the target visual piece and the target built-in base model.
3. The method of claim 2, wherein said training the pre-set model using the visual image data to obtain a trained optimization model comprises:
determining a meta-operator according to the target built-in basis model, wherein the meta-operator performs calculation processing on the visual image data;
and training the visual image data according to all the meta-operators to obtain a training optimization model.
4. The method of any of claims 1 to 3, wherein after said training said preset model using said visual image data to obtain a trained optimization model, said method further comprises:
acquiring parameter information of network equipment;
determining a deployment model according to the training optimization model and the parameter information of the network equipment;
and applying the deployment model to the network equipment so that the network equipment performs operation corresponding to the training target on the image data according to the deployment model.
5. The method of claim 4, wherein determining a deployment model based on the training optimization model and the parameter information for the network device comprises:
analyzing the training optimization model to obtain an analysis model;
and according to the parameter information of the network equipment, optimizing and compiling a calculation graph of the analysis model to generate a deployment model.
6. A model training optimization device based on visual images is characterized by comprising:
the data acquisition module is used for acquiring visual image data and user configuration parameters;
the training target determining module is used for determining a training target for the visual image data according to the user configuration parameters;
the target visual piece determining module is used for selecting a visual piece corresponding to the training target from a preset visual piece set to serve as a target visual piece;
the preset model determining module is used for determining a preset model according to the target visual piece;
and the training optimization module is used for training the preset model by adopting the visual image data to obtain a training optimization model.
7. The visual image-based model training optimization apparatus of claim 6, wherein the preset model determination module comprises:
the built-in base model determining unit is used for selecting a built-in base model corresponding to the target visual part from a preset built-in base model set to serve as a target built-in base model;
and the preset model determining unit is used for determining a preset model according to the target visual piece and the target built-in base model.
8. The visual image-based model training optimization apparatus of claim 6, wherein the training optimization module thereafter comprises:
the parameter information acquisition module is used for acquiring the parameter information of the network equipment;
the deployment model determining module is used for determining a deployment model according to the training optimization model and the parameter information of the network equipment;
and the application module is used for applying the deployment model to the network equipment so that the network equipment performs operation corresponding to the training target on the image data according to the deployment model.
9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the method for visual image-based model training optimization as claimed in any one of claims 1 to 5 when executing the computer program.
10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out a method for visual-image-based model training optimization according to any one of claims 1 to 5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110182525.1A CN112906554B (en) | 2021-02-08 | 2021-02-08 | Model training optimization method and device based on visual image and related equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110182525.1A CN112906554B (en) | 2021-02-08 | 2021-02-08 | Model training optimization method and device based on visual image and related equipment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112906554A true CN112906554A (en) | 2021-06-04 |
CN112906554B CN112906554B (en) | 2022-12-23 |
Family
ID=76123383
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110182525.1A Active CN112906554B (en) | 2021-02-08 | 2021-02-08 | Model training optimization method and device based on visual image and related equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112906554B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113688992A (en) * | 2021-10-25 | 2021-11-23 | 中电云数智科技有限公司 | Model optimization system and method |
CN114550241A (en) * | 2022-01-28 | 2022-05-27 | 智慧眼科技股份有限公司 | Face recognition method and device, computer equipment and storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109978179A (en) * | 2019-04-04 | 2019-07-05 | 拉扎斯网络科技(上海)有限公司 | Model training method and device, electronic equipment and readable storage medium |
CN110047044A (en) * | 2019-03-21 | 2019-07-23 | 深圳先进技术研究院 | A kind of construction method of image processing model, device and terminal device |
CN110517071A (en) * | 2019-08-15 | 2019-11-29 | 中国平安财产保险股份有限公司 | Information forecasting method, device, equipment and storage medium based on machine mould |
CN112085078A (en) * | 2020-08-31 | 2020-12-15 | 深圳思谋信息科技有限公司 | Image classification model generation system, method and device and computer equipment |
EP3754549A1 (en) * | 2019-06-17 | 2020-12-23 | Sap Se | A computer vision method for recognizing an object category in a digital image |
-
2021
- 2021-02-08 CN CN202110182525.1A patent/CN112906554B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110047044A (en) * | 2019-03-21 | 2019-07-23 | 深圳先进技术研究院 | A kind of construction method of image processing model, device and terminal device |
CN109978179A (en) * | 2019-04-04 | 2019-07-05 | 拉扎斯网络科技(上海)有限公司 | Model training method and device, electronic equipment and readable storage medium |
EP3754549A1 (en) * | 2019-06-17 | 2020-12-23 | Sap Se | A computer vision method for recognizing an object category in a digital image |
CN110517071A (en) * | 2019-08-15 | 2019-11-29 | 中国平安财产保险股份有限公司 | Information forecasting method, device, equipment and storage medium based on machine mould |
CN112085078A (en) * | 2020-08-31 | 2020-12-15 | 深圳思谋信息科技有限公司 | Image classification model generation system, method and device and computer equipment |
Non-Patent Citations (1)
Title |
---|
王高祖等: "基于神经网络组合模型的软件可靠性评估研究", 《计算机仿真》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113688992A (en) * | 2021-10-25 | 2021-11-23 | 中电云数智科技有限公司 | Model optimization system and method |
CN114550241A (en) * | 2022-01-28 | 2022-05-27 | 智慧眼科技股份有限公司 | Face recognition method and device, computer equipment and storage medium |
CN114550241B (en) * | 2022-01-28 | 2023-01-31 | 智慧眼科技股份有限公司 | Face recognition method and device, computer equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN112906554B (en) | 2022-12-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109740534B (en) | Image processing method, device and processing equipment | |
CN113065636B (en) | Pruning processing method, data processing method and equipment for convolutional neural network | |
JP7096888B2 (en) | Network modules, allocation methods and devices, electronic devices and storage media | |
CN112906554B (en) | Model training optimization method and device based on visual image and related equipment | |
CN112862828B (en) | Semantic segmentation method, model training method and device | |
CN108924420A (en) | Image capturing method, device, medium, electronic equipment and model training method | |
CN111414915B (en) | Character recognition method and related equipment | |
CN111695596A (en) | Neural network for image processing and related equipment | |
CN111797288A (en) | Data screening method and device, storage medium and electronic equipment | |
CN111652181A (en) | Target tracking method and device and electronic equipment | |
CN111222557A (en) | Image classification method and device, storage medium and electronic equipment | |
CN112241789A (en) | Structured pruning method, device, medium and equipment for lightweight neural network | |
CN114792355B (en) | Virtual image generation method and device, electronic equipment and storage medium | |
CN114037882A (en) | Edge artificial intelligence device, electronic device and method thereof | |
CN106484614A (en) | A kind of method of verification picture processing effect, device and mobile terminal | |
CN111507219A (en) | Action recognition method and device, electronic equipment and storage medium | |
CN114925651A (en) | Circuit routing determination method and related equipment | |
CN114462582A (en) | Data processing method, device and equipment based on convolutional neural network model | |
CN108875901A (en) | Neural network training method and generic object detection method, device and system | |
US20180359378A1 (en) | System, method, and non-transitory computer readable storage medium for image recognition based on convolutional neural networks | |
CN111353577A (en) | Optimization method and device of multi-task-based cascade combination model and terminal equipment | |
CN109685015A (en) | Processing method, device, electronic equipment and the computer storage medium of image | |
CN116579380A (en) | Data processing method and related equipment | |
CN113297973A (en) | Key point detection method, device, equipment and computer readable medium | |
CN113626022A (en) | Object model creating method and device, storage medium and electronic equipment |
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 | ||
CP03 | Change of name, title or address | ||
CP03 | Change of name, title or address |
Address after: No. 205, Building B1, Huigu Science and Technology Industrial Park, No. 336 Bachelor Road, Bachelor Street, Yuelu District, Changsha City, Hunan Province, 410000 Patentee after: Wisdom Eye Technology Co.,Ltd. Country or region after: China Address before: 410000 building 14, phase I, Changsha Zhongdian Software Park, No. 39, Jianshan Road, high tech Development Zone, Changsha City, Hunan Province Patentee before: Wisdom Eye Technology Co.,Ltd. Country or region before: China |