CN111144243A - Household pattern recognition method and device based on counterstudy - Google Patents
Household pattern recognition method and device based on counterstudy Download PDFInfo
- Publication number
- CN111144243A CN111144243A CN201911283893.4A CN201911283893A CN111144243A CN 111144243 A CN111144243 A CN 111144243A CN 201911283893 A CN201911283893 A CN 201911283893A CN 111144243 A CN111144243 A CN 111144243A
- Authority
- CN
- China
- Prior art keywords
- model
- pattern recognition
- initial
- recognition model
- training
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/40—Document-oriented image-based pattern recognition
- G06V30/42—Document-oriented image-based pattern recognition based on the type of document
- G06V30/422—Technical drawings; Geographical maps
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a house pattern recognition method based on antagonistic learning, which adopts a model dimension reduction part established by a convolutional neural network and a model dimension increasing part established by a deconvolution neural network to determine an initial house pattern recognition model, updates model parameters of the initial house pattern recognition model, updates model parameters of an antagonistic model, carries out antagonistic training between the two parts, determines a final house pattern recognition model according to the current parameters of the initial house pattern recognition model when a total cost function reaches a first set standard, adopts the final house pattern recognition model to recognize a house pattern to be recognized, introduces antagonistic behavior according to the characteristics of a generated antagonistic network in the recognition process of the house pattern to be recognized, judges the rationality of the house pattern recognized by the model by using a discriminator, carries out antagonistic learning, and solves the problem that the prior house pattern recognition method only focuses on local information and neglects the overall structure, thereby improving the accuracy of the house pattern recognition.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a family pattern recognition method and device based on counterstudy, computer equipment and a storage medium.
Background
In recent years, with the development of the internet, particularly the mobile internet, a big data age has come, and a large amount of data has been generated. Deep neural networks that require large amounts of data to train have thus far developed. The deep neural network has a great breakthrough in the fields of natural language processing, image recognition, target detection and the like, and the performance of some places exceeds that of human beings. And the method is successfully applied to a plurality of industrial fields, such as voice assistants, unmanned supermarkets, face payment, intelligent transportation, intelligent customer service and the like. However, it is only possible to manufacture it, and it is a real grasp. The generation of the countermeasure network is to make the neural network generate new and satisfactory things according to the knowledge learned by the neural network. The generation of the antagonistic network forms the antagonism through the alternate learning between two neural networks, one for generating is called a generator, and the other for judging the quality of the generation is called a discriminator. The arbiter and generator improve their capabilities by competing against each other. In the house type map recognition task, it is necessary to recognize the types of doors, windows, walls, and rooms. However, doors, windows and walls are a particularly small percentage of the room and are therefore difficult to identify. In the current house type graph recognition method, local information is mainly concerned, such as which category each pixel in the house type graph belongs to, whether the pixel belongs to a door, a window, a wall or a room, and the rationality of classification is not concerned on the overall structure, so that the accuracy of house type graph recognition is easily influenced.
Disclosure of Invention
In view of the above problems, the present invention provides a house type graph recognition method, apparatus, computer device and storage medium based on counterstudy.
In order to achieve the purpose of the invention, the invention provides a family pattern recognition method based on counterstudy, which comprises the following steps:
s10, determining an initial house pattern recognition model according to the model dimension reduction part and the model dimension increasing part by adopting the model dimension reduction part established by the convolutional neural network and the model dimension increasing part established by the deconvolution neural network;
s30, pre-training the initial house type pattern recognition model to update the model parameters of the initial house type pattern recognition model;
s40, pre-training the confrontation model by using the pre-trained initial house type pattern recognition model to update the model parameters of the confrontation model;
s50, performing the pre-training of the initial house type pattern recognition model and the confrontation model after the pre-training of the two models is completed, and determining the final house type pattern recognition model according to the current parameters of the initial house type pattern recognition model when the total cost function reaches the first set standard; the total cost function is the sum of the cost function of the initial house type graph identification model and the cost function of the countermeasure model;
and S60, adopting the final house type graph recognition model to recognize the house type graph to be recognized.
In one embodiment, the pre-training for the initial house pattern recognition model to update the model parameters of the initial house pattern recognition model comprises:
inputting a family pattern training sample into an initial family pattern recognition model, obtaining a probability pattern output by the initial family pattern recognition model, taking the square difference between the probability pattern and the truth value of the family pattern training sample as a cost function of the initial family pattern recognition model, and determining the model parameters of the initial family pattern recognition model after pre-training according to the current model parameters of the initial family pattern recognition model when the cost function of the initial family pattern recognition model reaches a second set standard.
In one embodiment, the pre-training the confrontation model by using the pre-trained initial house pattern recognition model to update the model parameters of the confrontation model comprises:
during training, inputting the house type graph into a house type graph recognition model, and outputting a probability graph by the house type graph recognition model; then the probability chart and the truth value are respectively input into a countermeasure model, and when the probability chart is input into the countermeasure model, the square difference between the output of the countermeasure model and 0 is used as a cost function; when true values are input to the countermeasure model, the squared difference of the output of the countermeasure model and 1 is taken as the cost function
Respectively inputting the probability graph output by the pre-trained initial house type graph recognition model aiming at the house type graph training sample and the truth value of the house type graph training sample into the confrontation model, taking the square difference between the output of the confrontation model and 0 as the cost function of the confrontation model, and determining the model parameters of the confrontation model after pre-training according to the current model parameters of the confrontation model when the cost function of the confrontation model reaches the third set standard.
In one embodiment, before the pre-training for the initial house pattern recognition model to update the model parameters of the initial house pattern recognition model, the method further includes:
and (5) constructing an antagonistic model.
As one embodiment, the constructing the countermeasure model includes:
and establishing a countermeasure model by adopting a convolutional neural network.
As an example, the model structure of the countermeasure model includes [ conv, pool, conv, pool, conv, pool, conv, pool, conv, pool ], wherein conv represents a convolutional layer having a convolution kernel size of 3x 3; pool represents the maximum pooling layer.
A family pattern recognition apparatus based on counterstudy, comprising:
the building module is used for determining an initial house pattern recognition model according to the model dimension reduction part and the model dimension increasing part established by the deconvolution neural network;
the first pre-training module is used for pre-training the initial house pattern recognition model so as to update the model parameters of the initial house pattern recognition model;
the second pre-training module is used for pre-training the confrontation model by utilizing the pre-trained initial house type pattern recognition model so as to update the model parameters of the confrontation model;
the countermeasure training module is used for carrying out countermeasure training between the initial house type pattern recognition model and the countermeasure model after the initial house type pattern recognition model and the countermeasure model are pre-trained, and determining a final house type pattern recognition model according to the current parameters of the initial house type pattern recognition model when the total cost function reaches a first set standard; the total cost function is the sum of the cost function of the initial house type graph identification model and the cost function of the countermeasure model;
and the identification module is used for identifying the user type graph to be identified by adopting the final user type graph identification model.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the counterstudy-based house pattern recognition method of any of the above embodiments when executing the computer program.
A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the counterlearning-based house pattern recognition method of any one of the above embodiments.
The user pattern recognition method based on the antagonistic learning comprises the steps of adopting a model dimension reduction part established by a convolutional neural network, adopting a model dimension increasing part established by a deconvolution neural network to determine an initial user pattern recognition model, pre-training aiming at the initial user pattern recognition model to update model parameters of the initial user pattern recognition model, pre-training an antagonistic model by utilizing the pre-trained initial user pattern recognition model to update model parameters of the antagonistic model, performing antagonistic training between the initial user pattern recognition model and the antagonistic model after the pre-training is completed, determining a final user pattern recognition model according to current parameters of the initial user pattern recognition model when a total cost function reaches a first set standard, recognizing a user pattern to be recognized by adopting the final user pattern recognition model, and generating the characteristics of the antagonistic network in the recognition process aiming at the user pattern to be recognized, and (3) introducing antagonism, judging the rationality of the house pattern identified by the model on the whole structure by using the discriminator, and performing antagonism learning so as to solve the problem that the existing house pattern identification method only focuses on local information and neglects the whole structure, thereby improving the accuracy of the house pattern identification.
Drawings
FIG. 1 is a flow diagram of a method for pattern recognition based on counterlearning according to an embodiment;
FIG. 2 is a schematic structural diagram of a family pattern recognition device based on counterstudy according to an embodiment;
FIG. 3 is a schematic diagram of a computer device of an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
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.
In one embodiment, as shown in fig. 1, there is provided a family pattern recognition method based on counterstudy, comprising the following steps:
and S10, determining an initial house pattern recognition model according to the model dimension reduction part and the model dimension increasing part by adopting the model dimension reduction part established by the convolutional neural network and the model dimension increasing part established by the deconvolution neural network.
Specifically, the model dimension reduction part is mainly used for extracting high-level features hidden in the house pattern, the high-level features are extracted by a neural network and contain information required by house pattern recognition, and the subsequent model dimension increasing part is used for generating a probability heat map output by the house pattern recognition model. The model dimension reduction part adopts a convolution layer part in the VGG16 model, and the model structure of the model dimension reduction part can be [ conv, conv, pool, conv, conv, pool, conv, conv, conv, pool ], wherein conv represents a convolution layer with the convolution kernel size of 3x 3; pool represents the maximum pooling layer, kernel size 2x2, step size 2. This part is primarily dimensionality reduced by the maximum pooling layer.
The model dimension raising part is used for generating a probability heat map of the output of the pattern recognition model. The model dimension-raising part comprises two branches: a room type identification branch and a door and window wall type identification branch. The two branches have basically the same structure and are built by adopting a deconvolution neural network, the structure is [ trans _ conv, trans _ conv, trans _ conv, trans _ conv, biliyet ], wherein trans _ conv represents a deconvolution layer with a convolution kernel size of 4x4, the step length is 2, and the feature map is subjected to dimension raising; the bilinear represents the nearest interpolation operation, and the output of the bilinear interpolation operation is a probability heat map with the same size as the pattern, namely the output of the pattern recognition model. The number of the probability heat maps output by the two branches is the same as the type to be identified, for example, the number of the probability heat maps output by the identification branches of the door and window wall type is the total number of the door and window wall type to be identified.
And S30, pre-training the initial house pattern recognition model to update the model parameters of the initial house pattern recognition model.
In one embodiment, the pre-training for the initial house pattern recognition model to update the model parameters of the initial house pattern recognition model comprises:
inputting a family pattern training sample into an initial family pattern recognition model, obtaining a probability pattern output by the initial family pattern recognition model, taking the square difference between the probability pattern and the truth value of the family pattern training sample as a cost function of the initial family pattern recognition model, and determining the model parameters of the initial family pattern recognition model after pre-training according to the current model parameters of the initial family pattern recognition model when the cost function of the initial family pattern recognition model reaches a second set standard.
The second setting criterion may be a criterion that needs to be achieved by optimizing the cost function of the initial house pattern recognition model by using a gradient descent method.
In one example, the above-mentioned house pattern training samples may include one or more house patterns of known truth values. The truth values include true attribute characteristics for each location of the corresponding floor plan.
Specifically, the user pattern recognition model may be pre-trained by using the R3D public data set, the user pattern training sample is input into the initial user pattern recognition model during training, the initial user pattern recognition model outputs a probability map, a square difference between truth values corresponding to the probability map and the user pattern training sample is used as a cost function (loss), the model is optimized by using a gradient descent method, and parameters of the initial user pattern recognition model are updated.
And S40, pre-training the confrontation model by using the pre-trained initial house type pattern recognition model so as to update the model parameters of the confrontation model.
In one embodiment, the pre-training the confrontation model by using the pre-trained initial house pattern recognition model to update the model parameters of the confrontation model comprises:
during training, inputting the house type graph into a house type graph recognition model, and outputting a probability graph by the house type graph recognition model; then the probability chart and the truth value are respectively input into a countermeasure model, and when the probability chart is input into the countermeasure model, the square difference between the output of the countermeasure model and 0 is used as a cost function; when true values are input to the countermeasure model, the squared difference of the output of the countermeasure model and 1 is taken as the cost function
Respectively inputting the probability graph output by the pre-trained initial house type graph recognition model aiming at the house type graph training sample and the truth value of the house type graph training sample into the confrontation model, taking the square difference between the output of the confrontation model and 0 as the cost function of the confrontation model, and determining the model parameters of the confrontation model after pre-training according to the current model parameters of the confrontation model when the cost function of the confrontation model reaches the third set standard.
The third setting criterion may be a criterion to be achieved for optimization by a gradient descent method with respect to the cost function of the countermeasure model.
Specifically, the embodiment pre-trains the confrontation model by using the pre-trained initial house pattern recognition model. In the pre-training process, the input of the confrontation model comprises two types, one type is a probability chart output by the pre-trained initial house type diagram recognition model, and the other type is a true value corresponding to a house type diagram training sample. During training, a house pattern training sample is firstly input into an initial house pattern recognition model, and the initial house pattern recognition model outputs a probability chart; then the probability chart and the truth value are respectively input into a countermeasure model, and when the probability chart is input into the countermeasure model, the square difference between the output of the countermeasure model and 0 is used as a cost function; when a true value is input to the countermeasure model, the squared difference of the output of the countermeasure model and 1 is taken as the cost function. The optimization method also adopts a gradient descent method.
S50, performing the pre-training of the initial house type pattern recognition model and the confrontation model after the pre-training of the two models is completed, and determining the final house type pattern recognition model according to the current parameters of the initial house type pattern recognition model when the total cost function reaches the first set standard; the total cost function is the sum of the cost function of the initial house pattern recognition model and the cost function of the countermeasure model.
Specifically, the confrontation training between the initial house pattern recognition model and the confrontation model is carried out after the pre-training of the two models is completed. During the fighting training, the fighting model and the initial house pattern recognition model play games, the fighting model tries to distinguish whether the input is a true value or the output of the initial house pattern recognition model, and the initial house pattern recognition model tries to output the same result as the true value, so that the fighting model cannot distinguish the difference between the output and the true value. In the process of the confrontation training, the confrontation model and the initial house pattern recognition model are trained in an alternate updating mode. When the countermeasure model is updated, the cost function is the same as that of the countermeasure model, and the optimization method also adopts a gradient descent method. When the initial house type graph recognition model is updated, the cost function comprises two parts, and one part is the same as the cost function of the initial house type graph recognition model; the other part is called the challenge cost function (i.e. the cost function of the challenge model), which is: firstly, inputting an initial house type graph recognition model by a house type graph training sample, and outputting a probability graph by the initial house type graph recognition model; the probability map is then input to a challenge model whose output is a challenge cost function with the squared difference of 1. The total cost function is the sum of two partial cost functions. The optimization method also adopts a gradient descent method.
And S60, adopting the final house type graph recognition model to recognize the house type graph to be recognized.
Specifically, in the embodiment, an initial house type pattern recognition model is established by using a convolutional neural network and a deconvolution neural network, and training and updating of each stage are performed on the initial house type pattern recognition model to obtain a final house type pattern recognition model. Finally, the input of the pattern recognition model is a pattern to be recognized, the probability heat map of the pattern to be recognized is output, each category corresponds to one heat map, the size of the heat map is the same as that of the input pattern, and each value in the heat map represents the probability that the corresponding position in the input pattern belongs to the category so as to accurately recognize each position of the corresponding pattern to be recognized.
The user pattern recognition method based on the antagonistic learning comprises the steps of adopting a model dimension reduction part established by a convolutional neural network, adopting a model dimension increasing part established by a deconvolution neural network to determine an initial user pattern recognition model, pre-training aiming at the initial user pattern recognition model to update model parameters of the initial user pattern recognition model, pre-training an antagonistic model by utilizing the pre-trained initial user pattern recognition model to update model parameters of the antagonistic model, performing antagonistic training between the initial user pattern recognition model and the antagonistic model after the pre-training is completed, determining a final user pattern recognition model according to current parameters of the initial user pattern recognition model when a total cost function reaches a first set standard, recognizing a user pattern to be recognized by adopting the final user pattern recognition model, and generating the characteristics of the antagonistic network in the recognition process aiming at the user pattern to be recognized, and (3) introducing antagonism, judging the rationality of the house pattern identified by the model on the whole structure by using the discriminator, and performing antagonism learning so as to solve the problem that the existing house pattern identification method only focuses on local information and neglects the whole structure, thereby improving the accuracy of the house pattern identification.
In one embodiment, before the pre-training for the initial house pattern recognition model to update the model parameters of the initial house pattern recognition model, the method further includes:
and (5) constructing an antagonistic model.
As one embodiment, the constructing the countermeasure model includes:
and establishing a countermeasure model by adopting a convolutional neural network.
As an example, the model structure of the countermeasure model includes [ conv, pool, conv, pool, conv, pool, conv, pool, conv, pool ], wherein conv represents a convolutional layer having a convolution kernel size of 3x 3; pool represents the maximum pooling layer.
Specifically, the countermeasure model is used for performing countermeasure learning with the initial house type pattern recognition model, and the countermeasure model mainly detects the rationality of the output of the initial house type pattern recognition model on the whole, so that the initial house type pattern recognition model is forced to output more reasonable prediction on the whole. The input of the confrontation model is two types, one type is the recognition result of the initial house type diagram recognition model, the other type is the true value corresponding to the house type diagram training sample, the true value is also a probability diagram, and each probability diagram corresponds to one category. The output of the confrontation model is a number between 0 and 1, representing the quality of its input, the larger the number the higher the quality, i.e. more reasonable in overall structure.
Further, the countermeasure model can be built by using a convolutional neural network, and the model structure is [ conv, pool, conv, pool, conv, pool, conv, pool, conv, pool ], wherein conv represents a convolutional layer with a convolutional kernel size of 3x 3; pool represents the maximum pooling layer, kernel size 2x2, step size 2.
In one embodiment, by introducing counterlearning during the training of the initial pattern recognition model, the pattern recognition accuracy is improved on the R3D data set, and the comparison results are shown in the following table:
| model (model) | Rate of accuracy |
| House type graph recognition model without countermeasure model | 0.87 |
| House pattern recognition model added with countermeasure model | 0.90 |
Referring to fig. 2, fig. 2 is a schematic structural diagram of a family pattern recognition device based on counterstudy according to an embodiment, including:
the building module 10 is used for determining an initial house pattern recognition model according to the model dimension reduction part and the model dimension increasing part established by the deconvolution neural network;
a first pre-training module 30, configured to pre-train the initial house pattern recognition model to update model parameters of the initial house pattern recognition model;
the second pre-training module 40 is configured to pre-train the countermeasure model with the pre-trained initial house type pattern recognition model to update model parameters of the countermeasure model;
the confrontation training module 50 is used for performing confrontation training between the initial house pattern recognition model and the confrontation model after the pre-training of the initial house pattern recognition model and the confrontation model is completed, and determining a final house pattern recognition model according to the current parameters of the initial house pattern recognition model when the total cost function reaches a first set standard; the total cost function is the sum of the cost function of the initial house type graph identification model and the cost function of the countermeasure model;
and the identifying module 60 is configured to identify the subscriber pattern to be identified by using the final subscriber pattern identification model.
The specific definition of the device for identifying a house pattern based on counterstudy can be referred to the above definition of the method for identifying a house pattern based on counterstudy, and is not described in detail herein. The modules in the above-mentioned antagonistic learning-based house type pattern recognition device can be wholly or partially realized by software, hardware and the 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 terminal, and its internal structure diagram may be as shown in fig. 3. The computer device includes a processor, a memory, a network interface, a display screen, and an input device 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 and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. 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 counterlearning based house pattern recognition method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the architecture shown in fig. 3 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
Based on the examples described above, there is also provided in one embodiment a computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor, when executing the program, implements any of the counterlearning based house pattern recognition methods described in the embodiments above.
It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments may be implemented by a computer program, which may be stored in a non-volatile computer-readable storage medium, and executed by at least one processor in a computer system, as in the embodiments of the present invention, to implement the processes including the embodiments of the above counterlearning-based user pattern recognition method. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
Accordingly, in an embodiment, a computer storage medium and a computer readable storage medium are provided, on which a computer program is stored, wherein the program is executed by a processor to implement any one of the above embodiments of the method for recognizing a house pattern based on counterstudy.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
It should be noted that the terms "first \ second \ third" referred to in the embodiments of the present application merely distinguish similar objects, and do not represent a specific ordering for the objects, and it should be understood that "first \ second \ third" may exchange a specific order or sequence when allowed. It should be understood that "first \ second \ third" distinct objects may be interchanged under appropriate circumstances such that the embodiments of the application described herein may be implemented in an order other than those illustrated or described herein.
The terms "comprising" and "having" and any variations thereof in the embodiments of the present application are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product, or device that comprises a list of steps or modules is not limited to the listed steps or modules but may alternatively include other steps or modules not listed or inherent to such process, method, product, or device.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (9)
1. A house type graph recognition method based on counterstudy is characterized by comprising the following steps:
s10, determining an initial house pattern recognition model according to the model dimension reduction part and the model dimension increasing part by adopting the model dimension reduction part established by the convolutional neural network and the model dimension increasing part established by the deconvolution neural network;
s30, pre-training the initial house type pattern recognition model to update the model parameters of the initial house type pattern recognition model;
s40, pre-training the confrontation model by using the pre-trained initial house type pattern recognition model to update the model parameters of the confrontation model;
s50, performing the pre-training of the initial house type pattern recognition model and the confrontation model after the pre-training of the two models is completed, and determining the final house type pattern recognition model according to the current parameters of the initial house type pattern recognition model when the total cost function reaches the first set standard; the total cost function is the sum of the cost function of the initial house type graph identification model and the cost function of the countermeasure model;
and S60, adopting the final house type graph recognition model to recognize the house type graph to be recognized.
2. The method of claim 1, wherein the pre-training for the initial pattern recognition model to update the model parameters of the initial pattern recognition model comprises:
inputting a family pattern training sample into an initial family pattern recognition model, obtaining a probability pattern output by the initial family pattern recognition model, taking the square difference between the probability pattern and the truth value of the family pattern training sample as a cost function of the initial family pattern recognition model, and determining the model parameters of the initial family pattern recognition model after pre-training according to the current model parameters of the initial family pattern recognition model when the cost function of the initial family pattern recognition model reaches a second set standard.
3. The method of claim 1, wherein the pre-training of the confrontation model with the pre-trained initial house pattern recognition model to update the model parameters of the confrontation model comprises:
during training, inputting the house type graph into a house type graph recognition model, and outputting a probability graph by the house type graph recognition model; then the probability chart and the truth value are respectively input into a countermeasure model, and when the probability chart is input into the countermeasure model, the square difference between the output of the countermeasure model and 0 is used as a cost function; when true values are input to the countermeasure model, the squared difference of the output of the countermeasure model and 1 is taken as the cost function
Respectively inputting the probability graph output by the pre-trained initial house type graph recognition model aiming at the house type graph training sample and the truth value of the house type graph training sample into the confrontation model, taking the square difference between the output of the confrontation model and 0 as the cost function of the confrontation model, and determining the model parameters of the confrontation model after pre-training according to the current model parameters of the confrontation model when the cost function of the confrontation model reaches the third set standard.
4. The method of any one of claims 1 to 3, wherein, in one embodiment, before the pre-training for the initial pattern recognition model to update the model parameters of the initial pattern recognition model, the method further comprises:
and (5) constructing an antagonistic model.
5. The method of claim 4, wherein the constructing the confrontation model comprises:
and establishing a countermeasure model by adopting a convolutional neural network.
6. The method of claim 5, wherein the model structure of the countermeasure model comprises [ conv, pool, conv, pool, conv, pool, conv, pool, conv, pool ], wherein conv represents a convolution layer with convolution kernel size of 3x 3; pool represents the maximum pooling layer.
7. A family pattern recognition apparatus based on counterstudy, comprising:
the building module is used for determining an initial house pattern recognition model according to the model dimension reduction part and the model dimension increasing part established by the deconvolution neural network;
the first pre-training module is used for pre-training the initial house pattern recognition model so as to update the model parameters of the initial house pattern recognition model;
the second pre-training module is used for pre-training the confrontation model by utilizing the pre-trained initial house type pattern recognition model so as to update the model parameters of the confrontation model;
the countermeasure training module is used for carrying out countermeasure training between the initial house type pattern recognition model and the countermeasure model after the initial house type pattern recognition model and the countermeasure model are pre-trained, and determining a final house type pattern recognition model according to the current parameters of the initial house type pattern recognition model when the total cost function reaches a first set standard; the total cost function is the sum of the cost function of the initial house type graph identification model and the cost function of the countermeasure model;
and the identification module is used for identifying the user type graph to be identified by adopting the final user type graph identification model.
8. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the counterlearning based house pattern recognition method of claims 1-6 when executing the computer program.
9. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the counterlearning-based house pattern recognition method of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911283893.4A CN111144243B (en) | 2019-12-13 | 2019-12-13 | Household pattern recognition method and device based on counterstudy |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911283893.4A CN111144243B (en) | 2019-12-13 | 2019-12-13 | Household pattern recognition method and device based on counterstudy |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111144243A true CN111144243A (en) | 2020-05-12 |
| CN111144243B CN111144243B (en) | 2022-07-08 |
Family
ID=70518221
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911283893.4A Active CN111144243B (en) | 2019-12-13 | 2019-12-13 | Household pattern recognition method and device based on counterstudy |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111144243B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113591929A (en) * | 2021-07-05 | 2021-11-02 | 华南师范大学 | Family pattern recognition method combining direction sensing kernel cluster |
| CN114648724A (en) * | 2022-05-18 | 2022-06-21 | 成都航空职业技术学院 | Lightweight efficient target segmentation and counting method based on generation countermeasure network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107742122A (en) * | 2017-10-27 | 2018-02-27 | 浙江大华技术股份有限公司 | The dividing method and device of a kind of x-ray image |
| CN109785252A (en) * | 2018-12-25 | 2019-05-21 | 山西大学 | Based on multiple dimensioned residual error dense network nighttime image enhancing method |
| CN110059750A (en) * | 2019-04-17 | 2019-07-26 | 广东三维家信息科技有限公司 | House type shape recognition process, device and equipment |
-
2019
- 2019-12-13 CN CN201911283893.4A patent/CN111144243B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107742122A (en) * | 2017-10-27 | 2018-02-27 | 浙江大华技术股份有限公司 | The dividing method and device of a kind of x-ray image |
| CN109785252A (en) * | 2018-12-25 | 2019-05-21 | 山西大学 | Based on multiple dimensioned residual error dense network nighttime image enhancing method |
| CN110059750A (en) * | 2019-04-17 | 2019-07-26 | 广东三维家信息科技有限公司 | House type shape recognition process, device and equipment |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113591929A (en) * | 2021-07-05 | 2021-11-02 | 华南师范大学 | Family pattern recognition method combining direction sensing kernel cluster |
| CN113591929B (en) * | 2021-07-05 | 2023-07-25 | 华南师范大学 | House pattern recognition method combining direction perception kernel cluster |
| CN114648724A (en) * | 2022-05-18 | 2022-06-21 | 成都航空职业技术学院 | Lightweight efficient target segmentation and counting method based on generation countermeasure network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111144243B (en) | 2022-07-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110046706B (en) | Model generation method and device and server | |
| CN111416797B (en) | Intrusion detection method for optimizing regularization extreme learning machine by improving longicorn herd algorithm | |
| JP2022141931A (en) | Method and device for training living body detection model, method and apparatus for living body detection, electronic apparatus, storage medium, and computer program | |
| CN113179263A (en) | Network intrusion detection method, device and equipment | |
| CN111144243B (en) | Household pattern recognition method and device based on counterstudy | |
| CN107995370A (en) | Call control method, device and storage medium and mobile terminal | |
| CN112819024B (en) | Model processing method, user data processing method and device and computer equipment | |
| JP2019152964A (en) | Learning method and learning device | |
| CN114261400A (en) | Automatic driving decision-making method, device, equipment and storage medium | |
| CN113656798B (en) | Regularization identification method and device for malicious software tag overturn attack | |
| CN116310386A (en) | Shallow adaptive enhanced context-based method for detecting small central Net target | |
| CN116304341A (en) | Fraud discrimination method and system based on user network big data | |
| CN113743594B (en) | Network traffic prediction model establishment method and device, electronic equipment and storage medium | |
| CN109978058B (en) | Method, device, terminal and storage medium for determining image classification | |
| CN114154622A (en) | Algorithm model for traffic operation system flow data acquisition missing completion | |
| CN117217779A (en) | Training method and device of prediction model and information prediction method and device | |
| CN111626098A (en) | Method, device, equipment and medium for updating parameter values of model | |
| CN114627085A (en) | Target image identification method and device, storage medium and electronic equipment | |
| CN114385846A (en) | Image classification method, electronic device, storage medium and program product | |
| CN111539390A (en) | Small target image identification method, equipment and system based on Yolov3 | |
| CN113191527A (en) | Prediction method and device for population prediction based on prediction model | |
| CN111179070A (en) | Loan risk timeliness prediction system and method based on LSTM | |
| CN110334244A (en) | A kind of method, apparatus and electronic equipment of data processing | |
| Zhou et al. | Tree boosted varying coefficient models | |
| CN112037325B (en) | Method, device, computer equipment and storage medium for constructing semantic map |
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 | ||
| CB02 | Change of applicant information |
Address after: 211100 floor 5, block a, China Merchants high speed rail Plaza project, No. 9, Jiangnan Road, Jiangning District, Nanjing, Jiangsu (South Station area) Applicant after: JIANGSU AIJIA HOUSEHOLD PRODUCTS Co.,Ltd. Address before: 211100 No. 18 Zhilan Road, Science Park, Jiangning District, Nanjing City, Jiangsu Province Applicant before: JIANGSU AIJIA HOUSEHOLD PRODUCTS Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |