CN116137061B - Training method and device for quantity statistical model, electronic equipment and storage medium - Google Patents
Training method and device for quantity statistical model, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN116137061B CN116137061B CN202310424471.4A CN202310424471A CN116137061B CN 116137061 B CN116137061 B CN 116137061B CN 202310424471 A CN202310424471 A CN 202310424471A CN 116137061 B CN116137061 B CN 116137061B
- Authority
- CN
- China
- Prior art keywords
- matrix
- statistic
- existence
- probability
- statistical
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000013179 statistical model Methods 0.000 title claims abstract description 139
- 238000012549 training Methods 0.000 title claims abstract description 65
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000003860 storage Methods 0.000 title claims abstract description 21
- 239000011159 matrix material Substances 0.000 claims abstract description 247
- 230000002776 aggregation Effects 0.000 claims description 6
- 238000004220 aggregation Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 5
- 238000005516 engineering process Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 6
- 230000015654 memory Effects 0.000 description 6
- 238000012360 testing method Methods 0.000 description 5
- 238000004891 communication Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Mathematical Physics (AREA)
- Computational Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Analysis (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Operations Research (AREA)
- Probability & Statistics with Applications (AREA)
- Biophysics (AREA)
- Algebra (AREA)
- Image Analysis (AREA)
Abstract
The present application relates to the field of quantitative statistics technologies, and in particular, to a training method and apparatus for a quantitative statistical model, an electronic device, and a storage medium, where the method includes: acquiring sample image features and whether matrix labels exist in statistics corresponding to all feature points in the sample image features; inputting the scaled sample image features into a quantity statistical model to obtain whether a statistical object exists a prediction matrix and a first statistical object existence probability matrix; determining an initial loss value of the quantity statistical model according to whether the statistics have matrix labels, whether the statistics have prediction matrixes and a first statistics existence probability matrix; determining a target loss value based on the initial loss value; and performing model training on the digital statistical model according to the target loss value until the target loss value is smaller than a preset threshold. By means of the method, the quantity statistical model for quantity statistics can be obtained, and accuracy of determining the quantity of the statistics is improved.
Description
Technical Field
The invention relates to the technical field of quantity statistics, in particular to a training method and device of a quantity statistical model, electronic equipment and a storage medium.
Background
With the development of society, the number of articles of manufacture, both population and production, has begun to increase. This adds significant difficulty to the statistics of the number. Currently, the aggregate area of the statistics in the image is generally selected by a target detection model, and then the number of the statistics is determined in each frame selected by the frame.
However, when the statistics are too dense, the frames overlap with each other, if the overlapping rate is too high, one of the frames is screened out, so that the statistics quantity of the statistics is smaller than the actual quantity, and the accuracy is lower.
Disclosure of Invention
In view of the foregoing, an object of the present application is to provide a training method, device, electronic apparatus, and storage medium for a quantity statistics model, which can obtain a quantity statistics model for quantity statistics, and improve accuracy in determining the quantity of statistics.
In a first aspect, an embodiment of the present application provides a training method of a quantitative statistical model, where the training method of the quantitative statistical model includes:
acquiring sample image features and whether matrix labels exist in statistics corresponding to all feature points in the sample image features;
inputting the scaled sample image features into a quantity statistical model to obtain whether a statistical object exists a prediction matrix and a first statistical object existence probability matrix;
Determining an initial loss value of the quantity statistical model according to whether the statistics have matrix labels, whether the statistics have prediction matrixes and a first statistics existence probability matrix;
determining a target loss value based on the initial loss value;
and performing model training on the digital statistical model according to the target loss value until the target loss value is smaller than a preset threshold.
In one possible embodiment, determining the initial loss value of the quantitative statistical model based on whether the statistic has a matrix tag, whether the statistic has a prediction matrix, and a first statistic presence probability matrix includes:
calculating whether a statistic exists a matrix label or not, and whether a result of existence of the statistic exists between prediction matrixes is a mean square error of feature points of the existence statistic;
calculating a binary cross entropy error of the first statistic existence probability matrix;
and calculating an initial loss value according to the mean square error and the binary cross entropy error.
In one possible implementation, determining the target loss value based on the initial loss value includes:
obtaining statistical existence probability matrix labels corresponding to all feature points in the sample image features;
inputting the sample image features into a quantity statistical model to obtain a second statistical existence probability matrix;
Determining the existence probability loss value of the quantity statistical model according to the statistical existence probability matrix label and the second statistical existence probability matrix;
and determining a target loss value of the quantity statistical model according to the initial loss value and the existence probability loss value.
In one possible embodiment, calculating whether the statistics have matrix labels and whether the statistics have matrix labels, and the presence result between the statistics and the prediction matrix is a mean square error of feature points of the presence statistics comprises:
scaling the size of whether the statistics have matrix labels to be the same as the size of whether the statistics have prediction matrixes;
calculating a mean square error by the following formula;
；
wherein,,is mean square error>For the number of feature points in which the presence result in the matrix label is the presence statistic for the scaled statistic, < >>For the presence result is the value corresponding to the presence statistic, < >>In order to determine whether the statistics exist in the prediction matrix, the i-th existence result in the scaled statistics existence matrix label is the existence result corresponding value of the feature point corresponding to the position coordinate of the feature point of the existence statistics.
In one possible implementation, calculating the binary cross entropy error of the first statistic existence probability matrix includes:
Calculating a binary cross entropy error by the following formula;
；
wherein,,is a binary cross entropy error, ">For the number of feature points in the first statistical presence probability matrix,/for the first statistical presence probability matrix>For the number of feature points with probability value greater than or equal to preset probability value in the first statistical existence probability matrix,/or->For the ith in the first statistical existence probability matrix, the probability value is larger than or equal to the probability value of the feature point of the preset probability value,/for the first statistical existence probability matrix>For the number of feature points in the first statistic existence probability matrix, the probability value of which is smaller than the preset probability value,/for the feature points>For the ith in the first statistic existence probability matrix, the probability value is smaller than the probability value of the feature point of the preset probability value, +.>For a preset balancing factor.
In a possible implementation manner, the number statistical model includes a regression branch for predicting whether the statistics of the image features exist in the prediction matrix and the first statistics existence probability matrix, and a probability matrix branch for predicting the second statistics existence probability matrix of the image features, and the method further includes:
and removing regression branches in the trained quantity statistical model, and reserving probability matrix branches to obtain a final trained quantity statistical model.
In one possible embodiment, the method further comprises:
Inputting the image features to be counted into the number statistical model after training is completed, and obtaining a statistic existence probability matrix corresponding to the image features to be counted;
the method comprises the steps of setting the existence result of scattered feature points as no statistic in a preset area of a statistic existence probability matrix corresponding to the image features to be counted through a greedy algorithm, wherein the probability value is larger than or equal to a preset probability value;
and rounding the sum of all probability values in the preset area in the statistical existence probability matrix corresponding to the image features to be counted after the setting is completed, so as to obtain the aggregation quantity of the statistical matters.
In a second aspect, an embodiment of the present application further provides a training device for a quantitative statistical model, where the training device for a quantitative statistical model includes:
the acquisition module is used for acquiring the sample image characteristics and whether matrix labels exist in the statistics corresponding to all the characteristic points in the sample image characteristics;
the input module is used for inputting the zoomed sample image characteristics into the quantity statistical model to obtain whether a statistical object exists a prediction matrix and a first statistical object existence probability matrix;
the determining module is used for determining an initial loss value of the quantity statistical model according to whether the statistical object has a matrix label, whether the statistical object has a prediction matrix and a first statistical object existence probability matrix;
The determining module is further used for determining a target loss value based on the initial loss value;
and the training module is used for carrying out model training on the digital statistical model according to the target loss value until the target loss value is smaller than a preset threshold value.
In one possible implementation manner, the determining module is specifically configured to calculate whether the statistics have matrix labels, and whether the existence result between the statistics have prediction matrices is a mean square error of feature points of the existence statistics; calculating a binary cross entropy error of the first statistic existence probability matrix; and calculating an initial loss value according to the mean square error and the binary cross entropy error.
In a possible implementation manner, the determining module is specifically configured to obtain statistics existence probability matrix labels corresponding to all feature points in the sample image features; inputting the sample image features into a quantity statistical model to obtain a second statistical existence probability matrix; determining the existence probability loss value of the quantity statistical model according to the statistical existence probability matrix label and the second statistical existence probability matrix; and determining a target loss value of the quantity statistical model according to the initial loss value and the existence probability loss value.
In a possible implementation manner, the determining module is further configured to scale the size of the matrix label of the statistic to be the same as the size of the prediction matrix of the statistic; calculating a mean square error by the following formula;
；
Wherein,,is mean square error>For the number of feature points in which the presence result in the matrix label is the presence statistic for the scaled statistic, < >>For the presence result is the value corresponding to the presence statistic, < >>In order to determine whether the statistics exist in the prediction matrix, the i-th existence result in the scaled statistics existence matrix label is the existence result corresponding value of the feature point corresponding to the position coordinate of the feature point of the existence statistics.
In a possible implementation manner, the determining module is further configured to calculate a binary cross entropy error by the following formula;
；
wherein,,is a binary cross entropy error, ">For the number of feature points in the first statistical presence probability matrix,/for the first statistical presence probability matrix>For the number of feature points with probability value greater than or equal to preset probability value in the first statistical existence probability matrix,/or->For the ith in the first statistical existence probability matrix, the probability value is larger than or equal to the probability value of the feature point of the preset probability value,/for the first statistical existence probability matrix>For the number of feature points in the first statistic existence probability matrix, the probability value of which is smaller than the preset probability value,/for the feature points>For the ith in the first statistic existence probability matrix, the probability value is smaller than the probability value of the feature point of the preset probability value, +. >For a preset balancing factor.
In a possible implementation manner, the number statistical model includes a regression branch for predicting whether the statistics of the image features exist in the prediction matrix and the first statistics existence probability matrix, and a probability matrix branch for predicting the second statistics existence probability matrix of the image features, and the apparatus further includes: a removal module;
and the removing module is used for removing the regression branches in the trained quantity statistical model, and reserving probability matrix branches to obtain a final trained quantity statistical model.
In one possible embodiment, the apparatus further comprises: setting a module;
the input module is also used for inputting the image features to be counted into the number statistical model after training is completed, and obtaining a statistical object existence probability matrix corresponding to the image features to be counted;
the setting module is used for setting the existence result of scattered feature points as no statistic in a preset area of the statistic existence probability matrix corresponding to the image features to be counted through a greedy algorithm, wherein the probability value is greater than or equal to the preset probability value;
the determining module is further configured to round the sum of all probability values in the preset area in the statistics existence probability matrix corresponding to the image feature to be counted after the setting is completed, so as to obtain the aggregation number of the statistics.
In a third aspect, an embodiment of the present application further provides an electronic device, including: the system comprises a processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the storage medium are communicated through the bus, and the processor executes the machine-readable instructions to execute the steps of the training method of the number statistical model according to any item of the first aspect.
In a fourth aspect, embodiments of the present application further provide a computer readable storage medium having a computer program stored thereon, which when executed by a processor performs the steps of the training method of any of the number statistical models of the first aspect.
The embodiment of the application provides a training method and device for a quantity statistical model, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring sample image features and whether matrix labels exist in statistics corresponding to all feature points in the sample image features; inputting the scaled sample image features into a quantity statistical model to obtain whether a statistical object exists a prediction matrix and a first statistical object existence probability matrix; determining an initial loss value of the quantity statistical model according to whether the statistics have matrix labels, whether the statistics have prediction matrixes and a first statistics existence probability matrix; determining a target loss value based on the initial loss value; and performing model training on the digital statistical model according to the target loss value until the target loss value is smaller than a preset threshold. According to the method, the device and the system, the scaled sample image features are input into a quantity statistical model to obtain whether a statistical object exists a prediction matrix and a first statistical object existence probability matrix, and whether the obtained statistical object exists a matrix label to determine an initial loss value of the quantity statistical model; determining a target loss value based on the initial loss value; and then, carrying out model training on the number statistical model according to the target loss value until the target loss value is smaller than a preset threshold value to obtain the number statistical model for number statistics, thereby improving the accuracy of determining the number of the statistics.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered limiting the scope, and that other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of a training method of a quantity statistical model according to an embodiment of the present application;
FIG. 2 is a flow chart illustrating another method of training a quantitative statistical model provided by embodiments of the present application;
FIG. 3 is a flowchart of another method for training a quantitative statistical model provided by an embodiment of the present application;
FIG. 4 is a schematic structural diagram of a training device for a quantity statistical model according to an embodiment of the present disclosure;
fig. 5 shows a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application, and it should be understood that the accompanying drawings in the present application are only for the purpose of illustration and description, and are not intended to limit the protection scope of the present application. In addition, it should be understood that the schematic drawings are not drawn to scale. A flowchart, as used in this application, illustrates operations implemented according to some embodiments of the present application. It should be understood that the operations of the flow diagrams may be implemented out of order and that steps without logical context may be performed in reverse order or concurrently. Moreover, one or more other operations may be added to the flow diagrams and one or more operations may be removed from the flow diagrams as directed by those skilled in the art.
In addition, the described embodiments are only some, but not all, of the embodiments of the present application. The components of the embodiments of the present application, which are generally described and illustrated in the figures herein, may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, as provided in the accompanying drawings, is not intended to limit the scope of the application, as claimed, but is merely representative of selected embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present application without making any inventive effort, are intended to be within the scope of the present application.
In order to enable one skilled in the art to use the present disclosure, the following embodiments are presented in connection with a specific application scenario "quantitative statistics technical field". It will be apparent to those having ordinary skill in the art that the general principles defined herein may be applied to other embodiments and applications without departing from the spirit and scope of the present application. Although the present application is described primarily in the context of "quantitative statistics technology," it should be understood that this is but one exemplary embodiment.
It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the presence of the features stated hereinafter, but not to exclude the addition of other features.
The following describes a training method of a quantity statistical model in detail.
Referring to fig. 1, a flow chart of a training method of a quantity statistical model according to an embodiment of the present application is shown, where a specific execution process of the training method of the quantity statistical model is as follows:
s101, acquiring sample image features and whether matrix labels exist in statistics corresponding to all feature points in the sample image features.
S102, inputting the scaled sample image features into a quantity statistical model to obtain a prediction matrix and a first statistical existence probability matrix of the statistical matters.
S103, determining an initial loss value of the quantity statistical model according to whether the statistical object has a matrix label, whether the statistical object has a prediction matrix and a first statistical object existence probability matrix.
S104, determining a target loss value based on the initial loss value.
S105, training the digital statistical model according to the target loss value until the target loss value is smaller than a preset threshold.
The embodiment of the application provides a training method and device for a quantity statistical model, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring sample image features and whether matrix labels exist in statistics corresponding to all feature points in the sample image features; inputting the scaled sample image features into a quantity statistical model to obtain whether a statistical object exists a prediction matrix and a first statistical object existence probability matrix; determining an initial loss value of the quantity statistical model according to whether the statistics have matrix labels, whether the statistics have prediction matrixes and a first statistics existence probability matrix; determining a target loss value based on the initial loss value; and performing model training on the digital statistical model according to the target loss value until the target loss value is smaller than a preset threshold. According to the method, the device and the system, the scaled sample image features are input into a quantity statistical model to obtain whether a statistical object exists a prediction matrix and a first statistical object existence probability matrix, and whether the obtained statistical object exists a matrix label to determine an initial loss value of the quantity statistical model; determining a target loss value based on the initial loss value; and then, carrying out model training on the number statistical model according to the target loss value until the target loss value is smaller than a preset threshold value to obtain the number statistical model for number statistics, thereby improving the accuracy of determining the number of the statistics.
Exemplary steps of embodiments of the present application are described below:
s101, acquiring sample image features and whether matrix labels exist in statistics corresponding to all feature points in the sample image features.
In the embodiment of the application, acquiring a statistic video with multiple resolutions of a statistic under multiple scenes; performing frame extraction on the statistic video to obtain a statistic sample image containing the statistic; the quantitative statistical model includes regression branches of a prediction matrix and a first statistical existence probability matrix for predicting the image feature, and probability matrix branches of a second statistical existence probability matrix for predicting the image feature. The statistical sample image was taken at 3:1: the scale of 1 is divided into a training set, a validation set and a test set.
And inputting the statistical sample images in the training set into the quantitative statistical model to obtain sample image features, wherein the sizes corresponding to the sample image features are the same as the sizes of the input sample images. Manually labeling each sample image feature with a matrix label and a statistic quantity label, wherein the matrix label contains a result of whether the statistic corresponding to each feature point in the sample image feature exists or not; the result of whether the statistic exists or not includes two results of the existence of the statistic and the nonexistence of the statistic, the result of the existence of the statistic is represented by 0, the result of the nonexistence of the statistic is represented by 1, and thus the label of the matrix of whether the statistic exists or not is a matrix composed of 0 and 1.
The number statistical model is a CNN network, the characteristic of the sample image is extracted through a convolutional neural network in the CNN network, and then the feature map is up-sampled by adopting a deconvolution machine network, so that the size corresponding to the characteristic of the output sample image is the same as the size of the input sample image. Thus, the problem of inaccurate results caused by excessive difference between the input image size and the output size can be avoided.
S102, inputting the scaled sample image features into a quantity statistical model to obtain a prediction matrix and a first statistical existence probability matrix of the statistical matters.
In the present embodiment, the size of the sample image feature is scaled to be the same as the output size of the regression branch matrix. The existence prediction matrix of the statistic and the existence probability matrix of the first statistic are respectively output corresponding to a first regression branch and a second regression branch in the regression branches.
Here, the scaling of the sample image features is performed to make the image size of the input image features identical to the output size, so that the problem of inaccurate results due to an excessively large difference between the input image size and the output size is avoided. The first and second regression branches in the example were 4*2 and 4*1 channels, respectively.
S103, determining an initial loss value of the quantity statistical model according to whether the statistical object has a matrix label, whether the statistical object has a prediction matrix and a first statistical object existence probability matrix.
In the embodiment of the application, the initial loss value of the quantitative statistical model is determined according to the number and position information of feature points of which the presence result is the presence statistic in the presence/absence statistic matrix label of the presence/absence statistic of the prediction matrix, the first statistic presence/absence statistic probability matrix and the presence/absence statistic of the first statistic. Therefore, the initial loss value in this embodiment is determined by whether the statistics exist in the matrix label or not and the feature points of which the existence result is the existence statistics are used for representing the accuracy of the regression branches in the quantitative statistical model; the smaller the initial loss value, the higher the accuracy of the regression branch; the larger the initial loss value, the lower the accuracy of the regression branch.
Determining an initial loss value of the quantitative statistical model by:
I. and calculating whether the statistics exist matrix labels or not, and whether the statistics exist prediction matrixes or not, wherein the existence result is the mean square error of feature points of the existence statistics.
Specifically, the size of the matrix label of the statistic is scaled to be the same as the size of the prediction matrix of the statistic. The mean square error is calculated by the following formula:
；
Wherein,,is mean square error>For the number of feature points in which the presence result in the matrix label is the presence statistic for the scaled statistic, < >>For the presence result is the value corresponding to the presence statistic, < >>In order to determine whether the statistics exist in the prediction matrix, the i-th existence result in the scaled statistics existence matrix label is the existence result corresponding value of the feature point corresponding to the position coordinate of the feature point of the existence statistics.
II. A binary cross entropy error of the first statistic presence probability matrix is calculated.
Specifically, the binary cross entropy error is calculated by the following formula:
；
wherein,,is a binary cross entropy error, ">For the number of feature points in the first statistical presence probability matrix,/for the first statistical presence probability matrix>For the first statistical existence probabilityThe number of feature points in the matrix with probability values greater than or equal to the preset probability value, < >>For the ith in the first statistical existence probability matrix, the probability value is larger than or equal to the probability value of the feature point of the preset probability value,/for the first statistical existence probability matrix>For the number of feature points in the first statistic existence probability matrix, the probability value of which is smaller than the preset probability value,/for the feature points>For the ith in the first statistic existence probability matrix, the probability value is smaller than the probability value of the feature point of the preset probability value, +. >For a preset balancing factor.
And III, calculating an initial loss value according to the mean square error and the binary cross entropy error.
Optionally, weighting is performed according to preset balance factors corresponding to the mean square error and the binary cross entropy error respectively, so as to obtain an initial loss value.
Alternatively, the mean value of the mean square error and the binary cross entropy error is used as the initial loss value.
Optionally, weighting is performed according to preset balance factors corresponding to the mean square error and the binary cross entropy error respectively, so as to obtain a first loss value. Taking the mean value of the mean square error and the binary cross entropy error as a second loss value; and continuing to weight average the first loss value and the second loss value to obtain an initial loss value.
S104, determining a target loss value based on the initial loss value.
In the embodiment of the application, the target loss value represents the comprehensive precision of the quantity statistical model; the smaller the target loss value is, the higher the accuracy of the quantity statistical model is; the greater the target loss value, the lower the accuracy of the quantitative statistical model.
Specifically, determining the target loss value includes the following two schemes:
firstly, obtaining statistical existence probability matrix labels corresponding to all feature points in sample image features; inputting the sample image features into a quantity statistical model to obtain a second statistical existence probability matrix; determining the existence probability loss value of the quantity statistical model according to the statistical existence probability matrix label and the second statistical existence probability matrix; and determining a target loss value of the quantity statistical model according to the initial loss value and the existence probability loss value.
Scheme II, taking the initial loss value as a target loss value.
S105, training the digital statistical model according to the target loss value until the target loss value is smaller than a preset threshold.
Optionally, removing regression branches in the trained quantity statistical model, and reserving probability matrix branches to obtain a final trained quantity statistical model.
In the embodiment of the application, when the number of the statistics is determined through the number statistics model, only probability matrix branches in the number statistics model are needed, and regression branches are only used for training the number statistics model, so that the accuracy of the number statistics model is improved. Therefore, after the removal, the processing efficiency of the model can be improved.
Further, the image features to be counted are input into the number statistical model after training is completed, and a statistical object existence probability matrix corresponding to the image features to be counted is obtained. And rounding the sum of all probability values in a preset area in the statistical existence probability matrix corresponding to the image features to be counted to obtain the number of the statistical objects. In this way, the total number of statistics contained in the image of the statistics to be counted can be determined using the number statistical model.
The preset area refers to an area needing to count the number of the statistics in the image features to be counted, and the preset area can be the whole image to be counted or some partial areas of the image to be counted.
Further, verifying the trained quantity statistical model according to the verification set; and testing the number statistical model according to the test set, wherein when the average precision of the number statistical model is more than or equal to 97% in the test set, the number statistical model is the final target number statistical model after training, otherwise, the number statistical model is continuously trained until the average precision of the number statistical model is more than or equal to 97% in the test set.
Further, as shown in fig. 2, another training method for a quantitative statistical model provided in the embodiment of the present application further includes:
s201, inputting the image features to be counted into the number statistical model after training is completed, and obtaining a statistic existence probability matrix corresponding to the image features to be counted.
In the embodiment of the present application, the number statistical model is a number statistical model obtained by training in the training manner shown in fig. 1.
S202, setting the existence result of scattered feature points as no statistic in a preset area of a statistic existence probability matrix corresponding to the image features to be counted through a greedy algorithm, wherein the probability value is larger than or equal to a preset probability value.
In this embodiment of the present application, the preset probability value is a minimum probability value of the pixel when the existence result of the pixel is the existence statistic. And (3) searching by using a mask algorithm in a preset area of a statistic existence probability matrix corresponding to the image features to be counted, so as to set the existence result of the feature points with the statistic scattered aside in the preset area as no statistic, namely, set to be 0.
The preset area refers to an area needing to count the number of the statistics in the image features to be counted, and the preset area can be the whole image to be counted or some partial areas of the image to be counted.
S203, rounding the sum of all probability values in a preset area in the statistical existence probability matrix corresponding to the image features to be counted after the setting is completed, so as to obtain the aggregation number of the statistical objects.
In the embodiment of the present application, the aggregation number of the statistics is the number of the statistics in the preset area, in which the statistics occur.
By means of the method, the number of the statistics objects aggregated in the image to be counted can be counted.
Referring to fig. 3, a flowchart of another training method for a quantity statistical model according to an embodiment of the present application is shown, and exemplary steps of the embodiment of the present application are described below:
s301, obtaining statistical existence probability matrix labels corresponding to all feature points in the sample image features.
In the embodiment of the present application, based on step S101, a gaussian kernel function is used to perform gaussian smoothing on whether a matrix label exists on a statistic, so as to obtain a statistic existence probability matrix label, so as to reduce an error of identifying the statistic by a quantity statistical model, and further improve accuracy of determining the quantity of the statistic by the quantity statistical model.
S302, inputting the sample image features into the quantity statistical model to obtain a second statistical existence probability matrix.
In this embodiment, the second statistical presence probability matrix is an output result of the probability matrix branch.
S303, determining the existence probability loss value of the quantity statistical model according to the statistical existence probability matrix label and the second statistical existence probability matrix.
In the embodiment of the application, the existence probability loss value is used for representing the accuracy of the probability matrix branches in the quantity statistical model, and the smaller the existence probability loss value is, the higher the accuracy of the probability matrix branches is; the larger the existence probability loss value, the lower the accuracy of the probability matrix branches.
Specifically, the existence probability loss value of the quantitative statistical model is calculated by the following formula;
；
wherein,,for the probability loss value +.>For the number of feature points in the statistics presence probability matrix label,/>For the number of rows in the statistics presence probability matrix label,/->For the number of columns in the statistics presence probability matrix label,/->For the value of the feature point of the ith row and jth column in the statistic presence probability matrix label,/>The value of the feature point in the ith row and jth column in the second statistic-presence probability matrix.
S304, determining a target loss value of the quantity statistical model according to the initial loss value and the existence probability loss value.
In the embodiment of the present application, the target loss value is obtained by performing weighting processing according to the balance factors corresponding to the initial loss value and the existence probability loss value, respectively.
By the method provided by the embodiment of the application, the accuracy of the target loss value can be improved, and the accuracy of determining the number of the statistics through the number statistics model is further improved.
Based on the same inventive concept, the embodiment of the present application further provides a training device for a quantity statistical model corresponding to the training method for a quantity statistical model, and since the principle of solving the problem by the device in the embodiment of the present application is similar to that of the training method for the quantity statistical model in the embodiment of the present application, the implementation of the device can refer to the implementation of the method, and the repetition is omitted.
Referring to fig. 4, a schematic structural diagram of a training device for a quantitative statistical model according to an embodiment of the present application is shown, where the training device for a quantitative statistical model includes:
an obtaining module 401, configured to obtain a sample image feature and whether a matrix label exists in statistics corresponding to all feature points in the sample image feature;
the input module 402 is configured to input the scaled sample image features into the quantitative statistical model to obtain a prediction matrix of whether a statistic exists or not and a first statistic existence probability matrix;
a determining module 403, configured to determine an initial loss value of the number statistical model according to whether the statistic has a matrix label, whether the statistic has a prediction matrix, and a first statistic presence probability matrix;
a determining module 403, configured to determine a target loss value based on the initial loss value;
and the training module 404 is configured to perform model training on the digital statistical model according to the target loss value until the target loss value is less than a preset threshold.
In a possible implementation manner, the determining module 403 is specifically configured to calculate whether a matrix label exists in the statistic and whether a matrix exists in the statistic, where a result of existence of the prediction matrix is a mean square error of a feature point of the statistic; calculating a binary cross entropy error of the first statistic existence probability matrix; and calculating a target loss value according to the mean square error and the binary cross entropy error.
In a possible implementation manner, the determining module 403 is specifically configured to obtain statistics existence probability matrix labels corresponding to all feature points in the sample image features; inputting the sample image features into a quantity statistical model to obtain a second statistical existence probability matrix; determining the existence probability loss value of the quantity statistical model according to the statistical existence probability matrix label and the second statistical existence probability matrix; and determining a target loss value of the quantity statistical model according to the initial loss value and the existence probability loss value.
In a possible implementation, the determining module 403 is further configured to scale the size of the matrix label of the statistic to be the same as the size of the prediction matrix of the statistic; calculating a mean square error by the following formula;
；
wherein,,is mean square error>For the number of feature points in which the presence result in the matrix label is the presence statistic for the scaled statistic, < >>For the presence result is the value corresponding to the presence statistic, < >>In order to determine whether the statistics exist in the prediction matrix, the i-th existence result in the scaled statistics existence matrix label is the existence result corresponding value of the feature point corresponding to the position coordinate of the feature point of the existence statistics.
In a possible implementation, the determining module 403 is further configured to calculate a binary cross entropy error by the following formula;
；
wherein,,is a binary cross entropy error, ">For the number of feature points in the first statistical presence probability matrix,/for the first statistical presence probability matrix>For the number of feature points with probability value greater than or equal to preset probability value in the first statistical existence probability matrix,/or->For the ith in the first statistical existence probability matrix, the probability value is larger than or equal to the probability value of the feature point of the preset probability value,/for the first statistical existence probability matrix>For the number of feature points in the first statistic existence probability matrix, the probability value of which is smaller than the preset probability value,/for the feature points>For the ith in the first statistic existence probability matrix, the probability value is smaller than the probability value of the feature point of the preset probability value, +.>For a preset balancing factor.
In a possible implementation manner, the number statistical model includes a regression branch for predicting whether the statistics of the image features exist in the prediction matrix and the first statistics existence probability matrix, and a probability matrix branch for predicting the second statistics existence probability matrix of the image features, and the apparatus further includes: a removal module 405;
and the removing module 405 is configured to remove the regression branches in the trained number statistical model, and reserve the probability matrix branches to obtain a final trained number statistical model.
In one possible embodiment, the apparatus further comprises: a setting module 406;
the input module 402 is further configured to input the image feature to be counted into the trained number statistical model, so as to obtain a statistic existence probability matrix corresponding to the image feature to be counted;
the setting module 406 is configured to set, by using a greedy algorithm, a probability value greater than or equal to a preset probability value in a preset area of a statistic existence probability matrix corresponding to the image feature to be counted, where the existence result of the scattered feature points is set to be that no statistic exists;
the determining module 403 is further configured to round the sum of all probability values in the preset area in the statistics existence probability matrix corresponding to the image feature to be counted after the setting is completed, so as to obtain the aggregation number of the statistics.
As shown in fig. 5, an electronic device 500 provided in an embodiment of the present application includes: processor 501, memory 502 and bus, memory 502 storing machine readable instructions executable by processor 501, processor 501 executing machine readable instructions to perform the steps of the training method of the quantitative statistical model described above when the electronic device is running, processor 501 communicates with memory 502 via the bus.
Specifically, the memory 502 and the processor 501 can be general-purpose memories and processors, which are not limited herein, and the training method of the number statistical model can be performed when the processor 501 runs a computer program stored in the memory 502.
Corresponding to the training method of the quantity statistical model, the embodiment of the application also provides a computer readable storage medium, wherein a computer program is stored on the computer readable storage medium, and the computer program executes the steps of the training method of the quantity statistical model when being executed by a processor.
It will be clearly understood by those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described system and apparatus may refer to corresponding procedures in the method embodiments, which are not described in detail in this application. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods may be implemented in other manners. The above-described apparatus embodiments are merely illustrative, and the division of the modules is merely a logical function division, and there may be additional divisions when actually implemented, and for example, multiple modules or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some communication interface, indirect coupling or communication connection of devices or modules, electrical, mechanical, or other form.
The modules described as separate components may or may not be physically separate, and components shown as modules may or may not be physical units, may be located in one place, or may be distributed over multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in each embodiment of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a non-volatile computer readable storage medium executable by a processor. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the information processing method described in the embodiments of the present application. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a ROM, a RAM, a magnetic disk, or an optical disk, etc.
The foregoing is merely a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think about changes or substitutions within the technical scope of the present application, and the changes or substitutions are covered in the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (7)
1. A method for training a quantitative statistical model, the method comprising:
acquiring sample image features and whether matrix labels exist in statistics corresponding to all feature points in the sample image features;
inputting the scaled sample image features into a quantity statistical model to obtain whether a statistical object exists a prediction matrix and a first statistical object existence probability matrix;
determining an initial loss value of the number statistical model according to whether the statistic has a matrix label, whether the statistic has a prediction matrix and the first statistic existence probability matrix, including: calculating whether a matrix label exists in the statistic and whether a mean square error exists between the existence result of the statistic and the existence prediction matrix is the feature point of the statistic; calculating a binary cross entropy error of the first statistic existence probability matrix; calculating the initial loss value according to the mean square error and the binary cross entropy error;
Determining a target loss value based on the initial loss value;
model training is carried out on the quantity statistical model according to the target loss value until the target loss value is smaller than a preset threshold value;
the calculating whether the statistics have matrix labels and whether the statistics have prediction matrixes has the mean square error of feature points of the statistics, wherein the mean square error comprises the following steps: scaling the size of the matrix label of the statistic to be the same as the size of the prediction matrix of the statistic; calculating the mean square error by the following formula;the method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Is mean square error>For the number of feature points in which the presence result in the matrix label is the presence statistic for the scaled statistic, < >>For the presence result is the value corresponding to the presence statistic, < >>The method comprises the steps of judging whether a statistic exists in a prediction matrix, wherein the ith existence result in a scaled statistic existence matrix label is the existence result corresponding value of a feature point of a position coordinate corresponding to a feature point of the existence statistic;
the calculating the binary cross entropy error of the first statistic existence probability matrix comprises: calculating the binary cross entropy error by the following formula; The method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Is a binary cross entropy error, ">For the number of feature points in the first statistical presence probability matrix,/for the first statistical presence probability matrix>For the number of feature points with probability value greater than or equal to preset probability value in the first statistical existence probability matrix,/or->For the probability value of the feature point with the ith probability value greater than or equal to the preset probability value in the first statistical existence probability matrix, the weight is +.>For the number of feature points in the first statistic existence probability matrix, the probability value of which is smaller than the preset probability value,/for the feature points>For the probability value of the feature point whose jth probability value is smaller than the preset probability value in the first statistic existence probability matrix,/the probability value of the feature point is smaller than the preset probability value>For a preset balancing factor.
2. The method of training a quantitative statistical model of claim 1, wherein the determining a target loss value based on the initial loss value comprises:
obtaining statistical existence probability matrix labels corresponding to all feature points in the sample image features;
inputting the sample image features into the quantity statistical model to obtain a second statistical existence probability matrix;
determining the existence probability loss value of the quantity statistical model according to the statistical existence probability matrix label and the second statistical existence probability matrix;
And determining a target loss value of the quantity statistical model according to the initial loss value and the existence probability loss value.
3. The method of training a quantitative statistical model according to claim 2, wherein the quantitative statistical model includes regression branches for predicting whether a prediction matrix exists for a statistic of image features and a first statistic existence probability matrix, and probability matrix branches for predicting a second statistic existence probability matrix of image features, the method further comprising:
and removing regression branches in the trained quantity statistical model, and reserving the probability matrix branches to obtain a final trained quantity statistical model.
4. A method of training a quantitative statistical model according to any one of claims 1 to 3, further comprising:
inputting the image features to be counted into the number statistical model after training is completed, and obtaining a statistic existence probability matrix corresponding to the image features to be counted;
the probability value is larger than or equal to a preset probability value in a preset area of a statistic existence probability matrix corresponding to the image features to be counted through a greedy algorithm, and the existence result of scattered feature points is set to be that no statistic exists;
And rounding the sum of all probability values in the preset area in a statistical existence probability matrix corresponding to the image features to be counted after the setting is completed, so as to obtain the aggregation number of the statistical objects.
5. A training device for a quantitative statistical model, wherein the training device for a quantitative statistical model comprises:
the acquisition module is used for acquiring sample image features and whether matrix labels exist in statistics corresponding to all feature points in the sample image features;
the input module is used for inputting the zoomed sample image characteristics into the quantity statistical model to obtain whether a statistical object exists a prediction matrix and a first statistical object existence probability matrix;
a determining module, configured to determine an initial loss value of the number statistical model according to whether the statistic has a matrix label, whether the statistic has a prediction matrix, and the first statistic presence probability matrix, where the determining module includes: calculating whether a matrix label exists in the statistic and whether a mean square error exists between the existence result of the statistic and the existence prediction matrix is the feature point of the statistic; calculating a binary cross entropy error of the first statistic existence probability matrix; calculating the initial loss value according to the mean square error and the binary cross entropy error;
The determining module is further used for determining a target loss value based on the initial loss value;
the training module is used for carrying out model training on the quantity statistical model according to the target loss value until the target loss value is smaller than a preset threshold value;
the determining module is further configured to: scaling the size of the matrix label of the statistic to be the same as the size of the prediction matrix of the statistic; calculating the mean square error by the following formula;the method comprises the steps of carrying out a first treatment on the surface of the Wherein the method comprises the steps of，/>Is mean square error>For the number of feature points in which the presence result in the matrix label is the presence statistic for the scaled statistic, < >>For the presence result is the value corresponding to the presence statistic, < >>The method comprises the steps of judging whether a statistic exists in a prediction matrix, wherein the ith existence result in a scaled statistic existence matrix label is the existence result corresponding value of a feature point of a position coordinate corresponding to a feature point of the existence statistic;
the determining module is further configured to: calculating the binary cross entropy error by the following formula;the method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Is a binary cross entropy error, ">For the number of feature points in the first statistical presence probability matrix,/for the first statistical presence probability matrix>For the number of feature points with probability value greater than or equal to preset probability value in the first statistical existence probability matrix,/or- >For the probability value of the feature point with the ith probability value greater than or equal to the preset probability value in the first statistical existence probability matrix, the weight is +.>For the number of feature points in the first statistic existence probability matrix, the probability value of which is smaller than the preset probability value,/for the feature points>For the probability value of the feature point whose jth probability value is smaller than the preset probability value in the first statistic existence probability matrix,/the probability value of the feature point is smaller than the preset probability value>For a preset balancing factor.
6. An electronic device, comprising: a processor, a storage medium and a bus, the storage medium storing machine-readable instructions executable by the processor, the processor and the storage medium communicating over the bus when the electronic device is running, the processor executing the machine-readable instructions to perform the steps of the method of training a quantitative statistical model according to any one of claims 1 to 4.
7. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a computer program which, when executed by a processor, performs the steps of the training method of a quantitative statistical model according to any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310424471.4A CN116137061B (en) | 2023-04-20 | 2023-04-20 | Training method and device for quantity statistical model, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310424471.4A CN116137061B (en) | 2023-04-20 | 2023-04-20 | Training method and device for quantity statistical model, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116137061A CN116137061A (en) | 2023-05-19 |
| CN116137061B true CN116137061B (en) | 2023-08-01 |
Family
ID=86333670
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310424471.4A Active CN116137061B (en) | 2023-04-20 | 2023-04-20 | Training method and device for quantity statistical model, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116137061B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117166996B (en) * | 2023-07-27 | 2024-03-22 | 中国地质大学（北京） | Method, device, equipment and storage medium for determining geological parameter threshold |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114596548A (en) * | 2022-02-22 | 2022-06-07 | 上海欧菲智能车联科技有限公司 | Target detection method, target detection device, computer equipment and computer-readable storage medium |
| CN114612728A (en) * | 2022-03-25 | 2022-06-10 | Oppo广东移动通信有限公司 | Model training method and device, computer equipment and storage medium |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112801164B (en) * | 2021-01-22 | 2024-02-13 | 北京百度网讯科技有限公司 | Training method, device, equipment and storage medium of target detection model |
| CN112990432B (en) * | 2021-03-04 | 2023-10-27 | 北京金山云网络技术有限公司 | Target recognition model training method and device and electronic equipment |
| CN113421192B (en) * | 2021-08-24 | 2021-11-19 | 北京金山云网络技术有限公司 | Training method of object statistical model, and statistical method and device of target object |
| CN114817480A (en) * | 2022-05-19 | 2022-07-29 | 中国建设银行股份有限公司 | Method and device for determining land resource value, electronic equipment and storage medium |
| CN114663731B (en) * | 2022-05-25 | 2022-09-20 | 杭州雄迈集成电路技术股份有限公司 | Training method and system of license plate detection model, and license plate detection method and system |
-
2023
- 2023-04-20 CN CN202310424471.4A patent/CN116137061B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114596548A (en) * | 2022-02-22 | 2022-06-07 | 上海欧菲智能车联科技有限公司 | Target detection method, target detection device, computer equipment and computer-readable storage medium |
| CN114612728A (en) * | 2022-03-25 | 2022-06-10 | Oppo广东移动通信有限公司 | Model training method and device, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116137061A (en) | 2023-05-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109886928B (en) | Target cell marking method, device, storage medium and terminal equipment | |
| CN108154105B (en) | Underwater biological detection and identification method and device, server and terminal equipment | |
| CN109086811B (en) | Multi-label image classification method and device and electronic equipment | |
| CN113159147A (en) | Image identification method and device based on neural network and electronic equipment | |
| CN112435214B (en) | Priori frame linear scaling-based pollen detection method and device and electronic equipment | |
| CN108805180B (en) | Target object detection method and device | |
| CN109461133B (en) | Bridge bolt falling detection method and terminal equipment | |
| CN112613553B (en) | Picture sample set generation method and device, computer equipment and storage medium | |
| CN109829371B (en) | Face detection method and device | |
| CN109685805B (en) | Image segmentation method and device | |
| CN111199186A (en) | Image quality scoring model training method, device, equipment and storage medium | |
| CN112950554B (en) | Lung lobe segmentation optimization method and system based on lung segmentation | |
| CN116137061B (en) | Training method and device for quantity statistical model, electronic equipment and storage medium | |
| CN113569968B (en) | Model training method, target detection method, device, equipment and storage medium | |
| CN113408561B (en) | Model generation method, target detection method, device, equipment and storage medium | |
| CN112132265A (en) | Model training method, cup-to-tray ratio determining method, device, equipment and storage medium | |
| CN112651953A (en) | Image similarity calculation method and device, computer equipment and storage medium | |
| CN113034514A (en) | Sky region segmentation method and device, computer equipment and storage medium | |
| CN111444807A (en) | Target detection method, device, electronic equipment and computer readable medium | |
| CN111582032A (en) | Pedestrian detection method and device, terminal equipment and storage medium | |
| CN114494775A (en) | Video segmentation method, device, equipment and storage medium | |
| CN113780287A (en) | Optimal selection method and system for multi-depth learning model | |
| CN115908363A (en) | Tumor cell counting method, device, equipment and storage medium | |
| CN112966687B (en) | Image segmentation model training method and device and communication equipment | |
| CN112215276A (en) | Training method and device for confrontation network, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |