CN110929785A - Data classification method and device, terminal equipment and readable storage medium - Google Patents
Data classification method and device, terminal equipment and readable storage medium Download PDFInfo
- Publication number
- CN110929785A CN110929785A CN201911150175.XA CN201911150175A CN110929785A CN 110929785 A CN110929785 A CN 110929785A CN 201911150175 A CN201911150175 A CN 201911150175A CN 110929785 A CN110929785 A CN 110929785A
- Authority
- CN
- China
- Prior art keywords
- label
- value
- sample data
- preset
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000013145 classification model Methods 0.000 claims abstract description 58
- 230000006870 function Effects 0.000 claims abstract description 56
- 238000012549 training Methods 0.000 claims abstract description 33
- 238000004590 computer program Methods 0.000 claims description 19
- 238000000926 separation method Methods 0.000 claims description 14
- 238000010801 machine learning Methods 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 13
- 238000004891 communication Methods 0.000 description 7
- 239000013598 vector Substances 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 238000013527 convolutional neural network Methods 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 241000282326 Felis catus Species 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 229910003460 diamond Inorganic materials 0.000 description 2
- 239000010432 diamond Substances 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application is applicable to the technical field of machine learning, and provides a data classification method, a device, a terminal device and a readable storage medium, wherein the data classification method comprises the following steps: firstly, data to be classified is received, then the data to be classified is input into the trained classification model, and at least one data label of the data to be classified is obtained. The classification model is obtained by training according to a distance focus loss function, and the distance focus loss function is used for representing the difference between the predicted label of the sample data and the preset label of the sample data according to the spacing distance between the predicted label of the sample data and the preset label of the sample data. When the classification training is performed on the sample data, the number of samples does not need to be considered, and only the classification boundary is determined according to the difference between the prediction label of the sample data and the preset label of the sample data, so that the condition of data imbalance can be effectively reduced, and the trained classification model can more accurately perform label classification on the data to be classified.
Description
Technical Field
The application belongs to the technical field of machine learning, and particularly relates to a data classification method and device, a terminal device and a readable storage medium.
Background
The multi-label classification is one of basic research tasks of machine learning, and aims to predict a plurality of class labels appearing in each sample data, and when a classification model is trained, because the type and the number of the class labels in each sample data are not fixed, the data amount of an individual class label in the predicted label of the sample data is greatly different from the data amount of other class labels, so that data imbalance occurs, and the accuracy of the machine learning model is further reduced.
In the prior art, when the condition of data imbalance is improved, a convolutional neural network can be used for extracting features of sample data, then the features are linearly combined through a full connection layer, then a sigmoid function is used for generating classification probability, then a focus loss function is used as a loss function for back propagation, a classification model is trained, and then classification is carried out through the trained classification model.
However, when a classification model is trained according to the prior art, because the number of each class in sample data is different, for classes with a small number, sufficient sample data features cannot be extracted, and only the weight in the focus loss function is used for adjustment, the situation that a classification boundary cannot be accurately obtained can be caused, so that data imbalance of a certain degree can still be caused, and the accuracy of label classification is affected.
Disclosure of Invention
The embodiment of the application provides a data classification method, a data classification device, a terminal device and a readable storage medium, so as to solve the problems that in the prior art, the classification boundary cannot be accurately obtained due to the fact that adjustment is performed only by means of the weight in the focus loss function, data imbalance to a certain degree is caused, and the accuracy of label classification is affected.
In a first aspect, an embodiment of the present application provides a data classification method, including:
firstly, data to be classified is received, then the data to be classified is input into the trained classification model, and at least one data label of the data to be classified is obtained. The classification model is obtained by training according to a distance focus loss function, and the distance focus loss function is used for representing the difference between the predicted label of the sample data and the preset label of the sample data according to the spacing distance between the predicted label of the sample data and the preset label of the sample data.
In some implementations, the training mode of the trained classification model is as follows: at least one sample data is obtained from a preset database, wherein each sample data comprises at least one preset label. And then, acquiring a prediction label of each sample data through a preset classification model. And then obtaining the spacing distance between the prediction label of the sample data and the preset label of the sample data. And calculating a maximum distance focus loss value through a distance focus loss function according to the distance between the sample data and the preset label of the sample data, wherein the maximum distance focus loss value is used for indicating the maximum value of the difference between the predicted label of the sample data and the preset label of the sample data. And finally, training a preset classification model according to the maximum distance focus loss value to obtain a trained classification model.
It should be noted that the prediction tag includes N classes, where N is an integer greater than 1.
Correspondingly, obtaining the spacing distance between the prediction label of the sample data and the preset label of the sample data comprises: and obtaining the interval distance between the ith type prediction label and the ith type preset label of the sample data according to the value of the ith type prediction label of the sample data and the value of the ith type preset label of the sample data, wherein i is an integer which is more than or equal to 1 and less than or equal to N.
In some implementation manners, the separation distance between the ith class prediction tag and the ith class preset tag of the sample data is obtained according to the value of the ith class prediction tag of the sample data and the value of the ith class preset tag of the sample data, and the absolute distance between the ith class prediction tag and the ith class preset tag of the sample data can be obtained by subtracting the value of the ith class preset tag from the value of the ith class preset tag. And multiplying the absolute distance by a preset scaling coefficient to obtain the interval distance between the ith type of prediction tag and the ith type of preset tag of the sample data.
Optionally, the maximum distance focus loss value is calculated and obtained through a distance focus loss function according to the distance between the first class and the second class, and a value range of the ith class prediction label value may be adjusted according to the distance between the first class and the second class, a value of the ith class prediction label, and a value range of the ith class prediction label value may be adjusted to obtain an ith class prediction label value after the range is adjusted. And obtaining a maximum distance focus loss value according to the ith type prediction label value and the distance focus loss function after the range adjustment.
In some implementation manners, adjusting a value range of the i-th class predicted tag value according to the separation distance, the value of the i-th class predicted tag, and the value of the i-th class preset tag to obtain the i-th class predicted tag value after the range adjustment includes: and multiplying the ith type preset label value by two and then subtracting one to obtain the mapped ith type preset label value. And subtracting the product of the spacing distance and the mapped ith type preset label value from the ith type predicted label value to obtain the mapped ith type predicted label value. And finally, multiplying the mapped ith type of prediction tag value by a preset range scaling coefficient to obtain the ith type of prediction tag value after range adjustment.
Optionally, obtaining a maximum distance focus loss value according to the range-adjusted i-th class predicted label value and the distance focus loss function, including: and firstly, carrying out secondary classification on the ith type of prediction tag value after range adjustment to obtain the ith type of prediction tag value after secondary classification. And then obtaining a maximum distance focus loss value according to the ith class prediction label value after the second classification and the distance focus loss function.
In some implementations, the data classification method is executed by a terminal having image processing capability. Illustratively, the terminal may be a physical terminal, such as a desktop computer, a server, a notebook computer, a tablet computer, or the like, or may be a virtual terminal, such as a cloud-end server, cloud computing, or the like. It should be understood that the above implementation bodies are only examples, and are not necessarily the above terminals.
In a second aspect, an embodiment of the present application provides a data classification apparatus, including: and the receiving module is used for receiving the data to be classified. And the classification module is used for inputting the data to be classified into the trained classification model to obtain at least one data label of the data to be classified, wherein the classification model is obtained by training according to a distance focus loss function, and the distance focus loss function is used for representing the difference between the predicted label of the sample data and the preset label of the sample data according to the interval distance between the predicted label of the sample data and the preset label of the sample data.
In some implementations, the apparatus further includes a training module configured to obtain a trained classification model according to the following steps: at least one sample data is obtained from a preset database, wherein each sample data comprises at least one preset label. And then, acquiring a prediction label of each sample data through a preset classification model. And then obtaining the spacing distance between the prediction label of the sample data and the preset label of the sample data. And calculating a maximum distance focus loss value through a distance focus loss function according to the distance between the sample data and the preset label of the sample data, wherein the maximum distance focus loss value is used for indicating the maximum value of the difference between the predicted label of the sample data and the preset label of the sample data. And finally, training a preset classification model according to the maximum distance focus loss value to obtain a trained classification model.
It should be noted that the prediction tag includes N classes, where N is an integer greater than 1.
Correspondingly, the training module is specifically configured to obtain a separation distance between the ith class prediction tag and the ith class preset tag of the sample data according to the value of the ith class prediction tag of the sample data and the value of the ith class preset tag of the sample data, where i is an integer greater than or equal to 1 and less than or equal to N.
In some implementation manners, the training module is specifically configured to subtract the value of the i-th class prediction label from the value of the i-th class preset label to obtain an absolute distance between the i-th class prediction label and the i-th class preset label of the sample data. And multiplying the absolute distance by a preset scaling coefficient to obtain the interval distance between the ith type of prediction tag and the ith type of preset tag of the sample data.
Optionally, the training module is specifically configured to adjust a value range of the ith class prediction label value according to the separation distance, the value of the ith class prediction label, and the value of the ith class preset label, so as to obtain the range-adjusted ith class prediction label value. And obtaining a maximum distance focus loss value according to the ith type prediction label value and the distance focus loss function after the range adjustment.
In some implementation manners, the training module is specifically configured to multiply the i-th class preset label value by two and then subtract by one to obtain the mapped i-th class preset label value. And subtracting the product of the spacing distance and the mapped ith type preset label value from the ith type predicted label value to obtain the mapped ith type predicted label value. And finally, multiplying the mapped ith type of prediction tag value by a preset range scaling coefficient to obtain the ith type of prediction tag value after range adjustment.
Optionally, the training module is specifically configured to perform second classification on the range-adjusted i-th class predicted tag value to obtain the i-th class predicted tag value after the second classification. And then obtaining a maximum distance focus loss value according to the ith class prediction label value after the second classification and the distance focus loss function.
In a third aspect, an embodiment of the present application provides a terminal device, including: a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the method as provided by the first aspect when executing the computer program.
In a fourth aspect, embodiments of the present application provide a computer-readable storage medium, which stores a computer program, and when executed by a processor, the computer program implements the method as provided in the first aspect.
In a fifth aspect, the present application provides a computer program product, which when run on a terminal device, causes the terminal device to execute the method provided in the first aspect.
It is understood that the beneficial effects of the second aspect to the fifth aspect can be referred to the related description of the first aspect, and are not described herein again.
Compared with the prior art, the embodiment of the application has the advantages that: and classifying the received data to be classified through the trained classification model to obtain at least one data label of the data to be classified. And the trained classification model is obtained by training a preset classification model according to the interval focus loss function. The distance focus loss function can represent the difference between the predicted label of the sample data and the preset label of the sample data according to the distance between the predicted label of the sample data and the preset label of the sample data. Therefore, the preset classification model is trained through the interval focus loss function, when sample data is classified and trained, the number of samples does not need to be considered, and only the classification boundary is determined according to the difference between the prediction label of the sample data and the preset label of the sample data, so that the condition of data imbalance can be effectively reduced, and the trained classification model can more accurately classify the label of the data to be classified.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise.
Fig. 1 is a schematic view of an application scenario of a data classification method according to an embodiment of the present application;
FIG. 2 is a schematic flow chart diagram illustrating a data classification method according to an embodiment of the present application;
FIG. 3 is a schematic flow chart diagram illustrating a data classification method according to another embodiment of the present application;
fig. 4 is a schematic diagram of a sample data tag in the data classification method according to an embodiment of the present application;
FIG. 5 is a schematic flow chart diagram illustrating a data classification method according to another embodiment of the present application;
FIG. 6 is a schematic flow chart diagram illustrating a data classification method according to another embodiment of the present application;
FIG. 7 is a schematic flow chart diagram illustrating a data classification method according to another embodiment of the present application;
FIG. 8 is a schematic flow chart diagram illustrating a data classification method according to another embodiment of the present application;
FIG. 9 is a schematic structural diagram of a data classification apparatus according to an embodiment of the present application;
FIG. 10 is a schematic structural diagram of a data sorting apparatus according to another embodiment of the present application;
fig. 11 is a schematic structural diagram of a terminal device according to an embodiment of the present application.
Detailed Description
In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc. in order to provide a thorough understanding of the embodiments of the present application. It will be apparent, however, to one skilled in the art that the present application may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.
Reference throughout this specification to "one embodiment" or "some embodiments," or the like, means that a particular feature, structure, or characteristic described in connection with the embodiment is included in one or more embodiments of the present application. Thus, appearances of the phrases "in one implementation," "in some implementations," "in other implementations," or the like, in various places throughout this specification are not necessarily all referring to the same embodiment, but rather "one or more but not all embodiments" unless specifically stated otherwise. The terms "comprising," "including," "having," and variations thereof mean "including, but not limited to," unless expressly specified otherwise.
The data classification method provided by the embodiment of the application can be applied to terminal devices such as mobile phones, tablet computers, wearable devices, vehicle-mounted devices, Augmented Reality (AR)/Virtual Reality (VR) devices, notebook computers, ultra-mobile personal computers (UMPCs), netbooks, Personal Digital Assistants (PDAs), security cameras, monitoring cameras and the like, and the embodiment of the application does not limit the specific types of the terminal devices at all.
Fig. 1 shows a schematic application scenario of the data classification method provided in the present application. Referring to fig. 1, in this scenario, the image capturing device 11, the server 12, and the database 13 are included, a Communication connection is performed between the image capturing device 11 and the server 12, a Communication connection is performed between the server 12 and the database 13, and the Communication connection may be a wired network or a Wireless network, where the Wireless network may include a solution for Communication such as a Wireless Local Area Network (WLAN) (e.g., a Wi-Fi network), bluetooth, Zigbee, a mobile Communication network, Near Field Communication (NFC), Infrared technology (Infrared, IR), and the like. The wired Network may include a fiber optic Network, a telecommunications Network, an intranet, etc., such as a Local Area Network (LAN), a Wide Area Network (WAN), a Metropolitan Area Network (MAN), a Public Switched Telephone Network (PSTN), etc. The types of wireless networks and wired networks are not limited herein.
By way of example only and not limitation, the image capturing device 11 may include a tablet computer 111, a laptop computer 112, a desktop computer 113, a smart phone 114, a digital camera 115, a surveillance camera 116, and the like, and the image may be captured by capturing a real-time image through the camera, or by calling up an image stored in the image capturing device 11, or by accessing a server, a database, and the like storing the image through the image capturing device 11 and forwarding the image to the server 12.
For example, when the image capturing device 11 is a device having a photographing function, such as a smartphone 114, a digital camera 115, or a monitoring camera 116, a real-time image can be captured by the camera and sent to the server 12.
When the image acquiring device 11 is a tablet computer 111, a notebook computer 112, a desktop computer 113, or the like, the image stored therein may be sent to the server 12, and at this time, the image acquiring device 11 and the server 12 may be two separate devices, that is, the server 12 is a cloud server, a rack server, a cabinet server, a blade server, or the like; alternatively, the image capturing device 11 and the server 12 may be the same device, for example, the server 12 may be a virtual server running on the desktop computer 113, which is not limited herein.
Similarly, the database 13 may be implemented on the same device as the server 12, or may be implemented on different devices, and the implementation manner is a conventional means for those skilled in the art, and will not be described herein.
Fig. 2 shows a flow chart of a data classification method provided in an embodiment of the present application, which may be applied to terminal devices in the above scenarios, such as a tablet 111, a notebook 112, a desktop 113, a smart phone 114, a digital camera 115, or a surveillance camera 116, by way of example and not limitation.
Referring to fig. 2, the data classification method includes:
and S21, receiving the data to be classified.
It should be noted that, in the present application, pictures are all used as data to be classified for description, but the type of the data to be classified is not limited to this, for example, the data to be classified may also be data in other forms such as video, text, audio, and the like, at this time, corresponding adjustment needs to be performed according to the type of the data, and the adjustment method is a conventional means of those skilled in the art, and is not described herein again.
And S22, inputting the data to be classified into the trained classification model to obtain at least one data label of the data to be classified.
The classification model is obtained by training according to a distance focus loss function, and the distance focus loss function is used for representing the difference between the predicted label of the sample data and the preset label of the sample data according to the spacing distance between the predicted label of the sample data and the preset label of the sample data.
It should be noted that if the difference between the predicted label of the sample data and the preset label of the sample data is small, it indicates that the category is easy to distinguish, and the boundary of the category may be closer (i.e., the interval distance); otherwise, the classification is difficult to distinguish, and the classified boundary needs to be set farther, so that the distinguishing difficulty is reduced.
In this implementation manner, the received data to be classified is classified by the trained classification model, so as to obtain at least one data tag of the data to be classified. And the trained classification model is obtained by training a preset classification model according to the interval focus loss function. The distance focus loss function can represent the difference between the predicted label of the sample data and the preset label of the sample data according to the distance between the predicted label of the sample data and the preset label of the sample data. Therefore, the preset classification model is trained through the interval focus loss function, when sample data is classified and trained, the number of samples does not need to be considered, and only the classification boundary is determined according to the difference between the prediction label of the sample data and the preset label of the sample data, so that the condition of data imbalance can be effectively reduced, and the trained classification model can more accurately classify the label of the data to be classified.
Referring to fig. 2, in another embodiment of the data classification method, the training mode of the trained classification model is as follows:
and S31, acquiring at least one sample data from the preset database.
Wherein each sample data comprises at least one preset tag.
In some implementations, a plurality of sample data and at least one type of preset tag corresponding to each sample data are stored in the preset database, for example, when the sample data is a picture, if a cat and a dog exist in the picture at the same time, there are two preset tags corresponding to the sample data, which are "cat" and "dog" respectively.
The preset label can be represented by a vector containing N elements, where N is the number of label categories, N is an integer greater than 1, and the range of each element is [0,1 ].
By way of example only, and not limitation, referring to fig. 4, a schematic diagram of sample data tags is shown in fig. 4, and assuming that there are 4 types of tags in the sample data, which are respectively a square, a circle, a triangle, and a diamond, in fig. 4, there are a square 15, a circle 16, and a triangle 17, then a tag vector y preset by the sample data can be represented as y ∈ [1,1,1,0 ].
Wherein, if the value of the first element is 1, it indicates that the probability of the first class of tag (i.e. the tag of the square 15) existing in the sample data is 100%; a value of 1 for the second element indicates that there is 100% probability of the second category of labels (i.e., the labels of circle 16) being present in the sample data; a value of 1 for the third element indicates that there is a probability of 100% of the third class of tags (i.e., the tags of triangle 17) in the sample data; a fourth element with a value of 0 indicates that there is a probability of 0% of the fourth class of tags (i.e., the diamond-shaped tags) in the sample data.
And S32, acquiring the prediction label of each sample data through a preset classification model.
In some embodiments, the predicted label may also be represented by a vector containing N elements, and generally, for the same batch of sample data, the number of label categories is constant, that is, the predicted label also contains labels of N categories, as shown in the example in S31 and fig. 4, the predicted label vector of the sample data may be used
Indicating that the probability of the presence of the first class of label (i.e., the label of the square 15) in the sample data is 90% if the first element has a value of 0.9; a value of 0.7 for the second element indicates that there is a 70% probability of the second category of labels (i.e., the labels of circle 16) being present in the sample data; a value of 0.6 for the third element indicates that there is a 60% probability of the third class of label (i.e., the label of triangle 17) being present in the sample data; a fourth element value of 0 indicates that there is an 80% probability of a fourth class of tags (i.e., diamond shaped tags) being present in the sample data.
And S33, acquiring the spacing distance between the prediction label of the sample data and the preset label of the sample data.
Refer to the predicted tag vectors of sample data given in S32 and S31 and the tag vectors preset by the sample data since
The prediction result is obtained by classifying through a preset classification model, and a certain difference exists between the prediction result and y, and the difference is between the prediction label of the sample data and the preset label of the sample dataThe separation distance.
In some embodiments, the value of the tag may be predicted according to the ith class of the sample data
Value y of sample data ith type preset labeliAnd acquiring the interval distance between the ith type prediction tag and the ith type preset tag of the sample data, wherein i is an integer which is greater than or equal to 1 and less than or equal to N.
Referring to fig. 5, the method for obtaining a separation distance between an i-th class prediction tag and an i-th class preset tag of sample data may include:
and S331, subtracting the value of the ith type prediction tag from the value of the ith type preset tag, and acquiring the absolute distance between the ith type prediction tag and the ith type preset tag of the sample data.
By way of example only, and not limitation, reference is made to S31 and S32
And y, value y of class 1 Preset tag1Is 1, the value of the predictive tag of the first class
If the absolute distance between the class 1 prediction label and the class 1 preset label of the sample data is 0.9, the absolute distance between the class 1 prediction label and the class 1 preset label of the sample data is
And S332, multiplying the absolute distance by a preset scaling coefficient to obtain the interval distance between the ith type of prediction tag and the ith type of preset tag of the sample data.
In some embodiments, the preset scaling factor can be represented by λ, and then the distance is separated
Due to the fact that
And elements in yHas a value range of [0,1]]So absolute distance
Also in the range of [0,1]]In the meantime. The smaller the absolute distance of the ith class is, the more easily the ith class is distinguished, and the larger the absolute distance of the ith class is, the more hardly the ith class is distinguished.
However, the absolute distance is [0,1]]In the meantime, since the value range is narrow, it is difficult to obtain the degree that the category is easily distinguished, so the absolute distance is amplified by λ times, the absolute distance is expanded, and whether the category is easily distinguished is more easily judged, for example, referring to examples in S31 and S32, y ∈ [1,1,1,0]，
When the two classes are not zoomed again, the absolute distance of the second class is 0.3, the absolute distance of the third class is 0.4, the two classes are close to the middle point 0.5 of the value range, which indicates that the two classes are between easy distinguishing and difficult distinguishing, lambda can be set to be 4, then the two classes are zoomed according to the absolute distance, the absolute distance of the second class is enlarged to 1.2 from 0.3, the absolute distance of the third class is enlarged to 1.6 from 0.4, and the distance between the two classes and the middle point 2 of the value range is also enlarged by four times, so that the two classes are farther from the middle point of the value range, and whether the classes are easy to distinguish is judged more easily.
In the above embodiment, by amplifying the absolute distance between the i-th class prediction tag and the i-th class preset tag of the sample data and taking the amplified absolute distance as the distance between the i-th class prediction tag and the i-th class preset tag of the sample data, the difference between the prediction tag of the sample data and the preset tag of the sample data is amplified, so that the decision boundary is clearer, and the prediction tag of the sample data is more accurate when being acquired. Meanwhile, the absolute distance is obtained by subtracting the value of the ith type prediction label from the value of the ith type preset label, and for each training, the absolute distance of the ith type can be adaptively changed according to the value of the ith type prediction label, so that the obtained interval distance is more accurate, the prediction of the type of prediction label is more accurate, and the prediction effect of the classification model is improved.
And S34, calculating the maximum interval focal loss value according to the interval distance through an interval focal loss function.
The maximum distance focus loss value is used for indicating the maximum value of the distance between the predicted label of the sample data and the preset label of the sample data.
Referring to fig. 6, the maximum pitch focus loss value may be calculated in the following manner.
And S341, adjusting the value range of the ith type prediction label value according to the interval distance, the value of the ith type prediction label and the value of the ith type preset label to obtain the range-adjusted ith type prediction label value.
In some embodiments, the value range of the i-th class prediction label value is adjusted by using the distance, the value of the i-th class prediction label and the value of the i-th class preset label, and the change curvature of the output curve of the prediction label value can be adjusted on the basis of S33, so that the decision boundary of the value of the i-th class prediction label after range adjustment is clearer, and the prediction effect of the classification model is improved.
Referring to fig. 7, adjusting a value range of the ith class prediction tag value according to the separation distance, the value of the ith class prediction tag, and the value of the ith class preset tag to obtain an ith class prediction tag value after the range adjustment, which may include:
s3411, multiplying the ith type preset label value by two and then subtracting one to obtain the mapped ith type preset label value.
S3412, subtracting the product of the interval distance and the mapped ith type preset label value from the ith type predicted label value to obtain the mapped ith type predicted label value.
And S3413, multiplying the mapped ith type of prediction tag value by a preset range scaling coefficient to obtain the range-adjusted ith type of prediction tag value.
In some embodiments, the steps in S3411, S3412, and S3413 may be expressed by a formula, that is, the range-adjusted i-th class predicted label value
The calculation method is as follows:
where s is a scaling factor.
By way of example only, and not limitation, with reference to the examples in S31, S32,
and yiAll values of (1) are [0,1]]And s can be set to a value of 10, then
Is in the range of [ -10mi,10+10mi]。
Relative to
When the number of the types of the predicted labels is the same, the change curvature of the output curve of the predicted label value is larger, the difference between the predicted label values of different types is larger, and the decision boundary of the ith type of predicted label is clearer.
And S342, obtaining a maximum interval focus loss value according to the ith class prediction label value and the interval focus loss function after the range adjustment.
The method for obtaining the maximum distance focal point loss value can comprise the following steps:
s3421, carrying out secondary classification on the ith class prediction label value after range adjustment, and obtaining the ith class prediction label value after secondary classification.
Since the decision boundary of the i-th class predicted label value after the range adjustment is already clear, the i-th class predicted label value after the range adjustment needs to be classified twice to determine whether each type of label exists in the sample data.
There are various ways of classifying two, such as using Sigmoid function, Logistic regression, etc.
By way of example only and not limitation, a class i predicted tag value after two classification when calculated using Sigmoid function
Can be expressed by the following formula:
s3422, obtaining a maximum distance focus loss value according to the ith class prediction label value after the second classification and the distance focus loss function.
In some embodiments, the pitch focus loss function is:
will be provided with
As
(i.e. the
) Substituting the formula to obtain a maximum distance focus loss function:
wherein, wi 0Representing the weight of the corresponding loss function when the prediction label of the ith category does not exist in the sample data; w is ai 1The weight of the corresponding loss function when the prediction label of the ith category exists in one sample data is represented, and the calculation mode is as follows:
α and β are preset parameters, and in some embodiments, α is 0.5 and β is 2, but not limited thereto.
And finally, calculating to obtain the maximum focus loss value of each category through the maximum distance focus loss function.
And S35, training a preset classification model according to the maximum distance focus loss value, and obtaining the trained classification model.
It should be noted that the maximum distance focus loss value may be used to perform back propagation, multiple iterations, and repeated training of the preset classification model to finally obtain the trained classification model, and the specific training method is not limited herein.
Here, an application scenario of the data classification method provided in the present application is described with reference to training of an automatic image classification model as an example.
First, a large number of image samples are collected, and D { < x may be used as sample datai,yi>I |, 1,2,3.. N }. Wherein xiIs an image sample, yiAre a plurality of category labels corresponding to the image samples.
Then, a machine-learned classification model is determined, where a convolutional neural network f may be usedθWhere θ is a parameter of the model.
Next, B image samples are input into a convolutional neural network fθAnd updating the parameter theta of the convolutional neural network according to the following formula:
wherein L is the maximum focus loss value calculated by the maximum focus loss function provided in this application.
Then, iterating the previous step for T times until the model converges or L is smaller than a preset threshold value, and obtaining the trained classification model
Finally, inputting the image x to be predicted into the trained classification model
Outputting multi-class label vector of image to be predicted
It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.
Fig. 9 shows a schematic structural diagram of a data classification apparatus provided in an embodiment of the present application, corresponding to the data classification method described in the foregoing embodiment, and only shows portions related to the embodiment of the present application for convenience of description.
Referring to fig. 9, the apparatus includes: the receiving module 51 is configured to receive data to be classified. The classification module 52 is configured to input the data to be classified into the trained classification model, so as to obtain at least one data label of the data to be classified, where the classification model is obtained by training according to a distance focus loss function, and the distance focus loss function is used to represent a difference between a predicted label of the sample data and a preset label of the sample data according to a separation distance between the predicted label of the sample data and the preset label of the sample data.
In some implementations, referring to fig. 10, the apparatus further includes a training module 53 for obtaining a trained classification model according to the following steps: at least one sample data is obtained from a preset database, wherein each sample data comprises at least one preset label. And then, acquiring a prediction label of each sample data through a preset classification model. And then obtaining the spacing distance between the prediction label of the sample data and the preset label of the sample data. And calculating a maximum distance focus loss value through a distance focus loss function according to the distance between the sample data and the preset label of the sample data, wherein the maximum distance focus loss value is used for indicating the maximum value of the difference between the predicted label of the sample data and the preset label of the sample data. And finally, training a preset classification model according to the maximum distance focus loss value to obtain a trained classification model.
It should be noted that the prediction tag includes N classes, where N is an integer greater than 1.
Correspondingly, the training module 53 is specifically configured to obtain a separation distance between the ith class prediction tag and the ith class preset tag of the sample data according to the value of the ith class prediction tag of the sample data and the value of the ith class preset tag of the sample data, where i is an integer greater than or equal to 1 and less than or equal to N.
In some implementation manners, the training module 53 is specifically configured to subtract the value of the i-th class prediction label from the value of the i-th class preset label to obtain an absolute distance between the i-th class prediction label and the i-th class preset label of the sample data. And multiplying the absolute distance by a preset scaling coefficient to obtain the interval distance between the ith type of prediction tag and the ith type of preset tag of the sample data.
Optionally, the training module 53 is specifically configured to adjust a value range of the ith class prediction label value according to the separation distance, the value of the ith class prediction label, and the value of the ith class preset label, so as to obtain the range-adjusted ith class prediction label value. And obtaining a maximum distance focus loss value according to the ith type prediction label value and the distance focus loss function after the range adjustment.
In some implementation manners, the training module 53 is specifically configured to multiply the i-th class preset label value by two and then subtract by one to obtain the mapped i-th class preset label value. And subtracting the product of the spacing distance and the mapped ith type preset label value from the ith type predicted label value to obtain the mapped ith type predicted label value. And finally, multiplying the mapped ith type of prediction tag value by a preset range scaling coefficient to obtain the ith type of prediction tag value after range adjustment.
Optionally, the training module 53 is specifically configured to perform secondary classification on the range-adjusted i-th class predicted label value to obtain the i-th class predicted label value after the secondary classification. And then obtaining a maximum distance focus loss value according to the ith class prediction label value after the second classification and the distance focus loss function.
It should be noted that, for the information interaction, the execution process, and other contents between the above-mentioned apparatuses, the specific functions and the technical effects of the embodiments of the method of the present application are based on the same concept, and specific reference may be made to the section of the embodiments of the method, which is not described herein again.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-mentioned functions. Each functional unit and module in the embodiments may be integrated in one processing unit, or each unit may exist alone physically, or two or more units are integrated in one unit, and the integrated unit may be implemented in a form of hardware, or in a form of software functional unit. In addition, specific names of the functional units and modules are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present application. The specific working processes of the units and modules in the system may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
Fig. 11 shows a schematic structural diagram of a terminal device provided in an embodiment of the present application, and referring to fig. 11, the terminal device 6 includes:
a memory 62, a processor 61 and a computer program 63 stored in the memory 62 and executable on the processor 61, the steps in the above described respective method embodiments being implemented when the computer program 63 is executed by the processor 61.
The embodiments of the present application further provide a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the computer program implements the steps in the above-mentioned method embodiments.
The embodiments of the present application provide a computer program product, which when running on a mobile terminal, enables the mobile terminal to implement the steps in the above method embodiments when executed.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium and can implement the steps of the embodiments of the methods described above when the computer program is executed by a processor. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer readable medium may include at least: any entity or device capable of carrying computer program code to a photographing apparatus/terminal apparatus, a recording medium, computer Memory, Read-Only Memory (ROM), random-access Memory (RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium. Such as a usb-disk, a removable hard disk, a magnetic or optical disk, etc. In certain jurisdictions, computer-readable media may not be an electrical carrier signal or a telecommunications signal in accordance with legislative and patent practice.
In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and reference may be made to the related descriptions of other embodiments for parts that are not described or illustrated in a certain embodiment.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
In the embodiments provided in the present application, it should be understood that the disclosed apparatus/network device and method may be implemented in other ways. For example, the above-described apparatus/network device embodiments are merely illustrative, and for example, the division of the modules or units is only one logical division, and there may be other divisions when actually implementing, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not implemented. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present application and are intended to be included within the scope of the present application.
Claims (10)
1. A method of data classification, comprising:
receiving data to be classified;
inputting the data to be classified into the trained classification model to obtain at least one data label of the data to be classified, wherein the classification model is obtained by training according to a distance focus loss function, and the distance focus loss function is used for representing the difference between a predicted label of sample data and a preset label of the sample data according to the interval distance between the predicted label of the sample data and the preset label of the sample data.
2. The method of claim 1, wherein the trained classification model is trained by:
acquiring at least one sample data from a preset database, wherein each sample data comprises at least one preset label;
acquiring a prediction label of each sample data through a preset classification model;
acquiring a spacing distance between a prediction label of the sample data and a preset label of the sample data;
calculating a maximum distance focus loss value through the distance focus loss function according to the spacing distance, wherein the maximum distance focus loss value is used for indicating the maximum value of the difference between the predicted label of the sample data and the preset label of the sample data;
and training the preset classification model according to the maximum distance focus loss value to obtain the trained classification model.
3. The method of claim 2, wherein the predictive tag comprises N classes, wherein N is an integer greater than 1;
correspondingly, the obtaining of the separation distance between the predicted label of the sample data and the preset label of the sample data includes:
and obtaining the interval distance between the ith type prediction label and the ith type preset label of the sample data according to the value of the ith type prediction label of the sample data and the value of the ith type preset label of the sample data, wherein i is an integer which is greater than or equal to 1 and less than or equal to N.
4. The method according to claim 3, wherein the obtaining a separation distance between the ith class prediction tag and the ith class preset tag of the sample data according to the value of the ith class prediction tag of the sample data and the value of the ith class preset tag of the sample data comprises:
subtracting the value of the ith type prediction label from the value of the ith type preset label to obtain the absolute distance between the ith type prediction label and the ith type preset label of the sample data;
and multiplying the absolute distance by a preset scaling coefficient to obtain the interval distance between the ith type of prediction label and the ith type of preset label of the sample data.
5. The method according to claim 3 or 4, wherein the calculating the maximum pitch focus loss value according to the separation distance by the pitch focus loss function comprises:
adjusting the value range of the ith type predicted tag value according to the interval distance, the value of the ith type predicted tag and the value of the ith type preset tag to obtain the range-adjusted ith type predicted tag value;
and obtaining the maximum distance focus loss value according to the ith type prediction label value after the range adjustment and the distance focus loss function.
6. The method according to claim 5, wherein the adjusting a value range of the i-th class predicted tag value according to the separation distance, the i-th class predicted tag value and the i-th class preset tag value to obtain the range-adjusted i-th class predicted tag value comprises:
multiplying the ith type preset label value by two and then subtracting one to obtain a mapped ith type preset label value;
subtracting the product of the spacing distance and the mapped ith type preset label value from the ith type predicted label value to obtain a mapped ith type predicted label value;
and multiplying the mapped ith type of prediction tag value by a preset range scaling coefficient to obtain the range-adjusted ith type of prediction tag value.
7. The method according to claim 5, wherein the obtaining the maximum distance focus loss value according to the i-th class predicted label value after the range adjustment and the distance focus loss function comprises:
performing secondary classification on the ith class prediction label value after the range adjustment to obtain the ith class prediction label value after the secondary classification;
and acquiring the maximum distance focus loss value according to the ith class prediction label value after the second classification and the distance focus loss function.
8. A data sorting apparatus, comprising:
the receiving module is used for receiving the data to be classified;
and the classification module is used for inputting the data to be classified into the trained classification model to obtain at least one data label of the data to be classified, wherein the classification model is obtained by training according to a distance focus loss function, and the distance focus loss function is used for representing the difference between a predicted label of sample data and a preset label of the sample data according to the interval distance between the predicted label of the sample data and the preset label of the sample data.
9. A terminal device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the method according to any of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1 to 7.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911150175.XA CN110929785B (en) | 2019-11-21 | 2019-11-21 | Data classification method, device, terminal equipment and readable storage medium |
| PCT/CN2020/128856 WO2021098618A1 (en) | 2019-11-21 | 2020-11-13 | Data classification method and apparatus, terminal device and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911150175.XA CN110929785B (en) | 2019-11-21 | 2019-11-21 | Data classification method, device, terminal equipment and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110929785A true CN110929785A (en) | 2020-03-27 |
| CN110929785B CN110929785B (en) | 2023-12-05 |
Family
ID=69850664
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911150175.XA Active CN110929785B (en) | 2019-11-21 | 2019-11-21 | Data classification method, device, terminal equipment and readable storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN110929785B (en) |
| WO (1) | WO2021098618A1 (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111330871A (en) * | 2020-03-31 | 2020-06-26 | 新华三信息安全技术有限公司 | Quality classification method and device |
| CN112054967A (en) * | 2020-08-07 | 2020-12-08 | 北京邮电大学 | Network traffic classification method and device, electronic equipment and storage medium |
| WO2021098618A1 (en) * | 2019-11-21 | 2021-05-27 | 中国科学院深圳先进技术研究院 | Data classification method and apparatus, terminal device and readable storage medium |
| CN112884569A (en) * | 2021-02-24 | 2021-06-01 | 中国工商银行股份有限公司 | Credit assessment model training method, device and equipment |
| CN113807400A (en) * | 2021-08-17 | 2021-12-17 | 西安理工大学 | Hyperspectral image classification method, system and equipment based on anti-attack |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117633456B (en) * | 2023-11-17 | 2024-05-31 | 国网江苏省电力有限公司 | Marine wind power weather event identification method and device based on self-adaptive focus loss |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106599913A (en) * | 2016-12-07 | 2017-04-26 | 重庆邮电大学 | Cluster-based multi-label imbalance biomedical data classification method |
| CN109189767A (en) * | 2018-08-01 | 2019-01-11 | 北京三快在线科技有限公司 | Data processing method, device, electronic equipment and storage medium |
| CN109635677A (en) * | 2018-11-23 | 2019-04-16 | 华南理工大学 | Combined failure diagnostic method and device based on multi-tag classification convolutional neural networks |
| CN109816092A (en) * | 2018-12-13 | 2019-05-28 | 北京三快在线科技有限公司 | Deep neural network training method, device, electronic equipment and storage medium |
| WO2019100724A1 (en) * | 2017-11-24 | 2019-05-31 | 华为技术有限公司 | Method and device for training multi-label classification model |
| WO2019100723A1 (en) * | 2017-11-24 | 2019-05-31 | 华为技术有限公司 | Method and device for training multi-label classification model |
| CN110147456A (en) * | 2019-04-12 | 2019-08-20 | 中国科学院深圳先进技术研究院 | A kind of image classification method, device, readable storage medium storing program for executing and terminal device |
| CN110163252A (en) * | 2019-04-17 | 2019-08-23 | 平安科技（深圳）有限公司 | Data classification method and device, electronic equipment, storage medium |
| CN110442722A (en) * | 2019-08-13 | 2019-11-12 | 北京金山数字娱乐科技有限公司 | Method and device for training classification model and method and device for data classification |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015070314A1 (en) * | 2013-11-13 | 2015-05-21 | Yong Liu | Supervised credit classifier with accounting ratios |
| CN110413791A (en) * | 2019-08-05 | 2019-11-05 | 哈尔滨工业大学 | File classification method based on CNN-SVM-KNN built-up pattern |
| CN110929785B (en) * | 2019-11-21 | 2023-12-05 | 中国科学院深圳先进技术研究院 | Data classification method, device, terminal equipment and readable storage medium |
-
2019
- 2019-11-21 CN CN201911150175.XA patent/CN110929785B/en active Active
-
2020
- 2020-11-13 WO PCT/CN2020/128856 patent/WO2021098618A1/en active Application Filing
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106599913A (en) * | 2016-12-07 | 2017-04-26 | 重庆邮电大学 | Cluster-based multi-label imbalance biomedical data classification method |
| WO2019100724A1 (en) * | 2017-11-24 | 2019-05-31 | 华为技术有限公司 | Method and device for training multi-label classification model |
| WO2019100723A1 (en) * | 2017-11-24 | 2019-05-31 | 华为技术有限公司 | Method and device for training multi-label classification model |
| CN109840530A (en) * | 2017-11-24 | 2019-06-04 | 华为技术有限公司 | The method and apparatus of training multi-tag disaggregated model |
| CN109840531A (en) * | 2017-11-24 | 2019-06-04 | 华为技术有限公司 | The method and apparatus of training multi-tag disaggregated model |
| CN109189767A (en) * | 2018-08-01 | 2019-01-11 | 北京三快在线科技有限公司 | Data processing method, device, electronic equipment and storage medium |
| CN109635677A (en) * | 2018-11-23 | 2019-04-16 | 华南理工大学 | Combined failure diagnostic method and device based on multi-tag classification convolutional neural networks |
| CN109816092A (en) * | 2018-12-13 | 2019-05-28 | 北京三快在线科技有限公司 | Deep neural network training method, device, electronic equipment and storage medium |
| CN110147456A (en) * | 2019-04-12 | 2019-08-20 | 中国科学院深圳先进技术研究院 | A kind of image classification method, device, readable storage medium storing program for executing and terminal device |
| CN110163252A (en) * | 2019-04-17 | 2019-08-23 | 平安科技（深圳）有限公司 | Data classification method and device, electronic equipment, storage medium |
| CN110442722A (en) * | 2019-08-13 | 2019-11-12 | 北京金山数字娱乐科技有限公司 | Method and device for training classification model and method and device for data classification |
Non-Patent Citations (2)
| Title |
|---|
| 刘端阳 等: "基于加权 SVM 主动学习的多标签分类", 《计算机工程》, vol. 37, no. 8, pages 181 - 185 * |
| 蒋俊钊 等: "基于标签相关性的卷积神经网络多标签分类算法", 《工业控制计算机》, vol. 31, no. 7, pages 105 - 109 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021098618A1 (en) * | 2019-11-21 | 2021-05-27 | 中国科学院深圳先进技术研究院 | Data classification method and apparatus, terminal device and readable storage medium |
| CN111330871A (en) * | 2020-03-31 | 2020-06-26 | 新华三信息安全技术有限公司 | Quality classification method and device |
| CN112054967A (en) * | 2020-08-07 | 2020-12-08 | 北京邮电大学 | Network traffic classification method and device, electronic equipment and storage medium |
| CN112884569A (en) * | 2021-02-24 | 2021-06-01 | 中国工商银行股份有限公司 | Credit assessment model training method, device and equipment |
| CN113807400A (en) * | 2021-08-17 | 2021-12-17 | 西安理工大学 | Hyperspectral image classification method, system and equipment based on anti-attack |
| CN113807400B (en) * | 2021-08-17 | 2024-03-29 | 西安理工大学 | Hyperspectral image classification method, hyperspectral image classification system and hyperspectral image classification equipment based on attack resistance |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110929785B (en) | 2023-12-05 |
| WO2021098618A1 (en) | 2021-05-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110929785B (en) | Data classification method, device, terminal equipment and readable storage medium | |
| US11816570B2 (en) | Method for accelerated detection of object in videos, server, and non-transitory computer readable storage medium | |
| CN109308490B (en) | Method and apparatus for generating information | |
| CN108197652B (en) | Method and apparatus for generating information | |
| TW202030637A (en) | Method and device, electronic equipment for face image recognition and storage medium thereof | |
| JP6731529B1 (en) | Single-pixel attack sample generation method, device, equipment and storage medium | |
| CA3066029A1 (en) | Image feature acquisition | |
| WO2020186887A1 (en) | Target detection method, device and apparatus for continuous small sample images | |
| CN113095346A (en) | Data labeling method and data labeling device | |
| CN109961032B (en) | Method and apparatus for generating classification model | |
| CN114169381A (en) | Image annotation method and device, terminal equipment and storage medium | |
| CN114282581A (en) | Training sample obtaining method and device based on data enhancement and electronic equipment | |
| EP4343616A1 (en) | Image classification method, model training method, device, storage medium, and computer program | |
| CN114170484B (en) | Picture attribute prediction method and device, electronic equipment and storage medium | |
| CN113140012B (en) | Image processing method, device, medium and electronic equipment | |
| US20100169292A1 (en) | Slideshow video file detection | |
| CN113919361A (en) | Text classification method and device | |
| CN113052025A (en) | Training method of image fusion model, image fusion method and electronic equipment | |
| CN112766402A (en) | Algorithm selection method and device and electronic equipment | |
| KR20220058915A (en) | Image detection and related model training methods, apparatus, apparatus, media and programs | |
| CN111898658B (en) | Image classification method and device and electronic equipment | |
| CN112214639B (en) | Video screening method, video screening device and terminal equipment | |
| CN115170838A (en) | Data screening method and device | |
| CN114549502A (en) | Method and device for evaluating face quality, electronic equipment and storage medium | |
| CN114170271A (en) | Multi-target tracking method with self-tracking consciousness, 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 |