CN111325278A - Image processing method, device and storage medium - Google Patents
Image processing method, device and storage medium Download PDFInfo
- Publication number
- CN111325278A CN111325278A CN202010121222.4A CN202010121222A CN111325278A CN 111325278 A CN111325278 A CN 111325278A CN 202010121222 A CN202010121222 A CN 202010121222A CN 111325278 A CN111325278 A CN 111325278A
- Authority
- CN
- China
- Prior art keywords
- data
- training
- prediction
- image data
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/55—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Databases & Information Systems (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the application provides an image processing method, an image processing device and a storage medium, wherein the method comprises the following steps: acquiring a prediction result set of a first prediction model; the prediction result set comprises a prediction result for each image data to be predicted in at least one image data to be predicted; according to the prediction result set, determining training data with abnormal prediction from the at least one data to be predicted; training a preset model by using the training data and correct marking result information of the training data to obtain a second prediction model; and predicting the first data to be predicted by using the second prediction model to obtain a prediction result of the first data to be predicted. By the method and the device, the prediction accuracy of the model can be effectively improved.
Description
Technical Field
The present application relates to the field of computer technologies, and in particular, to an image processing method and apparatus, and a storage medium.
Background
With the popularization of artificial intelligence, deep learning technology has become a research field of great interest as an inevitable way to implement artificial intelligence. When the deep learning technique is applied to target detection, a large amount of image data is required to be used for model training, so as to predict image data to be predicted. These image data used for model training often require a long time for manual collection and sorting. However, in the practical application process, we can easily find that the prediction accuracy of the model trained in this way is always limited and difficult to improve.
Disclosure of Invention
The embodiment of the application provides an image processing method, an image processing device and a storage medium, which can use image data to be predicted for model training, and further effectively improve the prediction precision of a model.
In a first aspect, an embodiment of the present application provides an image processing method, including:
acquiring a prediction result set of a first prediction model; the prediction result set comprises a prediction result for each image data to be predicted in at least one image data to be predicted;
according to the prediction result set, determining training data with abnormal prediction from the at least one data to be predicted;
training a preset model by using the training data and correct marking result information of the training data to obtain a second prediction model;
and predicting the first data to be predicted by using the second prediction model to obtain a prediction result of the first data to be predicted.
Optionally, the method further comprises:
determining target training data with normal prediction from the at least one image data to be predicted;
adding the target training data with normal prediction and the training data with abnormal prediction to a specified data set;
the training of the preset model by using the training data and the correct marking result information of the training data to obtain a second prediction model comprises the following steps:
and training a preset model by using each image data in the specified data set and the correct marking result information of each image data to obtain a second prediction model.
Optionally, before the preset model is trained by using each image data in the designated data set and the correct labeling result information of each image data to obtain a second prediction model, the method further includes:
if the number of the image data included in the specified data set is smaller than the first number, increasing the number of the image data included in the specified data set by adopting a data enhancement method;
and after the number of the image data included in the specified data set is larger than or equal to a first number, performing training on a preset model by using each image data in the specified data set and correct marking result information of each image data to obtain a second prediction model.
Optionally, before the preset model is trained by using each image data in the designated data set and the correct labeling result information of each image data to obtain a second prediction model, the method further includes:
when the specified data set comprises the image data of the target category with the quantity smaller than that of the image data of other categories except the target category, increasing the quantity of the image data of the target category by adopting a data enhancement method;
and after the number of the image data of the target category is larger than or equal to a second number, performing training on a preset model by using each image data in the specified data set and correct marking result information of each image data to obtain a second prediction model.
Optionally, the determining, according to the prediction result set, training data with abnormal prediction from the at least one data to be predicted includes:
outputting a predicted result interface, wherein the predicted result interface comprises the predicted result set;
and acquiring a prediction result of second image data to be predicted selected on the prediction result interface, and determining the second image data to be predicted as training data with abnormal prediction.
Optionally, the method further comprises:
in a training time period, training a preset model by using the first to-be-predicted image data and the corresponding correct marking result information to obtain a second prediction model; or the like, or, alternatively,
and in a non-working time period, training a preset model by using the first to-be-predicted image data and the corresponding correct marking result information to obtain a second prediction model.
Optionally, the method further comprises:
interrupting the model training process in the working time period, and continuously executing the interrupted model training process in the training time period or the non-working time period;
the model training process refers to a process of training a preset model by using the training data and the correct marking result information of the training data to obtain a second prediction model.
Optionally, the method further comprises:
when a function call request for an image detection function is monitored in the model training process, interrupting the model training process and providing a corresponding function according to the function call request;
and if the function quitting request of the image detection function is monitored, continuously executing the interrupted model training process within a preset training time period or a non-working time period.
In a second aspect, an embodiment of the present application provides a data processing apparatus, including:
the acquisition module is used for acquiring a prediction result set of the first prediction model; the prediction result set comprises a prediction result for each image data to be predicted in at least one image data to be predicted;
the determining module is used for determining training data with abnormal prediction from the at least one data to be predicted according to the prediction result set;
and the processing module is used for training a preset model by using the training data and the correct marking result information of the training data to obtain a second prediction model, and predicting the first to-be-predicted image data by using the second prediction model to obtain a prediction result of the first to-be-predicted image data.
In a third aspect, an embodiment of the present application provides an electronic device, including a processor and a memory, where the processor and the memory are connected to each other, where the memory is used to store a computer program, and the computer program includes program instructions, and the processor is configured to call the program instructions to execute the method according to the first aspect.
In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program, where the computer program is executed by a processor to implement the method according to the first aspect.
In summary, the electronic device may obtain a prediction result set of the first prediction model, and determine training data with abnormal prediction from at least one data to be predicted according to the prediction result set, so as to train a preset model by using the training data and correct labeling result information of the training data to obtain a second prediction model, and further predict the first data to be predicted by using the second prediction model to obtain a prediction result of the first data to be predicted, so as to improve prediction accuracy of the model.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic diagram of an image processing scheme provided by an embodiment of the present application;
fig. 2 is a schematic flowchart of an image processing method according to an embodiment of the present application;
FIG. 3a is a schematic diagram of a predicted result interface provided by an embodiment of the present application;
FIG. 3b is a schematic diagram of another predicted outcome interface provided in the present application based on FIG. 3 a;
fig. 4 is a schematic flowchart of another image processing method according to an embodiment of the present application.
Fig. 5 is a schematic structural diagram of an image processing apparatus according to an embodiment of the present application;
fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application.
In order to effectively predict the image data, such as performing classification prediction, a deep learning method may be used to predict the image data to obtain a prediction result. For example, a target detection method based on deep learning may be adopted to perform prediction (including target object positioning and target object classification) on the image data, and obtain a prediction result. In order to improve the prediction accuracy and reduce the labor cost, the target detection has been switched from the traditional way of manually extracting features to the current way of extracting features based on a deep learning model. Therefore, in the image prediction process, the prediction accuracy can be improved, high-efficiency service can be well provided for the user, and the method can continuously and autonomously learn and is a key point needing optimization at present.
Based on this, the embodiment of the application provides an image processing scheme, and the image processing scheme can be applied to electronic equipment. The electronic device may be a terminal or a server. The terminal includes but is not limited to a notebook, a computer and other intelligent terminals. The server may be a server or a cluster of servers. In the image processing scheme, on the one hand, a set of prediction results of the first prediction model may be obtained; the set of prediction results includes a prediction result for each of the at least one image data to be predicted. On the other hand, training data with abnormal prediction can be determined from the at least one data to be predicted according to the prediction result set. In another aspect, the training data and the correct labeling result information of the training data may be used to train a preset model to obtain a second prediction model, so that the second prediction model may be used to predict the first to-be-predicted image data to obtain a prediction result of the first to-be-predicted image data. By adopting the image processing scheme, the prediction precision can be improved, high-efficiency service can be provided for users well, and the users can learn independently and continuously.
In one embodiment, the electronic device may be run with an image detection system, which may refer to a service program capable of providing image detection and the like. The image processing scheme may be implemented by invoking the image detection function. In one embodiment, the image detection system may also be referred to as an object detection system.
The image processing scheme is described below in conjunction with fig. 1. The image processing scheme shown in fig. 1 may include steps S101-S112. Specifically, the method comprises the following steps:
in steps S101 to S103, target detection may be performed on the new data by using an existing network model, so as to obtain a detection result. That is, the first prediction model may be used to predict at least one piece of data to be predicted, and a prediction result for each piece of data to be predicted in the at least one piece of data to be predicted is obtained.
In step S104, data with a wrong prediction can be determined from the new data by combining with a manual screening method, and image data with a correct prediction can be obtained.
In step S105 and step S106, a correct marking result of the prediction error data may be obtained in combination with a manual correction. That is, the training data with abnormal prediction can be determined from the at least one data to be predicted according to the prediction result set.
In step S107, a correct data set including image data that is predicted correctly and data that is predicted incorrectly may be obtained. That is, a specified data set including target training data that is predicted to be normal and training data that is predicted to be abnormal can be obtained.
In steps S108 to S110, after the correct data set is expanded by using the data enhancement method, the preset model may be trained in an autonomous learning manner to obtain a new network model. That is, the preset model may be trained by using each image data in the designated data set and the correct labeling result information of each image data, so as to obtain the second prediction model.
In steps S111 and S112, object detection may be performed using the new network model instead of the existing network model. That is, the prediction may be performed using the second prediction model instead of the first prediction model.
Through the process, the network model can be continuously updated and iterated, new data are fully utilized, and manual intervention and labor cost are effectively reduced. In addition, along with the long-time continuous operation of system, detection effect also can promote gradually, effectively reduces false detection rate and missing rate, promotes the prediction precision.
Referring to fig. 2, a flow chart of an image processing method according to an embodiment of the present application is shown. The image processing method can be applied to electronic equipment. Specifically, the method may comprise the steps of:
s201, obtaining a prediction result set of the first prediction model.
The electronic device can acquire at least one piece of data to be predicted, and predict the at least one piece of data to be predicted to obtain a prediction result set. The first model to be predicted is a trained model, such as a trained convolutional neural network model. The set of prediction results includes a prediction result for each of the at least one image data to be predicted.
In one embodiment, the electronic device may acquire the at least one image data to be predicted transmitted by the other device, or the electronic device may also acquire the at least one image data to be predicted that is locally saved. For example, when the electronic device is a server, the electronic device may acquire the at least one image data to be predicted transmitted by the first user terminal. The first user terminal may be a terminal corresponding to a developer, a terminal corresponding to a system administrator, or a terminal corresponding to a user using an image detection function. Accordingly, the user corresponding to the first user terminal may be a developer, a system administrator, or a user using an image detection function. In an application scenario, taking an electronic device as a server as an example, the first user terminal may display an image detection button, and a user corresponding to the first user terminal may click the image detection button. The first user terminal may send an image detection request to the electronic device in response to a click operation of the image detection button, the image detection request carrying image data to be predicted. The electronic device can predict the data of the image to be predicted by using the first prediction model, and obtain a prediction result of the data of the image to be predicted.
S202, according to the prediction result set, determining training data with abnormal prediction from the at least one data to be predicted.
Wherein the training data of the prediction anomaly may include image data for false detection and/or missed detection of the at least one image data to be predicted.
In this embodiment of the application, according to the prediction result set, the electronic device may determine, from the at least one to-be-predicted image data, training data with abnormal prediction in the following manner: the electronic equipment outputs a prediction result interface, wherein the prediction result interface comprises the prediction result set; and the electronic equipment acquires the prediction result of the second image data to be predicted selected on the prediction result interface, and determines the second image data to be predicted as the training data with abnormal prediction.
In one application scenario, see the prediction results interface shown in fig. 3a, which includes a set of prediction results including the prediction results of image data 1 (i.e., the prediction results of fig. 1), image data 2 (i.e., the prediction results of fig. 2), and image data 3 (i.e., the prediction results of fig. 3). Taking the electronic device as a server as an example, the electronic device may send the prediction result interface to the second user terminal. According to an actual application scenario, when the second user terminal is a terminal corresponding to a developer, the second user terminal may be a terminal corresponding to the developer or a terminal corresponding to a system administrator, and so on. When the first user terminal is a terminal corresponding to a system administrator, the second user terminal may be a terminal corresponding to the system administrator, and so on. When the first user terminal is a terminal corresponding to a user using the image detection function, the second user terminal may be a terminal corresponding to a developer, a terminal corresponding to a system manager, or a terminal corresponding to a user using the image detection function. And so on. Correspondingly, when the user corresponding to the second user terminal is the terminal corresponding to the system administrator, the user corresponding to the second user terminal is the system administrator. And when the second user terminal is the terminal corresponding to the user using the image detection function, the user corresponding to the second terminal is the user using the image detection function. The second user terminal may display the prediction result interface, and the user corresponding to the second user terminal may select the prediction result of the image data that is detected incorrectly and/or that is not detected correctly on the prediction result interface, for example, the user corresponding to the second user terminal may select the prediction result of the image data 1 and the prediction result of the image data 2 on the prediction result interface shown in fig. 3 a. After the user corresponding to the second user terminal clicks the confirmation button, the second user terminal may respond to the click operation on the confirmation button, obtain the prediction result of the image data 1 and the prediction result of the image data 2, which are selected by the user corresponding to the second user terminal, and return the prediction result of the image data 1 and the prediction result of the image data 2 to the electronic device, and the electronic device may determine the image data 1 and the image data 2 as image data with abnormal prediction.
In yet another application scenario, referring to fig. 3b, fig. 3b is another predicted result interface based on the predicted result interface shown in fig. 3 a. The prediction results interface shown in FIG. 3b may also include a specified list. Specifically, after the user corresponding to the second user terminal selects the prediction result of the image data subjected to false detection and/or missed detection on the prediction result interface, the second user terminal may add information such as the prediction result of the selected image data to the specified list. After the user corresponding to the second user terminal clicks the confirmation button, the second user terminal may respond to the click operation of the confirmation button, acquire the specified list, and send the specified list to the electronic device. The electronic device may determine image data 1 and image data 2 as the image data of the predicted abnormality according to the designation list.
In one embodiment, the electronic device may further acquire correct flag result information set for the second image data to be predicted. In one embodiment, the electronic device may further obtain a prediction result of second image data to be predicted selected on the prediction result interface, output the second image data to be predicted, and obtain correct flag result information set for the second image data to be predicted.
S203, training a preset model by using the training data and the correct marking result information of the training data to obtain a second prediction model.
And S204, predicting the first to-be-predicted image data by using the second prediction model to obtain a prediction result of the first to-be-predicted image data.
In steps S203 to S204, the electronic device may use the training data and the correct labeling result information of the training data as input data of a preset model to train the preset model, so as to obtain a second prediction model. The electronic device may use the first to-be-predicted image data as input data of the second prediction model, and predict the first to-be-predicted image data through the second prediction model to obtain a prediction result of the first to-be-predicted image data. The preset model may be the first prediction model or a new model. The new model refers to an untrained model. The new model may be obtained from the system backup, or may be downloaded from a designated path, and the obtaining manner is not limited in the embodiment of the present application. The correct labeling result information of the training data may refer to correct labeling result information of each image data included in the training data, such as correct labeling result information of image data 1 and correct labeling result information of image data 2. The correct marking result information here may include the correct category of the corresponding image data. In one embodiment, the correct identification result information may further include a correct annotation box for the corresponding image data.
In one embodiment, in order to avoid the problem that the model is over-fitted due to over-training of the model, and thus the prediction accuracy of the prediction model is low, the electronic device may select a suitable model according to the number of iterations to train. Specifically, the electronic device queries the iteration number of the first prediction model, and if the iteration number is greater than or equal to a preset iteration number, the training data and correct marking result information of the training data are used for training a new model to obtain a second prediction model; and if the iteration times are less than the preset iteration times, training a new model by using the training data and the correct marking result information of the training data to obtain a second prediction model.
In an embodiment, to avoid a problem that the prediction accuracy of the prediction model is low due to model under-fitting caused by insufficient training data, the electronic device may increase the number of image data included in the training data by using a data enhancement method when the number of image data included in the training data is less than a preset number, and perform a step of training the preset model by using the training data and correct labeling result information of the training data after the number of image data included in the training data is greater than or equal to the preset number to obtain a second prediction model. In one embodiment, the method of data enhancement may be a method of supervised data enhancement or a method of unsupervised data enhancement. The method of supervised data enhancement includes at least a method of geometric transformation and/or a method of color transformation class. The method of geometric transformation class includes, but is not limited to, at least one of: the image random cutting method, the image turning method, the image rotation method, the image translation method and the image scaling method. The method of color transformation class includes, but is not limited to, at least one of: a method of color dithering, a method of adding noise, a method of blurring processing, a method of color filling. The unsupervised data enhancement method can at least comprise a method for generating image data by adopting a specified model (such as a generative confrontation network model) or a method for determining data enhancement suitable for the current model training process according to a preset rule.
In an embodiment, to avoid the problem of low training efficiency caused by unbalanced training data classes, the electronic device may increase the number of image data of a first class by using a data enhancement method when the number of image data of the first class included in the training data is less than the number of image data of a second class other than the first class, and perform the step of training a preset model by using the training data and correct labeling result information of the training data to obtain a second prediction model after the number of image data of the first class is greater than or equal to a target number. The second category is a category other than the first category in at least two categories corresponding to the training data. When the number of the second classes is plural, the training data may include image data of a first class whose number is smaller than the number of image data of the second classes other than the first class, the training data may include image data of a first class whose number is smaller than the number of image data of any one of the plurality of second classes other than the first class, or the training data may include image data of a first class whose number is smaller than the number of image data of each of the plurality of second classes other than the first class. The target number may refer to the number of image data of any second category, or may also refer to the number of image data of a second category that satisfies a preset condition in the plurality of second categories, for example, the number of image data that satisfies the preset condition may be the largest number of image data, or may also refer to obtaining a number median value or a number average value according to the number of image data of each second category in the plurality of second categories, and the determination manner of the target number in the embodiment of the present application is not limited.
For example, the training data includes image data of category 1, image data of category 2, and image data of category 3, where the number of image data of category 1 is a, the number of image data of category 2 is a, the number of image data of category 3 is b, and b is much smaller than a. The electronic device may determine that the number of image data of category 3 is less than the number of image data of category 1, and the number of image data of category 3 is less than the number of image data of category 2. The electronic device may employ a data enhancement method to expand the number of image data of category 3 from b to c, where c is greater than b, for example, c may be a.
In one embodiment, the electronic device may calculate a difference between a number of first category image data included in the training data and a number of second category image data except for the first category image data, increase the number of the first category image data by using a data enhancement method when the difference is greater than a preset value, and perform the step of training a preset model by using the training data and correct labeling result information of the training data after the number of the first category image data is greater than or equal to a target number to obtain a second prediction model.
In an embodiment, the electronic device may acquire other training data besides the training data, and train the preset model in combination with the other training data and the correct labeling result information of the other training data to obtain the second prediction model. Wherein the other training data refers to training data other than the training data of the predicted abnormality. In one embodiment, the electronic device may obtain the other training data transmitted by the other device, or the electronic device may also obtain the other training data stored locally, such as training data recorded within a preset time range.
It can be seen that, in the embodiment shown in fig. 2, the electronic device may obtain a prediction result set of the first prediction model, and determine training data with abnormal prediction from the at least one data to be predicted according to the prediction result set, so as to train a preset model by using the training data and correct labeling result information of the training data to obtain a second prediction model, and further predict the first data to be predicted by using the second prediction model to obtain a prediction result of the first data to be predicted, so as to improve the prediction accuracy of the prediction model.
Please refer to fig. 4, which is a flowchart illustrating another image processing method according to an embodiment of the present disclosure. The method can be applied to electronic devices. The electronic device may be a terminal or a server. The terminal includes but is not limited to a notebook, a computer and other intelligent terminals. The server may be a server or a cluster of servers. Specifically, the method may comprise the steps of:
s401, obtaining a prediction result set of the first prediction model.
S402, according to the prediction result set, determining training data with abnormal prediction from the at least one data to be predicted.
Steps S401 to S402 can refer to steps S201 to S202 in the embodiment of fig. 2, which is not described herein again in this embodiment of the present application.
And S403, determining target training data with normal prediction from the at least one piece of data to be predicted.
S404, adding the target training data with normal prediction and the training data with abnormal prediction to a specified data set.
The electronic device may determine, as the target training data whose prediction is normal, data to be predicted other than the training data whose prediction is abnormal among the at least one data to be predicted. In one embodiment, the electronic device may select, according to a preset rule, a preset number of pieces of data to be predicted from data to be predicted, except for the training data with abnormal prediction, in the at least one piece of data to be predicted, so as to determine the target training data with normal prediction. In one embodiment, the electronic device may receive an image data selecting operation, and select, according to the image data selecting operation, corresponding image data to be predicted, from the at least one image data to be predicted, except for the training data with abnormal prediction, to serve as target training data with normal prediction.
In one embodiment, in order to avoid the problem that insufficient image data of the specified data set leads to model under-fitting and thus lower prediction accuracy of the prediction model, the electronic device may further add other training data to the specified data set. Wherein the other training data refers to training data other than the training data of the predicted normality and the predicted abnormality. In one embodiment, the electronic device may obtain the other training data transmitted by the other device, or the electronic device may also obtain the other training data stored locally, such as training data recorded within a preset time range.
In an embodiment, to avoid a problem that the number of image data included in a given data set is less than a first number, the electronic device may increase the number of image data included in the given data set by using a data enhancement method, and perform the step of training a preset model by using each image data in the given data set and correct labeling result information of each image data after the number of image data included in the given data set is greater than or equal to the first number, so as to obtain a second prediction model, when the number of image data included in the given data set is less than the first number, the electronic device may further perform the step of training the preset model by using each image data in the given data set and correct labeling result information of each image data. The data enhancement method can be a supervised data enhancement method or an unsupervised data enhancement method. Specific contents of the method for enhancing supervised data and the method for enhancing unsupervised data can be referred to in the embodiment of fig. 1, which is not described herein again in this embodiment of the present application.
In an embodiment, to avoid a problem that the image data of the designated data set is insufficient to cause model under-fitting and further cause low prediction accuracy of the prediction model, the electronic device may increase the number of image data of the target category by using a data enhancement method when the number of image data of the target category included in the designated data set is less than the number of image data of other categories except the target category, and perform the step of training the preset model by using each image data in the designated data set and correct labeling result information of each image data after the number of image data of the target category is greater than or equal to a second number to obtain a second prediction model. Wherein the other category refers to a category other than the target category in at least two categories corresponding to the specified data set. When the number of other categories is plural, the specified data set may include image data of the target category which is smaller in number than the image data of the other categories except the target category, may include image data of the target category which is smaller in number than the image data of any one of the plurality of other categories except the target category for the specified data set, or may include image data of the target category which is smaller in number than the image data of each of the plurality of other categories except the target category. The target number may refer to the number of image data of any other category, or may also refer to the number of image data of other categories satisfying a preset condition in the plurality of other categories, for example, the number of image data satisfying the preset condition may be the largest number, or may also refer to a number median value or a number average value obtained according to the number of image data of each other category in the plurality of other categories, and the determination manner of the target number in the embodiment of the present application is not limited.
For example, the specified data set includes image data of category 1, image data of category 2, image data of category 3, image data of category 4, and image data of category 5, where the number of image data of category 1 is a, the number of image data of category 2 is a, the number of image data of category 3 is b, the number of image data of category 4 is a, and the number of image data of category 5 is a. b is much smaller than a. The electronic device may determine that the number of image data of category 3 is less than the number of image data of category 1, and the number of image data of category 3 is less than the number of image data of category 2, and the number of image data of category 3 is less than the number of image data of category 4, and the number of image data of category 3 is less than the number of image data of category 5. The electronic device may employ a data enhancement method to expand the number of image data of category 3 from b to c, where c is greater than b, for example, c may be a.
In an embodiment, step S403 may also be performed before step S402, which is not limited in this embodiment.
S405, training a preset model by using each image data in the designated data set and the correct marking result information of each image data to obtain a second prediction model.
The electronic device may use each image data in the designated data set and correct labeling result information of each image data as input data of a preset model to train the preset model, so as to obtain a second prediction model. The predetermined model may be the first predictive model or a new model. The new model may be a new convolutional neural network model. The correct marking result information of each image data in the target training data which is normally predicted can be a prediction result of the image data. In one embodiment, the correct labeling result information of each image data in the target training data with normal prediction can be obtained by modifying according to the prediction result of the image data.
S406, predicting the first to-be-predicted image data by using the second prediction model to obtain a prediction result of the first to-be-predicted image data.
Step S406 may refer to step S204, which is not described herein again in this embodiment of the present application.
In one embodiment, in order to make the model training process more automated, the electronic device may perform, within a training time period, a step of training a preset model by using the first image data to be predicted and the corresponding correct labeling result information to obtain a second prediction model. For example, the preset training time period is 10:00 pm-6:00 am every day, the electronic device may perform the step of training the preset model by using the first to-be-predicted image data and the corresponding correct labeling result information within 10:00 pm-6:00 am every day to obtain the second prediction model. In one embodiment, the electronic device may perform, within a training time period, a step of training a preset model by using each image data in the designated data set and correct labeling result information of each image data to obtain a second prediction model.
Or, the electronic device may perform the step of training the preset model by using the first image data to be predicted and the corresponding correct labeling result information to obtain the second prediction model in the non-working time period. For example, the working period may be 9:00am-11:00pm per day, and the electronic device may perform the step of training the preset model by using the first image data to be predicted and the corresponding correct labeling result information to obtain the second prediction model in a period other than the working period. In an embodiment, the electronic device may further perform, during a non-operating period, a step of training a preset model by using each image data in the designated data set and correct labeling result information of each image data to obtain a second prediction model.
In one embodiment, the electronic device may interrupt the model training process during the active time period and continue to perform the interrupted model training process during the training time period or during the inactive time period; the model training process refers to a process of training a preset model by using the training data and correct marking result information of the training data to obtain a second prediction model. In an embodiment, the model training process may refer to a process of training a preset model by using each image data in the designated data set and correct labeling result information of each image data to obtain a second prediction model. In one embodiment, when the model training process is interrupted, training process data is recorded during the interruption, and the interrupted model training process is continued according to the training process data during the training period or during the non-working period.
In one embodiment, the electronic device may interrupt the model training process when a function call request for the image detection function is monitored in the model training process, and provide a corresponding function according to the function call request; when a function quit request of the image detection function is monitored, the interrupted model training process is continuously executed in a preset training time period or a non-working time period. In an application scenario, taking an electronic device as a server as an example, a third user terminal (the third user terminal may be the first user terminal, the second user terminal, or another user terminal) may access an image detection function in the image detection system (for example, the image detection function in the image detection system may be accessed according to a designated link address), and the electronic device may monitor a function call request for the image detection function and provide a corresponding function according to the function call request. If a function call request for the image detection function is monitored in the model training process, the model training process can be interrupted, and a corresponding function is provided according to the function call request. The electronic device may continue to perform the interrupted model training process within a preset training time period or within a non-operating time period when a function exit request for the image detection function is monitored. For example, the electronic device may monitor, when monitoring a function exit request for the image detection function, a function exit request from the third user terminal for the image detection function after the third user terminal stops accessing the image detection function in the image detection system or does not access the image detection function within a preset time range, and continue to perform the interrupted model training process within a preset training time period or during a non-operating time period.
It can be seen that, in the embodiment shown in fig. 4, the electronic device may add the target training data with normal prediction and the training data with abnormal prediction to the specified data set, and train the preset model by using each image data in the specified data set and the correct labeling result information of each image data to obtain the second prediction model, so as to predict the first image data to be predicted by using the second prediction model to obtain the prediction result of the first image data to be predicted, and the method enriches the image data used for model training, thereby improving the prediction accuracy of the model.
Fig. 5 is a schematic structural diagram of an image processing apparatus according to an embodiment of the present disclosure. The device can be applied to electronic equipment such as a terminal or a server. Specifically, the image processing apparatus may include:
an obtaining module 501, configured to obtain a prediction result set of a first prediction model; the set of prediction results includes a prediction result for each of the at least one image data to be predicted.
A determining module 502, configured to determine, according to the prediction result set, training data with abnormal prediction from the at least one data to be predicted.
The processing module 503 is configured to train a preset model by using the training data and the correct labeling result information of the training data to obtain a second prediction model, and predict the first to-be-predicted image data by using the second prediction model to obtain a prediction result of the first to-be-predicted image data.
In an optional embodiment, the processing module 503 is further configured to determine target training data with a normal prediction from the at least one image data to be predicted, and add the target training data with a normal prediction and the training data with an abnormal prediction to a specified data set.
In an optional implementation manner, the processing module 503 trains the preset model by using the training data and the correct labeling result information of the training data to obtain the second prediction model, specifically, trains the preset model by using each image data in the specified data set and the correct labeling result information of each image data to obtain the second prediction model.
In an optional implementation manner, the processing module 503 is further configured to, before the preset model is trained by using each image data in the specified data set and correct labeling result information of each image data to obtain a second prediction model, increase the number of image data included in the specified data set by using a data enhancement method if the number of image data included in the specified data set is smaller than a first number; and after the number of the image data included in the specified data set is larger than or equal to a first number, performing the operation of training a preset model by using each image data in the specified data set and the correct marking result information of each image data to obtain a second prediction model.
In an optional implementation manner, the processing module 503 is further configured to, before the preset model is trained by using each image data in the specified data set and correct labeling result information of each image data to obtain a second prediction model, increase the number of image data of the target category by using a data enhancement method when the number of image data of the target category included in the specified data set is smaller than the number of image data of other categories except the target category; and after the number of the image data of the target category is larger than or equal to a second number, performing training on a preset model by using each image data in the specified data set and correct marking result information of each image data to obtain a second prediction model.
In an optional implementation, the determining module 502 is specifically configured to output a prediction result interface, where the prediction result interface includes the prediction result set; and acquiring a prediction result of second image data to be predicted selected on the prediction result interface, and determining the second image data to be predicted as training data with abnormal prediction.
In an optional implementation manner, the processing module 503 is further configured to perform, in a training time period, an operation of training a preset model by using the first image data to be predicted and the corresponding correct labeling result information to obtain a second prediction model; or, in a non-working time period, performing the operation of training a preset model by using the first to-be-predicted image data and the corresponding correct marking result information to obtain a second prediction model.
In an optional embodiment, the processing module 503 is further configured to interrupt the model training process during the working time period, and continue to execute the interrupted model training process during the training time period or during the non-working time period; the model training process refers to a process of training a preset model by using the training data and the correct marking result information of the training data to obtain a second prediction model.
In an optional implementation manner, the processing module 503 is further configured to, when a function call request for the image detection function is monitored in the model training process, interrupt the model training process, and provide a corresponding function according to the function call request; and if the function quitting request of the image detection function is monitored, continuously executing the interrupted model training process within a preset training time period or a non-working time period.
In the embodiment shown in fig. 5, the image processing apparatus may obtain a prediction result set of the first prediction model, and determine training data with abnormal prediction from the at least one data to be predicted according to the prediction result set, so as to train a preset model by using the training data and correct labeling result information of the training data to obtain a second prediction model, and further predict the first data to be predicted by using the second prediction model to obtain a prediction result of the first data to be predicted, so as to improve the prediction accuracy of the prediction model.
Fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure. The electronic device in the present embodiment shown in fig. 6 may include: one or more processors 601; one or more input devices 602, one or more output devices 603, and memory 604. The processor 601, input device 602, output device 603, and memory 604 are connected by a bus or other means. In one embodiment, input device 602 and output device 603 are optional devices. The memory 604 is used for storing a computer program comprising program instructions, and the processor 601 is used for executing the program instructions stored by the memory 604.
In one embodiment, the processor 601 may be a Central Processing Unit (CPU), or other general-purpose processor, i.e., a microprocessor or any conventional processor. The memory 604 may include both read-only memory and random access memory, and provides instructions and data to the processor 601. Therefore, the processor 601 and the memory 604 are not limited herein.
In the embodiment of the present application, one or more instructions stored in a computer storage medium are loaded and executed by the processor 601 to implement the corresponding steps of the method in the corresponding embodiments; in particular implementations, at least one instruction in the computer storage medium is loaded by the processor 601 and performs the following steps:
acquiring a prediction result set of a first prediction model; the prediction result set comprises a prediction result for each image data to be predicted in at least one image data to be predicted;
according to the prediction result set, determining training data with abnormal prediction from the at least one data to be predicted;
training a preset model by using the training data and correct marking result information of the training data to obtain a second prediction model;
and predicting the first data to be predicted by using the second prediction model to obtain a prediction result of the first data to be predicted.
Optionally, the loading of the at least one instruction by the processor 601 further performs the steps of:
determining target training data with normal prediction from the at least one image data to be predicted;
adding the target training data with normal prediction and the training data with abnormal prediction to a specified data set;
the at least one instruction is loaded and executed by the processor 601, and the preset model is trained by using the training data and the correct labeling result information of the training data to obtain a second prediction model, which is specifically used for executing the following steps:
and training a preset model by using each image data in the specified data set and the correct marking result information of each image data to obtain a second prediction model.
Optionally, before training a preset model by using each image data in the designated data set and correct labeling result information of each image data to obtain a second prediction model, the at least one instruction is loaded by the processor 601 and further performs the following steps:
if the number of the image data included in the specified data set is smaller than the first number, increasing the number of the image data included in the specified data set by adopting a data enhancement method;
and after the number of the image data included in the specified data set is larger than or equal to a first number, performing the operation of training a preset model by using each image data in the specified data set and the correct marking result information of each image data to obtain a second prediction model.
Optionally, before training a preset model by using each image data in the designated data set and correct labeling result information of each image data to obtain a second prediction model, the at least one instruction is loaded by the processor 601 and further performs the following steps:
when the specified data set comprises the image data of the target category with the quantity smaller than that of the image data of other categories except the target category, increasing the quantity of the image data of the target category by adopting a data enhancement method;
and after the number of the image data of the target category is greater than or equal to a second number, performing the operation of training a preset model by using each image data in the specified data set and correct marking result information of each image data to obtain a second prediction model.
Optionally, the at least one instruction is loaded and executed by the processor 601, and training data with abnormal prediction is determined from the at least one data to be predicted according to the prediction result set, and is specifically configured to perform the following steps:
outputting, by an output device 603, a prediction result interface, the prediction result interface including the set of prediction results;
and acquiring a prediction result of second image data to be predicted selected on the prediction result interface, and determining the second image data to be predicted as training data with abnormal prediction.
Optionally, the at least one instruction is loaded by the processor 601 and further performs the steps of:
in a training time period, training a preset model by using the first to-be-predicted image data and the corresponding correct marking result information to obtain a second prediction model; or the like, or, alternatively,
and in a non-working time period, training a preset model by using the first to-be-predicted image data and the corresponding correct marking result information to obtain a second prediction model.
Optionally, the at least one instruction is loaded by the processor 601 and further performs the steps of:
interrupting the model training process in the working time period, and continuously executing the interrupted model training process in the training time period or the non-working time period;
the model training process refers to a process of training a preset model by using the training data and the correct marking result information of the training data to obtain a second prediction model.
Optionally, the at least one instruction is loaded by the processor 601 and further performs the steps of:
when a function call request for an image detection function is monitored in the model training process, interrupting the model training process and providing a corresponding function according to the function call request;
and if the function quitting request of the image detection function is monitored, continuously executing the interrupted model training process within a preset training time period or a non-working time period.
It will be understood by those skilled in the art that all or part of the processes of 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 when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
While the present disclosure has been described with reference to particular embodiments, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present disclosure.
Claims (10)
1. An image processing method, comprising:
acquiring a prediction result set of a first prediction model; the prediction result set comprises a prediction result for each image data to be predicted in at least one image data to be predicted;
according to the prediction result set, determining training data with abnormal prediction from the at least one data to be predicted;
training a preset model by using the training data and correct marking result information of the training data to obtain a second prediction model;
and predicting the first data to be predicted by using the second prediction model to obtain a prediction result of the first data to be predicted.
2. The method of claim 1, further comprising:
determining target training data with normal prediction from the at least one image data to be predicted;
adding the target training data with normal prediction and the training data with abnormal prediction to a specified data set;
the training of the preset model by using the training data and the correct marking result information of the training data to obtain a second prediction model comprises the following steps:
and training a preset model by using each image data in the specified data set and the correct marking result information of each image data to obtain a second prediction model.
3. The method of claim 2,
before the training of a preset model by using each image data in the specified data set and the correct marking result information of each image data to obtain a second prediction model, the method further comprises:
if the number of the image data included in the specified data set is smaller than the first number, increasing the number of the image data included in the specified data set by adopting a data enhancement method;
and after the number of the image data included in the specified data set is larger than or equal to a first number, performing training on a preset model by using each image data in the specified data set and correct marking result information of each image data to obtain a second prediction model.
4. The method according to claim 2, wherein before the training of the preset model by using the image data in the designated data set and the correct labeling result information of the image data to obtain the second prediction model, the method further comprises:
when the specified data set comprises the image data of the target category with the quantity smaller than that of the image data of other categories except the target category, increasing the quantity of the image data of the target category by adopting a data enhancement method;
and after the number of the image data of the target category is larger than or equal to a second number, performing training on a preset model by using each image data in the specified data set and correct marking result information of each image data to obtain a second prediction model.
5. The method according to claim 1, wherein the determining training data with abnormal prediction from the at least one data to be predicted according to the prediction result set comprises:
outputting a predicted result interface, wherein the predicted result interface comprises the predicted result set;
and acquiring a prediction result of second image data to be predicted selected on the prediction result interface, and determining the second image data to be predicted as training data with abnormal prediction.
6. The method of claim 1, further comprising:
in a training time period, training a preset model by using the first to-be-predicted image data and the corresponding correct marking result information to obtain a second prediction model; or the like, or, alternatively,
and in a non-working time period, training a preset model by using the first to-be-predicted image data and the corresponding correct marking result information to obtain a second prediction model.
7. The method of claim 5, further comprising:
interrupting the model training process in the working time period, and continuously executing the interrupted model training process in the training time period or the non-working time period;
the model training process refers to a process of training a preset model by using the training data and the correct marking result information of the training data to obtain a second prediction model.
8. The method of claim 5, further comprising:
when a function call request for an image detection function is monitored in the model training process, interrupting the model training process and providing a corresponding function according to the function call request;
and if the function quitting request of the image detection function is monitored, continuously executing the interrupted model training process within a preset training time period or a non-working time period.
9. An image processing apparatus characterized by comprising:
the acquisition module is used for acquiring a prediction result set of the first prediction model; the prediction result set comprises a prediction result for each image data to be predicted in at least one image data to be predicted;
the determining module is used for determining training data with abnormal prediction from the at least one data to be predicted according to the prediction result set;
and the processing module is used for training a preset model by using the training data and the correct marking result information of the training data to obtain a second prediction model, and predicting the first to-be-predicted image data by using the second prediction model to obtain a prediction result of the first to-be-predicted image data.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program which is executed by a processor to implement the method according to any one of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010121222.4A CN111325278B (en) | 2020-02-26 | 2020-02-26 | Image processing method, device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010121222.4A CN111325278B (en) | 2020-02-26 | 2020-02-26 | Image processing method, device and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111325278A true CN111325278A (en) | 2020-06-23 |
| CN111325278B CN111325278B (en) | 2023-08-29 |
Family
ID=71173107
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010121222.4A Active CN111325278B (en) | 2020-02-26 | 2020-02-26 | Image processing method, device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111325278B (en) |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8554703B1 (en) * | 2011-08-05 | 2013-10-08 | Google Inc. | Anomaly detection |
| US20160299938A1 (en) * | 2015-04-10 | 2016-10-13 | Tata Consultancy Services Limited | Anomaly detection system and method |
| US20160342903A1 (en) * | 2015-05-21 | 2016-11-24 | Software Ag Usa, Inc. | Systems and/or methods for dynamic anomaly detection in machine sensor data |
| US20180322234A1 (en) * | 2017-05-08 | 2018-11-08 | Globalfoundries Inc. | Prediction of process-sensitive geometries with machine learning |
| CN108960232A (en) * | 2018-06-08 | 2018-12-07 | Oppo广东移动通信有限公司 | Model training method, device, electronic equipment and computer readable storage medium |
| 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 |
| US20190080225A1 (en) * | 2017-09-11 | 2019-03-14 | Tata Consultancy Services Limited | Bilstm-siamese network based classifier for identifying target class of queries and providing responses thereof |
| CN109685131A (en) * | 2018-12-20 | 2019-04-26 | 斑马网络技术有限公司 | Automobile vehicle device system exception recognition methods and device |
| WO2019085793A1 (en) * | 2017-11-01 | 2019-05-09 | 腾讯科技（深圳）有限公司 | Image classification method, computer device and computer readable storage medium |
| CN109951635A (en) * | 2019-03-18 | 2019-06-28 | Oppo广东移动通信有限公司 | It takes pictures processing method, device, mobile terminal and storage medium |
| CN110414313A (en) * | 2019-06-06 | 2019-11-05 | 平安科技（深圳）有限公司 | Abnormal behaviour alarm method, device, server and storage medium |
| WO2019223582A1 (en) * | 2018-05-24 | 2019-11-28 | Beijing Didi Infinity Technology And Development Co., Ltd. | Target detection method and system |
| EP3579148A1 (en) * | 2018-06-08 | 2019-12-11 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Method and apparatus for training, classification model, mobile terminal, and readable storage medium |
| US20200036611A1 (en) * | 2018-07-24 | 2020-01-30 | Baidu Online Network Technology (Beijing) Co., Ltd. | Method and apparatus for warning |
-
2020
- 2020-02-26 CN CN202010121222.4A patent/CN111325278B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8554703B1 (en) * | 2011-08-05 | 2013-10-08 | Google Inc. | Anomaly detection |
| US20160299938A1 (en) * | 2015-04-10 | 2016-10-13 | Tata Consultancy Services Limited | Anomaly detection system and method |
| US20160342903A1 (en) * | 2015-05-21 | 2016-11-24 | Software Ag Usa, Inc. | Systems and/or methods for dynamic anomaly detection in machine sensor data |
| US20180322234A1 (en) * | 2017-05-08 | 2018-11-08 | Globalfoundries Inc. | Prediction of process-sensitive geometries with machine learning |
| US20190080225A1 (en) * | 2017-09-11 | 2019-03-14 | Tata Consultancy Services Limited | Bilstm-siamese network based classifier for identifying target class of queries and providing responses thereof |
| WO2019085793A1 (en) * | 2017-11-01 | 2019-05-09 | 腾讯科技（深圳）有限公司 | Image classification method, computer device and computer readable storage medium |
| WO2019223582A1 (en) * | 2018-05-24 | 2019-11-28 | Beijing Didi Infinity Technology And Development Co., Ltd. | Target detection method and system |
| CN108960232A (en) * | 2018-06-08 | 2018-12-07 | Oppo广东移动通信有限公司 | Model training method, device, electronic equipment and computer readable storage medium |
| EP3579148A1 (en) * | 2018-06-08 | 2019-12-11 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Method and apparatus for training, classification model, mobile terminal, and readable storage medium |
| US20200036611A1 (en) * | 2018-07-24 | 2020-01-30 | Baidu Online Network Technology (Beijing) Co., Ltd. | Method and apparatus for warning |
| 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 |
| CN109685131A (en) * | 2018-12-20 | 2019-04-26 | 斑马网络技术有限公司 | Automobile vehicle device system exception recognition methods and device |
| CN109951635A (en) * | 2019-03-18 | 2019-06-28 | Oppo广东移动通信有限公司 | It takes pictures processing method, device, mobile terminal and storage medium |
| CN110414313A (en) * | 2019-06-06 | 2019-11-05 | 平安科技（深圳）有限公司 | Abnormal behaviour alarm method, device, server and storage medium |
Non-Patent Citations (2)
| Title |
|---|
| SEBASTIÁN M等: "Abnormal Pattern Prediction: Detecting Fraudulent Insurance Property Claims with Semi-Supervised Machine-Learning", 《DATA SCIENCE JOURNAL 》, vol. 18, no. 1, pages 1 - 15 * |
| 张天: "基于神经网络的空气质量预测预警系统的设计与实现", 《中国优秀硕士学位论文全文数据库 信息科技辑》, no. 2, pages 140 - 1137 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111325278B (en) | 2023-08-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110276446B (en) | Method and device for training model and selecting recommendation information | |
| CN110598802B (en) | Memory detection model training method, memory detection method and device | |
| CN110781960B (en) | Training method, classification method, device and equipment of video classification model | |
| CN115421950B (en) | Automatic system operation and maintenance management method and system based on machine learning | |
| CN111400294B (en) | Data anomaly monitoring method, device and system | |
| CN110909768B (en) | Method and device for acquiring marked data | |
| CN110602207A (en) | Method, device, server and storage medium for predicting push information based on off-network | |
| CN113254153B (en) | Method and device for processing flow task, computer equipment and storage medium | |
| CN112612393B (en) | Interaction method and device of interface function | |
| CN113918438A (en) | Method and device for detecting server abnormality, server and storage medium | |
| CN110704614B (en) | Information processing method and device for predicting user group type in application | |
| CN113010785A (en) | User recommendation method and device | |
| CN111325278A (en) | Image processing method, device and storage medium | |
| CN114116688B (en) | Data processing and quality inspection method and device and readable storage medium | |
| CN115980585A (en) | Battery fault detection method and device, computer equipment and storage medium | |
| CN114842476A (en) | Watermark detection method and device and model training method and device | |
| CN111428886B (en) | Method and device for adaptively updating deep learning model of fault diagnosis | |
| CN114385398A (en) | Request response state determination method, device, equipment and storage medium | |
| CN114203304A (en) | Information pushing method based on smart medical big data and smart medical cloud server | |
| CN114265757A (en) | Equipment anomaly detection method and device, storage medium and equipment | |
| CN116450632B (en) | Geographic sample data quality evaluation method, device and storage medium | |
| CN114844778B (en) | Abnormality detection method and device for core network, electronic equipment and readable storage medium | |
| CN111352840B (en) | Online behavior risk assessment method, device, equipment and readable storage medium | |
| US10489272B2 (en) | Automatic instrumentation of code | |
| CN114238055A (en) | Task data processing method and device, 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 |