CN114519404A - Image sample classification labeling method, device, equipment and storage medium - Google Patents
Image sample classification labeling method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN114519404A CN114519404A CN202210413215.0A CN202210413215A CN114519404A CN 114519404 A CN114519404 A CN 114519404A CN 202210413215 A CN202210413215 A CN 202210413215A CN 114519404 A CN114519404 A CN 114519404A
- Authority
- CN
- China
- Prior art keywords
- image
- annotation
- classification
- labeling
- sample
- 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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Abstract
The application discloses a method, a device, equipment and a medium for image sample classification and annotation, which relate to the technical field of image processing and comprise the following steps: selecting a pre-trained image annotation model as a current target image annotation model, carrying out image annotation on an original image sample set by using the current target image annotation model, classifying and outputting a corresponding image sample set; and sequentially utilizing the other unselected image labeling models to label and classify the image sample set output by the previous target image labeling model, and eliminating the image samples which are inconsistent with the target classification type from the current classification result until the image generated by labeling and classifying the current target image labeling model meets the preset image labeling quantity, stopping image labeling and classifying and outputting the corresponding classified and labeled image sample set. By the method and the device, time cost and labor cost can be reduced, errors of manual labeling are reduced, and the accuracy of labeling is improved.
Description
Technical Field
The present invention relates to the field of image processing technologies, and in particular, to a method, an apparatus, a device, and a storage medium for classifying and labeling image samples.
Background
At present, in the early stage of image classification based on supervised learning, preprocessing labeling of images is required to obtain an image dataset carrying type labels, at present, in image classification, manual classification, namely manual labeling, is required to be performed firstly, and in order to improve the training accuracy of a model, a large amount of data labeling is required, at least 10 data labeling are required4In addition, the manual labeling for a long time always has errors.
In conclusion, how to realize simple and accurate image sample classification labeling, reduce labor and time costs, and reduce labeling errors is a problem to be solved in the field.
Disclosure of Invention
In view of the above, the present invention provides a method, an apparatus, a device, and a storage medium for image sample classification and annotation, which can realize simple and accurate image sample classification and annotation, reduce labor and time costs, and reduce annotation errors. The specific scheme is as follows:
in a first aspect, the application discloses an image sample classification and annotation method, which includes:
selecting one image annotation model from a plurality of pre-trained image annotation models as a current target image annotation model, and performing image annotation and classification on an original image sample set by using the current target image annotation model to output a current image sample set containing corresponding target classification categories;
selecting one image annotation model from the other unselected image annotation models as a current target image annotation model;
carrying out image annotation and classification on an image sample set output by a previous target image annotation model by using the current target image annotation model, and removing image samples inconsistent with the target classification type from a current classification result so as to output the current image sample set containing the target classification type;
and skipping to the other unselected image annotation models again to select one image annotation model as the current target image annotation model until a preset skipping ending condition is met, and outputting the corresponding classified and labeled image sample set.
Optionally, before selecting one image annotation model from the plurality of pre-trained image annotation models as the current target image annotation model, the method further includes:
and respectively carrying out image annotation training on different image annotation models by using different sample image data sets to obtain a plurality of trained image annotation models.
Optionally, before the image annotation training is performed on different image annotation models respectively by using different sample image data sets to obtain a plurality of trained image annotation models, the method further includes:
sampling an original image sample set, labeling corresponding data type labels for sample image data obtained by sampling, and grouping to obtain a plurality of sample image data sets.
Optionally, after the sampling an original image sample set, labeling corresponding data type labels for sample image data obtained by the sampling, and grouping the sample image data to obtain a plurality of sample image data sets, the method further includes:
the sample images in each sample image dataset are expanded to obtain an expanded sample image dataset.
Optionally, the expanding the sample image in each sample image data set to obtain an expanded sample image data set includes:
the method further includes expanding the subsample image of each category in the sample image dataset using one or more transform enhancement methods and based on preset image enhancement coefficients to obtain an expanded sample image dataset.
Optionally, the image labeling and classifying the image sample set output by the previous target image labeling model by using the current target image labeling model includes:
selecting a target classification category, determining a target image sample set corresponding to the target classification category from an image sample set output by a last target image labeling model, and labeling and classifying the target image sample set by using the current target image labeling model.
Optionally, the re-skipping to the image annotation model that is selected from the other unselected image annotation models as the current target image annotation model until a preset skipping end condition is met, and outputting a corresponding classified labeled image sample set includes:
and skipping to the other unselected image annotation models again to select one image annotation model as the current target image annotation model until the number of image samples in each target classification category reaches the preset number of image samples, and outputting the corresponding classified and labeled image sample set.
In a second aspect, the present application discloses an image sample classifying and labeling device, which includes:
the first classification labeling module is used for selecting one image labeling model from a plurality of image labeling models trained in advance as a current target image labeling model, performing image labeling and classification on an original image sample set by using the current target image labeling model, and outputting a current image sample set containing corresponding target classification categories;
the model determining module is used for selecting one image annotation model from the other unselected image annotation models as a current target image annotation model;
the second classification labeling module is used for performing image labeling and classification on an image sample set output by a previous target image labeling model by using the current target image labeling model, and eliminating image samples inconsistent with the target classification type from a current classification result so as to output the current image sample set containing the target classification type;
and the sample acquisition module is used for skipping to the other unselected image annotation models again to select one image annotation model as the current target image annotation model until a preset skipping ending condition is met, and outputting a corresponding classified and labeled image sample set.
In a third aspect, the present application discloses an electronic device, comprising:
a memory for storing a computer program;
a processor for executing the computer program to realize the steps of the image sample classification labeling method disclosed in the foregoing.
In a fourth aspect, the present application discloses a computer readable storage medium for storing a computer program; wherein the computer program realizes the steps of the image sample classification and labeling method disclosed in the foregoing when being executed by a processor.
Therefore, the application discloses an image sample classification and annotation method, which comprises the following steps: selecting one image annotation model from a plurality of pre-trained image annotation models as a current target image annotation model, and performing image annotation and classification on an original image sample set by using the current target image annotation model to output a current image sample set containing corresponding target classification categories; selecting one image annotation model from the other unselected image annotation models as a current target image annotation model; carrying out image annotation and classification on an image sample set output by a previous target image annotation model by using the current target image annotation model, and removing image samples inconsistent with the target classification type from a current classification result so as to output the current image sample set containing the target classification type; and skipping to the other unselected image annotation models again to select one image annotation model as the current target image annotation model until a preset skipping ending condition is met, and outputting the corresponding classified and labeled image sample set. Therefore, the image labeling and classifying method has the advantages that the trained image labeling models are used for sequentially performing image labeling and classifying on the original image sample set, image data which do not belong to the target classification category are eliminated when the second image labeling model is selected for performing label classification, and finally a data set with very accurate image labeling classification is obtained to serve as the image sample set, so that a large number of classification image sample sets are prevented from being manually labeled, the time cost and the labor cost are reduced, labeling errors caused by long-time manual labeling are reduced, and the labeling accuracy is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention 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 embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a flowchart of an image sample classification and annotation method disclosed in the present application;
FIG. 2 is a flow chart of a data set classification labeling process disclosed herein;
FIG. 3 is a flowchart of a specific image sample classification labeling method disclosed in the present application;
FIG. 4 is a block diagram of an exemplary image sample classification and labeling apparatus disclosed herein;
fig. 5 is a block diagram of an electronic device disclosed in the present application.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Currently, in the early stage of image classification based on supervised learning, preprocessing annotation of images is required to obtain portabilityAt present, in image classification, image data sets with type labels need to be manually classified once, namely manually labeled, and in order to improve the training accuracy of models, a large amount of data labels are required, which are at least 104In addition, the manual labeling for a long time always has errors.
Therefore, the invention correspondingly provides an image sample classifying and labeling scheme, which can realize simple and accurate image sample classifying and labeling, reduce the labor and time cost and reduce the labeling error.
Referring to fig. 1, an embodiment of the present invention discloses a flowchart of an image sample classification and annotation method, specifically including:
step S11: selecting one image annotation model from a plurality of pre-trained image annotation models as a current target image annotation model, and performing image annotation and classification on an original image sample set by using the current target image annotation model to output a current image sample set containing corresponding target classification categories.
In this embodiment, referring to fig. 2, for example: pre-training n image labeling models, selecting model n1Performing prediction classification on the original image sample set to obtain a classified image sample set xn1，xn1There are p classes.
In this embodiment, before selecting one image annotation model from a plurality of pre-trained image annotation models as the current target image annotation model, the method further includes: and respectively carrying out image annotation training on different image annotation models by using different sample image data sets to obtain a plurality of trained image annotation models. It can be understood that, image annotation training is performed on the disclosed image annotation model by using different sample image data sets to obtain a corresponding trained target image annotation model, where the image annotation model specifically may include: the private image annotation model and the public image annotation model may specifically include, but are not limited to: VGG (Visual Geometry Group, super-resolution test sequence), AlexNet, google lenet, and the like.
In this embodiment, before the image annotation training is performed on different image annotation models respectively by using different sample image data sets to obtain a plurality of trained image annotation models, the method further includes: sampling an original image sample set, labeling corresponding data type labels for sample image data obtained by sampling, and grouping to obtain a plurality of sample image data sets. It will be appreciated that the sampling is done manually from the original image sample set, for example: the original image sample set x aims at classifying and marking p types of data, wherein each type comprises q pictures, then p x q pictures are used for conducting supervision image classification based on deep learning, and conventionally, p x q pictures are manually selected from the x pictures. In this embodiment, p classes are labeled and classified from x pictures, each class of m × 100 pictures, where m × 100 < q is at least one order of magnitude, then the classified m × 100 pictures are classified into m1 × n × 100 pictures again, and finally the obtained data sets are sorted into n data sets of m1 × 100 pictures with p classes as n sample image data sets.
In this embodiment, referring to fig. 2, after sampling an original image sample set, labeling corresponding data type labels for sample image data obtained by the sampling, and grouping the sample image data to obtain a plurality of sample image data sets, the method further includes: the sample images in each sample image dataset are expanded to obtain an expanded sample image dataset. It will be appreciated that m x 100 pictures of each classification p of the n sample image datasets are expanded to yield n expanded target sample image datasets. It should be noted that, because the sample image data set classified by manual labeling is small, even if the sample image data set is expanded to obtain the corresponding target sample image data set, the number of images in the target sample image data set is still small, and therefore the target image labeling model trained by the target sample image data set is also a low-accuracy target image labeling model.
In the present embodimentExpanding the sample image in each sample image dataset to obtain an expanded sample image dataset, comprising: and expanding the subsample image of each category in the sample image dataset by using one or more transformation enhancement methods and based on preset image enhancement coefficients to obtain an expanded sample image dataset. It will be appreciated that, by expanding the n sample image datasets separately by a parametric transform enhancement method to each class m1 x 100 x k, where k is an enhancement factor, the enhancement factor for each sample image dataset may be different, but should have a minimum value kmin，kminWhich may be 20, the transform enhancement method may include, but is not limited to: clipping, flipping, rotating, warping, etc., then selecting 1 or more transform enhancement methods for each sample image data set, and it should be noted that, in order to avoid over-fitting, the sample image data sets in this embodiment cannot be enhanced and then classified.
Step S12: and selecting one image annotation model from the rest unselected image annotation models as a current target image annotation model.
In this embodiment, any one of the image annotation models that are not selected after training is selected as the current target image annotation model, for example: selecting another model n from n-1 image labeling models2As a current target image annotation model.
Step S13: and carrying out image annotation and classification on the image sample set output by the previous target image annotation model by using the current target image annotation model, and removing image samples inconsistent with the target classification type from the current classification result so as to output the current image sample set containing the target classification type.
In this embodiment, it can be understood that the image sample set xn including p classifications output by the last target image annotation model is obtained1Selecting xn1P in the data set1Sub-classified sub-image sample set, using model n2For the p1Performing prediction classification again on the sub-image sample sets of the sub-classifications to obtain a classified sub-data set xn2Discarding the data set xn2Middle non p1Classified image data, resulting in a new data set xn2. Correspondingly, for the image sample set xn1The sub-image sample sets corresponding to other sub-classifications are also modeled using model n2Repeating the steps of prediction classification in the embodiment to obtain a new data set of the classified corresponding sub-classification.
Step S14: and skipping to the other unselected image annotation models again to select one image annotation model as the current target image annotation model until a preset skipping ending condition is met, and outputting the corresponding classified and labeled image sample set.
In this embodiment, the process of returning to the step of selecting one image annotation model from the rest unselected image annotation models as the current target image annotation model is repeated, and it can be understood that one model n is selected from the n-2 models again3For model n2Predicting a sorted data set xn2Performing predictive classification, discarding image data that does not conform to the selected target class label, and generating an image sample set xn3If the image sample set xn at that time3Stopping skipping to select the image annotation model as the current target image annotation model if the preset image sample quantity condition is met, and outputting the corresponding classified and annotated image sample set xn3And finishing the acquisition of the image sample set.
Therefore, the application discloses an image sample classification and annotation method, which comprises the following steps: selecting one image annotation model from a plurality of pre-trained image annotation models as a current target image annotation model, and performing image annotation and classification on an original image sample set by using the current target image annotation model to output a current image sample set containing corresponding target classification categories; selecting one image annotation model from the other unselected image annotation models as a current target image annotation model; carrying out image annotation and classification on an image sample set output by a previous target image annotation model by using the current target image annotation model, and removing image samples inconsistent with the target classification type from a current classification result so as to output the current image sample set containing the target classification type; and skipping to the other unselected image annotation models again to select one image annotation model as the current target image annotation model until a preset skipping ending condition is met, and outputting the corresponding classified and labeled image sample set. Therefore, the image labeling and classifying method has the advantages that the trained image labeling models are used for sequentially performing image labeling and classifying on the original image sample set, image data which do not belong to the target classification category are eliminated when the second image labeling model is selected for performing label classification, and finally a data set with very accurate image labeling classification is obtained to serve as the image sample set, so that a large number of classification image sample sets are prevented from being manually labeled, the time cost and the labor cost are reduced, labeling errors caused by long-time manual labeling are reduced, and the labeling accuracy is improved.
Referring to fig. 3, the embodiment of the present invention discloses a specific image sample classification and annotation method, and compared with the previous embodiment, the present embodiment further describes and optimizes the technical solution. Specifically, the method comprises the following steps:
step S21: selecting one image annotation model from a plurality of pre-trained image annotation models as a current target image annotation model, and performing image annotation and classification on an original image sample set by using the current target image annotation model to output a current image sample set containing corresponding target classification categories.
Step S22: and selecting one image annotation model from the rest unselected image annotation models as a current target image annotation model.
Step S23: selecting a target classification category, determining a target image sample set corresponding to the target classification category from an image sample set output by a last target image labeling model, and labeling and classifying the target image sample set by using the current target image labeling model.
For more specific processing procedures in the steps S21, S22, and S23, reference may be made to corresponding contents disclosed in the foregoing embodiments, and details are not repeated here.
Step S24: and skipping to the other unselected image annotation models again to select one image annotation model as the current target image annotation model until the number of image samples in each target classification category reaches the preset number of image samples, and outputting the corresponding classified and labeled image sample set.
In this embodiment, the data set xn is skipped again and the other models are verifiednWhen the number of the images is slightly larger than a preset image number training target p x q, the repetition is stopped, and the data set xn is usednAs an image sample set, the data size of the image sample set is smaller than that of an original image sample set by one data magnitude, and is already pre-classified, and the total data size of manual labeling classification is greatly reduced, wherein, assuming that sample data is uniformly distributed, the multi-model product of each classification is finally approximately equal to the template data size q, and the multi-model product formula is as follows:
wherein the content of the first and second substances,
representing the amount of original image sample data,
representing the number of categories to be classified,
representing the number of image annotation models,
represents eachThe model accuracy of the individual image annotation models,
the amount of image data representing each classified category.
In this embodiment, for example: dividing 100 ten thousand image data into 3 classes in advance, if machine learning is needed, then the image samples needing to be labeled and classified need at least 3 ten thousand, 3000 images can be manually labeled, each class is 1000 images, then 1000 images are divided into 10 parts, a sample data set 10 x 3 x 100 is generated, the data set of 10 small samples is respectively expanded by different transformation enhancing methods and enhancing parameters to the image data of corresponding classes, 10 image labeling models are trained by the expanded data set of 10 small samples, 10 trained image labeling models are generated, then the 100 ten thousand data sets are sequentially predicted by 10 models, after each model is predicted, the data set of the same class as the last prediction in the prediction classification is selected to be trained by other image prediction models for the next time until the predicted data set is slightly larger than 1/3 ten thousand target data, when manual verification is carried out, only extremely individual abnormal data basically need to be removed through the manual verification, and therefore the task of marking 3 thousands of images can be indirectly completed only by manually marking 3000 images without manually marking 3 thousands of images at one time, marking time is greatly shortened, and marking speed is improved.
Therefore, in the embodiment, multi-level classification labeling is performed on the original image sample data set once by repeatedly using different image labeling models, and finally a small data set is obtained to serve as a final image sample set.
Referring to fig. 4, an embodiment of the present invention discloses a specific image sample classification and labeling apparatus, including:
the first classification and annotation module 11 is configured to select one image annotation model from a plurality of pre-trained image annotation models as a current target image annotation model, perform image annotation and classification on an original image sample set by using the current target image annotation model, and output a current image sample set including corresponding target classification categories;
the model determining module 12 is configured to select one image annotation model from the other unselected image annotation models as a current target image annotation model;
the second classification labeling module 13 is configured to label and classify an image sample set output by a previous target image labeling model by using the current target image labeling model, and remove an image sample inconsistent with the target classification type from a current classification result to output a current image sample set including the target classification type;
and the sample acquisition module 14 is configured to skip to the image annotation model that is not selected from the other unselected image annotation models again to select one image annotation model as the current target image annotation model until a preset skipping end condition is met, and output a corresponding classified and labeled image sample set.
Therefore, the application discloses an image sample classification and annotation method, which comprises the following steps: selecting one image annotation model from a plurality of pre-trained image annotation models as a current target image annotation model, and performing image annotation and classification on an original image sample set by using the current target image annotation model to output a current image sample set containing corresponding target classification categories; selecting one image annotation model from the other unselected image annotation models as a current target image annotation model; carrying out image annotation and classification on an image sample set output by a previous target image annotation model by using the current target image annotation model, and removing image samples inconsistent with the target classification type from a current classification result so as to output the current image sample set containing the target classification type; and skipping to the other unselected image annotation models again to select one image annotation model as the current target image annotation model until a preset skipping ending condition is met, and outputting the corresponding classified and labeled image sample set. Therefore, the image labeling and classifying method has the advantages that the trained image labeling models are used for sequentially performing image labeling and classifying on the original image sample set, image data which do not belong to the target classification category are eliminated when the second image labeling model is selected for performing label classification, and finally a data set with very accurate image labeling classification is obtained to serve as the image sample set, so that a large number of classification image sample sets are prevented from being manually labeled, the time cost and the labor cost are reduced, labeling errors caused by long-time manual labeling are reduced, and the labeling accuracy is improved.
Further, an electronic device is disclosed in the embodiments of the present application, and fig. 5 is a block diagram of the electronic device 20 according to an exemplary embodiment, which should not be construed as limiting the scope of the application.
Fig. 5 is a schematic structural diagram of an electronic device 20 according to an embodiment of the present disclosure. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input output interface 25, and a communication bus 26. The memory 22 is used for storing a computer program, and the computer program is loaded and executed by the processor 21 to implement relevant steps in the image sample classification and labeling method disclosed in any of the foregoing embodiments. In addition, the electronic device 20 in the present embodiment may be specifically an electronic computer.
In this embodiment, the power supply 23 is configured to provide a working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and an external device, and a communication protocol followed by the communication interface is any communication protocol applicable to the technical solution of the present application, and is not specifically limited herein; the input/output interface 25 is configured to obtain external input data or output data to the outside, and a specific interface type thereof may be selected according to specific application requirements, which is not specifically limited herein.
The processor 21 may include one or more processing cores, such as a 4-core processor, an 8-core processor, and the like. The processor 21 may be implemented in at least one hardware form of a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), and a PLA (Programmable Logic Array). The processor 21 may also include a main processor and a coprocessor, where the main processor is a processor for Processing data in an awake state, and is also called a Central Processing Unit (CPU); a coprocessor is a low power processor for processing data in a standby state. In some embodiments, the processor 21 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content required to be displayed on the display screen. In some embodiments, the processor 21 may further include an AI (Artificial Intelligence) processor for processing a calculation operation related to machine learning.
In addition, the storage 22 is used as a carrier for resource storage, and may be a read-only memory, a random access memory, a magnetic disk or an optical disk, etc., and the resources stored thereon may include an operating system 221, a computer program 222, etc., and the storage manner may be a transient storage or a permanent storage.
The operating system 221 is used for managing and controlling each hardware device and the computer program 222 on the electronic device 20, so as to realize the operation and processing of the mass data 223 in the memory 22 by the processor 21, and may be Windows Server, Netware, Unix, Linux, and the like. The computer program 222 may further include a computer program that can be used to perform other specific tasks in addition to the computer program that can be used to perform the image sample classification labeling method performed by the electronic device 20 disclosed in any of the foregoing embodiments. The data 223 may include data received by the electronic device and transmitted from an external device, or may include data collected by the input/output interface 25 itself.
Further, the present application also discloses a computer-readable storage medium for storing a computer program; wherein the computer program realizes the image sample classification labeling method disclosed in the foregoing when being executed by a processor. For the specific steps of the method, reference may be made to the corresponding contents disclosed in the foregoing embodiments, which are not described herein again.
The embodiments are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same or similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. 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. The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in Random Access Memory (RAM), memory, Read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The image sample classification and annotation method, apparatus, device, and storage medium provided by the present invention are described in detail above, and a specific example is applied in the present disclosure to explain the principle and the implementation of the present invention, and the description of the above embodiment is only used to help understanding the method and the core idea of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (10)
1. An image sample classification and annotation method is characterized by comprising the following steps:
selecting one image annotation model from a plurality of pre-trained image annotation models as a current target image annotation model, and performing image annotation and classification on an original image sample set by using the current target image annotation model to output a current image sample set containing corresponding target classification categories;
selecting one image annotation model from the rest unselected image annotation models as a current target image annotation model;
carrying out image annotation and classification on an image sample set output by a previous target image annotation model by using the current target image annotation model, and removing image samples inconsistent with the target classification type from a current classification result so as to output the current image sample set containing the target classification type;
and skipping to the other unselected image annotation models again to select one image annotation model as the current target image annotation model until a preset skipping ending condition is met, and outputting the corresponding classified and labeled image sample set.
2. The image sample classification and annotation method according to claim 1, wherein before selecting one of the image annotation models trained in advance as the current target image annotation model, the method further comprises:
and respectively carrying out image annotation training on different image annotation models by using different sample image data sets to obtain a plurality of trained image annotation models.
3. The image sample classification and annotation method according to claim 2, wherein before the image annotation training is performed on different image annotation models by using different sample image data sets to obtain a plurality of trained image annotation models, the method further comprises:
sampling an original image sample set, labeling corresponding data type labels for sample image data obtained by sampling, and grouping to obtain a plurality of sample image data sets.
4. The method for classifying and labeling image samples according to claim 3, wherein after sampling an original image sample set, labeling corresponding data type labels for the sampled sample image data, and grouping the labeled sample image data to obtain a plurality of sample image data sets, the method further comprises:
the sample images in each sample image dataset are expanded to obtain an expanded sample image dataset.
5. The method for classifying and labeling image samples according to claim 4, wherein the expanding the sample images in each sample image dataset to obtain an expanded sample image dataset comprises:
and expanding the subsample image of each category in the sample image dataset by using one or more transformation enhancement methods and based on preset image enhancement coefficients to obtain an expanded sample image dataset.
6. The image sample classification and annotation method according to any one of claims 1 to 5, wherein the image annotation and classification of the image sample set output by the previous target image annotation model by using the current target image annotation model comprises:
selecting a target classification category, determining a target image sample set corresponding to the target classification category from an image sample set output by a last target image labeling model, and labeling and classifying the target image sample set by using the current target image labeling model.
7. The image sample classification and annotation method according to claim 1, wherein the re-skipping to the image annotation model that is selected from the other unselected image annotation models as the current target image annotation model until a preset skipping end condition is satisfied, and outputting a corresponding classification-labeled image sample set comprises:
and skipping to the other unselected image annotation models again to select one image annotation model as the current target image annotation model until the number of image samples in each target classification category reaches the preset number of image samples, and outputting the corresponding classified and labeled image sample set.
8. An image sample classifying and labeling device is characterized by comprising:
the first classification labeling module is used for selecting one image labeling model from a plurality of image labeling models trained in advance as a current target image labeling model, performing image labeling and classification on an original image sample set by using the current target image labeling model, and outputting a current image sample set containing corresponding target classification categories;
the model determining module is used for selecting one image annotation model from the other unselected image annotation models as a current target image annotation model;
the second classification labeling module is used for performing image labeling and classification on an image sample set output by a previous target image labeling model by using the current target image labeling model, and eliminating image samples inconsistent with the target classification type from a current classification result so as to output the current image sample set containing the target classification type;
and the sample acquisition module is used for skipping to the other unselected image annotation models again to select one image annotation model as the current target image annotation model until a preset skipping ending condition is met, and outputting a corresponding classified and labeled image sample set.
9. An electronic device, comprising:
a memory for storing a computer program;
a processor for executing the computer program for implementing the steps of the image sample classification labeling method according to any one of claims 1 to 7.
10. A computer-readable storage medium for storing a computer program; wherein the computer program realizes the steps of the image sample classification labeling method according to any one of claims 1 to 7 when being executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210413215.0A CN114519404B (en) | 2022-04-20 | 2022-04-20 | Image sample classification labeling method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210413215.0A CN114519404B (en) | 2022-04-20 | 2022-04-20 | Image sample classification labeling method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114519404A true CN114519404A (en) | 2022-05-20 |
| CN114519404B CN114519404B (en) | 2022-07-12 |
Family
ID=81600077
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210413215.0A Active CN114519404B (en) | 2022-04-20 | 2022-04-20 | Image sample classification labeling method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114519404B (en) |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109784391A (en) * | 2019-01-04 | 2019-05-21 | 杭州比智科技有限公司 | Sample mask method and device based on multi-model |
| US20200184270A1 (en) * | 2018-10-30 | 2020-06-11 | Eyezon Ltd | Automatically tagging images to create labeled dataset for training supervised machine learning models |
| US20200265630A1 (en) * | 2017-06-27 | 2020-08-20 | The Boeing Company | Generative image synthesis for training deep learning machines |
| US20200356842A1 (en) * | 2019-05-09 | 2020-11-12 | Shenzhen Malong Technologies Co., Ltd. | Decoupling Category-Wise Independence and Relevance with Self-Attention for Multi-Label Image Classification |
| CN112182269A (en) * | 2020-09-27 | 2021-01-05 | 北京达佳互联信息技术有限公司 | Training of image classification model, image classification method, device, equipment and medium |
| CN112328823A (en) * | 2020-11-25 | 2021-02-05 | Oppo广东移动通信有限公司 | Training method and device for multi-label classification model, electronic equipment and storage medium |
| CN112348122A (en) * | 2020-12-03 | 2021-02-09 | 苏州挚途科技有限公司 | Method and device for marking drivable area and electronic equipment |
| CN112825144A (en) * | 2019-11-20 | 2021-05-21 | 深圳云天励飞技术有限公司 | Picture labeling method and device, electronic equipment and storage medium |
| US20210166822A1 (en) * | 2019-12-02 | 2021-06-03 | International Business Machines Corporation | Method and apparatus for selecting radiology reports for image labeling by modality and anatomical region of interest |
| CN113065013A (en) * | 2021-03-25 | 2021-07-02 | 携程计算机技术（上海）有限公司 | Image annotation model training and image annotation method, system, device and medium |
| CN113505261A (en) * | 2021-08-04 | 2021-10-15 | 城云科技（中国）有限公司 | Data annotation method and device and data annotation model training method and device |
| WO2021243246A1 (en) * | 2020-05-29 | 2021-12-02 | University Of Florida Research Foundation | A machine learning system and method for predicting alzheimer's disease based on retinal fundus images |
| CN113822263A (en) * | 2021-06-18 | 2021-12-21 | 腾讯科技（深圳）有限公司 | Image annotation method and device, computer equipment and storage medium |
| CN114169381A (en) * | 2020-08-19 | 2022-03-11 | 深圳云天励飞技术股份有限公司 | Image annotation method and device, terminal equipment and storage medium |
| CN114332553A (en) * | 2021-11-03 | 2022-04-12 | 腾讯科技（深圳）有限公司 | Image processing method, device, equipment and storage medium |
-
2022
- 2022-04-20 CN CN202210413215.0A patent/CN114519404B/en active Active
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200265630A1 (en) * | 2017-06-27 | 2020-08-20 | The Boeing Company | Generative image synthesis for training deep learning machines |
| US20200184270A1 (en) * | 2018-10-30 | 2020-06-11 | Eyezon Ltd | Automatically tagging images to create labeled dataset for training supervised machine learning models |
| CN109784391A (en) * | 2019-01-04 | 2019-05-21 | 杭州比智科技有限公司 | Sample mask method and device based on multi-model |
| US20200356842A1 (en) * | 2019-05-09 | 2020-11-12 | Shenzhen Malong Technologies Co., Ltd. | Decoupling Category-Wise Independence and Relevance with Self-Attention for Multi-Label Image Classification |
| CN112825144A (en) * | 2019-11-20 | 2021-05-21 | 深圳云天励飞技术有限公司 | Picture labeling method and device, electronic equipment and storage medium |
| US20210166822A1 (en) * | 2019-12-02 | 2021-06-03 | International Business Machines Corporation | Method and apparatus for selecting radiology reports for image labeling by modality and anatomical region of interest |
| WO2021243246A1 (en) * | 2020-05-29 | 2021-12-02 | University Of Florida Research Foundation | A machine learning system and method for predicting alzheimer's disease based on retinal fundus images |
| CN114169381A (en) * | 2020-08-19 | 2022-03-11 | 深圳云天励飞技术股份有限公司 | Image annotation method and device, terminal equipment and storage medium |
| CN112182269A (en) * | 2020-09-27 | 2021-01-05 | 北京达佳互联信息技术有限公司 | Training of image classification model, image classification method, device, equipment and medium |
| CN112328823A (en) * | 2020-11-25 | 2021-02-05 | Oppo广东移动通信有限公司 | Training method and device for multi-label classification model, electronic equipment and storage medium |
| CN112348122A (en) * | 2020-12-03 | 2021-02-09 | 苏州挚途科技有限公司 | Method and device for marking drivable area and electronic equipment |
| CN113065013A (en) * | 2021-03-25 | 2021-07-02 | 携程计算机技术（上海）有限公司 | Image annotation model training and image annotation method, system, device and medium |
| CN113822263A (en) * | 2021-06-18 | 2021-12-21 | 腾讯科技（深圳）有限公司 | Image annotation method and device, computer equipment and storage medium |
| CN113505261A (en) * | 2021-08-04 | 2021-10-15 | 城云科技（中国）有限公司 | Data annotation method and device and data annotation model training method and device |
| CN114332553A (en) * | 2021-11-03 | 2022-04-12 | 腾讯科技（深圳）有限公司 | Image processing method, device, equipment and storage medium |
Non-Patent Citations (2)
| Title |
|---|
| 常致富等: "基于深度学习的图像自动标注方法综述", 《山东大学学报(工学版)》 * |
| 张蕾等: "融合卷积神经网络与主题模型的图像标注", 《激光与光电子学进展》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114519404B (en) | 2022-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20230274125A1 (en) | Learning observation representations by predicting the future in latent space | |
| CN111602148B (en) | Regularized neural network architecture search | |
| US11205099B2 (en) | Training neural networks using data augmentation policies | |
| CN110807515A (en) | Model generation method and device | |
| CN109376267B (en) | Method and apparatus for generating a model | |
| US20220121906A1 (en) | Task-aware neural network architecture search | |
| WO2020182123A1 (en) | Method and device for pushing statement | |
| CN114067119B (en) | Training method of panorama segmentation model, panorama segmentation method and device | |
| US11501655B2 (en) | Automated skill tagging, knowledge graph, and customized assessment and exercise generation | |
| CN108717547A (en) | The method and device of sample data generation method and device, training pattern | |
| CN112819099A (en) | Network model training method, data processing method, device, medium and equipment | |
| CN114330588A (en) | Picture classification method, picture classification model training method and related device | |
| CN111143517B (en) | Human selection label prediction method, device, equipment and storage medium | |
| CN110909768A (en) | Method and device for acquiring marked data | |
| CN112712795B (en) | Labeling data determining method, labeling data determining device, labeling data determining medium and electronic equipment | |
| CN110574047A (en) | Generating output examples using bit blocks | |
| CN114519404B (en) | Image sample classification labeling method, device, equipment and storage medium | |
| CN114118068B (en) | Method and device for amplifying training text data and electronic equipment | |
| CN113688232B (en) | Method and device for classifying bid-inviting text, storage medium and terminal | |
| CN115688002A (en) | Classification method and device, method and device for training classification model and classification model | |
| US20220284280A1 (en) | Data labeling for synthetic data generation | |
| CN114861873A (en) | Multi-stage computationally efficient neural network inference | |
| CN114141236A (en) | Language model updating method and device, electronic equipment and storage medium | |
| CN117636100B (en) | Pre-training task model adjustment processing method and device, electronic equipment and medium | |
| CN110796262B (en) | Test data optimization method and device of machine learning model and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |