CN111783812A - Prohibited image recognition method, apparatus and computer-readable storage medium - Google Patents

Abstract

The present disclosure relates to a prohibited image recognition method, device and computer-readable storage medium, and relates to the field of computer technology. The method of the present disclosure includes: inputting an image to be recognized into a classification model to obtain image category information of the output image to be recognized; determining whether the image to be recognized belongs to a candidate prohibited image according to the image category information of the image to be recognized; In the case of an image, the image to be recognized is input into the target detection model to obtain target category information of each target in the image to be recognized; according to the target category information of each target in the image to be recognized, it is determined whether the image to be recognized is a prohibited image.

Description

Prohibited image recognition method, apparatus and computer-readable storage medium

technical field

The present disclosure relates to the field of computer technologies, and in particular, to a prohibited image recognition method, device, and computer-readable storage medium.

Background technique

With the development of Internet technology, the dissemination of information has become more and more widespread and faster. The Internet is flooded with all kinds of information. In order to maintain network order and purify the network environment, it is necessary to remove some illegal and bad information, such as prohibited images such as pornography and violence. Due to the complex content of prohibited images and many details, at present, prohibited images are generally reviewed manually.

SUMMARY OF THE INVENTION

The inventors found that manual review of prohibited images is inefficient.

A technical problem to be solved by the present disclosure is: how to improve the efficiency of prohibited image recognition.

According to some embodiments of the present disclosure, a method for recognizing prohibited images is provided, comprising: inputting an image to be recognized into a classification model to obtain image category information of the output image to be recognized; determining the image to be recognized according to the image category information of the image to be recognized Whether it belongs to a candidate prohibited image; if the image to be recognized belongs to a candidate prohibited image, input the image to be recognized into the target detection model to obtain the target category information of each target in the image to be recognized; according to the target category information of each target in the image to be recognized , to determine whether the image to be recognized is a prohibited image.

In some embodiments, determining whether the to-be-recognized image belongs to the candidate prohibited images according to the image category information of the to-be-recognized image includes: determining an application scenario of the to-be-recognized image; According to the matching result, it is determined whether the image to be identified belongs to the candidate prohibited images; wherein, the preset image categories to which the candidate prohibited images belong in different application scenarios are different.

In some embodiments, the image category information of the image to be identified includes: the probability that the image to be identified belongs to each image category; the matching of the preset image category to which the candidate prohibited image belongs in the application scenario and the image category information of the to-be-identified image includes: Compare the probability that the image to be recognized belongs to each image category and the probability of the preset image category corresponding to the application scene to determine the image category of the image to be identified; Image categories are matched; wherein, the preset image category probabilities corresponding to different application scenarios are different.

In some embodiments, determining whether the to-be-recognized image is a prohibited image according to target category information of each target in the to-be-recognized image includes: determining an application scenario of the to-be-recognized image; The target category information of each target is matched; according to the matching result, it is determined whether the image to be recognized belongs to a prohibited image; wherein, the preset prohibited target categories in different application scenarios are different.

In some embodiments, the target category information of each target in the image to be recognized includes: the probability that each target belongs to each target category; the preset prohibited target category in the application scenario is matched with the target category information of each target in the to-be-recognized image. Including: comparing the probability of each target in the image to be identified belonging to each target category with the preset target category probability corresponding to the application scenario, and determining the target category of each target in the image to be identified; The category is matched with the target category of each target in the image to be recognized; wherein, the preset target category probabilities corresponding to different application scenarios are different.

In some embodiments, the method further includes: acquiring a first sample image marked with an image category as a first training sample set; using the images in the first training sample set to train a classification model to obtain parameters of the classification model.

In some embodiments, using the images of the first training sample set to train the classification model includes: using the images of the first training sample set to initially train the classification model; inputting the images of the first training sample set into the classification model after the initial training model to obtain the classification results of the images of the first training sample set; determine the difficult sample images according to the difference between the output classification results of the images of the first training sample set and the accurate classification results; The classification model is trained again.

In some embodiments, the method further includes: acquiring a second sample image marked with a target category of the target as a second training sample set; and using the images of the second training sample set to train the target detection model to obtain target detection parameters.

In some embodiments, acquiring the second sample image marked with the target category of the target includes: inputting the candidate sample image into the classification model to obtain image category information of the output candidate sample image; determining the candidate sample according to the image category information of the candidate sample image Whether it belongs to the candidate prohibited images; the candidate sample images belonging to the candidate prohibited images are regarded as the second sample images.

According to other embodiments of the present disclosure, a prohibited image recognition device is provided, comprising: an image category determination module, configured to input an image to be identified into a classification model to obtain image category information of the output image to be identified; an image screening module, It is used to determine whether the image to be recognized belongs to the candidate prohibited images according to the image category information of the image to be recognized; the target category determination module is used to input the image to be recognized into the target detection model when the image to be recognized belongs to the candidate prohibited image, and obtain the target detection model. Recognize target category information of each target in the image; a prohibited image determination module is used to determine whether the to-be-recognized image is a prohibited image according to the target category information of each target in the to-be-recognized image.

In some embodiments, the image screening module is used to determine the application scene of the image to be recognized; match the preset image category to which the candidate prohibited images in the application scene belong with the image category information of the image to be recognized; determine the image to be recognized according to the matching result Whether it belongs to the candidate prohibited images; among them, the preset image categories to which the candidate prohibited images belong in different application scenarios are different.

In some embodiments, the image category information of the image to be identified includes: the probability that the image to be identified belongs to each image category; the image screening module is configured to compare the probability of the image to be identified belonging to each image category with the preset image category probability corresponding to the application scene Perform comparison to determine the image category of the image to be identified; match the preset image category to which the candidate prohibited images in the application scenario belong with the image category of the image to be identified; wherein the preset image category probabilities corresponding to different application scenarios are different.

In some embodiments, the prohibited image determination module is used to determine the application scene of the image to be recognized; match the preset prohibited target category under the application scene with the target category information of each target in the to-be-recognized image; determine the to-be-recognized image according to the matching result Whether the image is a prohibited image; among them, the preset prohibited target categories in different application scenarios are different.

In some embodiments, the target category information of each target in the to-be-recognized image includes: the probability that each target belongs to each target category; the prohibited image determination module is used to compare the probability of each target in the to-be-recognized image belonging to each target category with the application scenario The corresponding preset target category probabilities are compared to determine the target category of each target in the image to be recognized; the preset prohibited target category in the application scenario is matched with the target category of each target in the to-be-recognized image; among them, different application scenarios The corresponding preset target category probabilities are different.

In some embodiments, the apparatus further includes: a first training module, configured to obtain a first sample image marked with an image category as a first training sample set; and use the images of the first training sample set to train the classification model , get the parameters of the classification model.

In some embodiments, the first training module is used to perform initial training on the classification model using the images of the first training sample set; input the images of the first training sample set into the classification model after the initial training, to obtain the first training sample set of images The classification result of the image; according to the difference between the classification result of the output first training sample set and the accurate classification result, the difficult sample image is determined; using the difficult sample image, the classification model after the initial training is retrained.

In some embodiments, the apparatus further includes: a second training module, configured to obtain a second sample image of a target category marked with a target, as a second training sample set; a target detection model using the images of the second training sample set Perform training to obtain parameters for target detection.

In some embodiments, the second training module is used to input the candidate sample image into the classification model, and obtain the image category information of the output candidate sample image; determine whether the candidate sample belongs to the candidate prohibited image according to the image category information of the candidate sample image; The candidate sample images of the candidate prohibited images are used as the second sample images.

According to further embodiments of the present disclosure, there is provided a prohibited image recognition device, comprising: a processor; and a memory coupled to the processor for storing instructions, and when the instructions are executed by the processor, the processor executes the above-mentioned The prohibited image identification method of any embodiment.

According to still other embodiments of the present disclosure, a non-transitory computer-readable storage medium is provided, and a computer program is stored thereon, and when the program is executed by a processor, the prohibited image recognition method of any of the foregoing embodiments is implemented.

The present disclosure proposes a method for automatically recognizing prohibited images by a machine. First, an image to be recognized is input into a classification model to obtain a coarsely classified image category of the image to be recognized, and candidate prohibited images are screened out according to the coarse classification category, and a target detection model is further used to treat the image. The target in the image is recognized for identification, the target category information of each target is obtained, and whether the to-be-recognized image is a prohibited image is determined according to the target category information of each target. The method of the present disclosure combines the classification model and the target detection model to apply the to-be-recognized image firstly to perform rough classification to recognize the features of the whole image, and then to perform sub-classification to finely recognize the details of the target in the image. Orientation recognition of prohibited images improves the accuracy and efficiency of recognition.

Other features of the present disclosure and advantages thereof will become apparent from the following detailed description of exemplary embodiments of the present disclosure with reference to the accompanying drawings.

Description of drawings

In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only These are some embodiments of the present disclosure, and for those of ordinary skill in the art, other drawings can also be obtained from these drawings without creative effort.

FIG. 1 shows a schematic flowchart of a prohibited image recognition method according to some embodiments of the present disclosure.

FIG. 2 shows a schematic flowchart of a prohibited image recognition method according to other embodiments of the present disclosure.

FIG. 3 shows a schematic structural diagram of a prohibited image recognition apparatus according to some embodiments of the present disclosure.

FIG. 4 shows a schematic structural diagram of a prohibited image recognition apparatus according to other embodiments of the present disclosure.

FIG. 5 shows a schematic structural diagram of a prohibited image recognition apparatus according to further embodiments of the present disclosure.

Detailed ways

The technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, but not all of the embodiments. The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the disclosure, its application or uses in any way. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present disclosure.

Aiming at the problem of low efficiency in the prior art mainly using manual review of prohibited images, a method for automatically recognizing a scarf image by a machine is proposed. Some embodiments of the solution are described below with reference to FIG. 1 .

FIG. 1 is a flowchart of some embodiments of a prohibited image recognition method of the present disclosure. As shown in FIG. 1 , the method of this embodiment includes steps S102 to S108.

In step S102, the to-be-recognized image is input into the classification model to obtain image category information of the output to-be-recognized image.

The to-be-recognized image is obtained from the network, and the to-be-recognized image may be an image frame extracted from the to-be-recognized video. The to-be-recognized image can be preprocessed, for example, the to-be-recognized image is rotated, scaled, color adjusted, etc., to make subsequent recognition of the to-be-recognized image more accurate. Input the preprocessed image to be recognized into a classification model, extract features of the image to be recognized (eg, CNN (Convolutional Neural Network) features), and determine image category information of the image to be recognized according to the features. The classification model is, for example, a neural network model, a convolutional neural network model, or more specifically, an existing model such as ResNeXt, ResNeXt50 or SE-ResNeXt50, and a model with better and more accurate classification effect can be selected according to actual needs.

The classification model can configure the divided image categories according to actual requirements. After pre-training, the classification model can determine that the image to be recognized belongs to one or more of the configured image categories. For example, the image classes determined by the classification model include two image classes, normal images and prohibited images. For another example, the image categories determined by the classification model include three image categories, normal images, intermediate images, and prohibited images. The intermediate image belongs to the image category between normal and prohibited. For example, for pornographic prohibited image recognition, the intermediate image can represent a sexy image. . For another example, the image categories determined by the classification model include four image categories: normal images, partial normal images, partial prohibited images, and prohibited images. For example, for pornographic prohibited image recognition, partial normal images may represent sexy images, partial prohibited images may represent vulgar image. The image category determined by the classification model is determined according to the labeling of the training samples and the training process, and can be determined according to actual needs. The training process will be described later.

The classification model can determine the probability that the image to be recognized belongs to each image category, and further determine one or more image categories of the image to be recognized according to the probability that the image to be recognized belongs to each image category.

In step S104, it is determined whether the to-be-recognized image belongs to candidate prohibited images according to the image category information of the to-be-recognized image.

Candidate prohibited images represent images that are likely to be prohibited images, and images that need to be sub-classified by the subsequent object detection model. In some embodiments, the application scene of the image to be recognized is determined; the preset image category to which the candidate prohibited image belongs in the application scenario is matched with the image category information of the image to be recognized; according to the matching result, it is determined whether the image to be recognized belongs to the candidate prohibited image ; wherein, the preset image categories to which the candidate prohibited images belong in different application scenarios are different.

Different application scenarios have different discriminative scales for prohibited images, so the discriminative scales of candidate prohibited images are also different. For example, in more serious news content, sexy images may belong to prohibited images, therefore, images to be identified that belong to the category of sexy images can be determined as candidate prohibited images, while for entertainment content, sexy images are not prohibited images, Therefore, images to be identified that belong to the category of sexy images may not be determined as candidate prohibited images. The preset image category to which the candidate prohibited images corresponding to different application scenarios belong can be preset, the preset image category corresponding to the application scene of the image to be identified can be searched, and then the image category information of the image to be identified is compared with the corresponding preset image category. right.

Further, in some embodiments, the image category information of the image to be identified includes: the probability that the image to be identified belongs to each image category. The probability of the image to be recognized belonging to each image category is compared with the preset probability of the image category corresponding to the application scene, and the image category of the image to be recognized is determined. Match the preset image category to which the candidate prohibited images in the application scenario belong with the image category of the image to be recognized. If the image category of the image to be identified includes one or more of preset image categories corresponding to the application scene to which it belongs, the image to be identified belongs to a candidate prohibited image. By comparing the probability that the image to be recognized belongs to each image category and the preset probability of the image category corresponding to the application scenario, the image to be recognized may be determined into multiple categories at the same time. That is, for the probability that the image to be recognized belongs to each image category, if the probability of belonging to the image category exceeds the preset probability of the image category, the image to be identified belongs to the image category.

The preset image category probabilities corresponding to different application scenarios are different. Since the discriminant scales for prohibited images are different in different application scenarios, by adjusting the preset image category probability in different application scenarios, the image category of the image to be recognized can be adjusted, thereby adjusting the selection of candidate prohibited images, and the model can be adapted to different application scenarios.

Further, different image categories in the same application scenario may correspond to different preset image category probabilities. For the probability that the image to be recognized belongs to each image category, the probability of belonging to the image category is compared with the preset image category probability of the image category pair in the application scenario to which it belongs, so as to determine whether the image to be identified belongs to the image category.

In step S106, if the image to be recognized belongs to a candidate prohibited image, the image to be recognized is input into the target detection model to obtain target category information of each target in the image to be recognized.

Select the to-be-recognized images that belong to the candidate prohibited images and input the target detection for sub-classification, and further accurately identify whether they belong to the prohibited images. The target detection model is, for example, an existing model, such as Faster-RCNN (Faster Recurrent Convolutional Neural Network), etc., and a model with better effect can be selected according to actual needs.

In step S108, it is determined whether the to-be-recognized image is a prohibited image according to the target category information of each target in the to-be-recognized image.

In some embodiments, the application scene of the image to be recognized is determined; the preset prohibited target category in the application scene is matched with the target category information of each target in the image to be recognized; whether the image to be recognized belongs to a prohibited image is determined according to the matching result. The preset prohibited target categories are different in different application scenarios. Preset prohibited target categories corresponding to different application scenarios can be preset. For example, swimwear belongs to the prohibited target category in some application scenarios (news scenarios), but does not belong to the prohibited target category in some application scenarios (shopping platform scenarios). By adjusting the preset prohibited target categories in different application scenarios, prohibited images that conform to different application scenarios can be determined, and prohibited images can be flexibly determined according to the application scenarios.

Further, in some embodiments, the target category information of each target in the image to be identified includes: the probability that each target belongs to each target category. The target detection model can determine the probability that each target in the image to be recognized belongs to each target category, and compare the probability of each target in the image to be recognized belonging to each target category with the preset target category probability corresponding to the application scene, and determine the image to be recognized. The target category of each target in the image; match the preset prohibited target category in the application scenario with the target category of each target in the image to be recognized. If the target in the to-be-recognized image includes one or more of the preset prohibited target categories corresponding to the application scene to which it belongs, the to-be-recognized image belongs to a prohibited image. For example, if the image to be recognized includes some exposed human body parts or includes some specific actions, it can be determined as a prohibited image.

By comparing the probability of each target in the image to be recognized belonging to each target category with the preset target category probability corresponding to the application scenario, the target may be determined into multiple categories at the same time. That is, for the probability of each target belonging to each target category, if the probability of the target belonging to the target category exceeds the preset target category probability, the target belongs to the target category.

The preset target category probabilities corresponding to different application scenarios are different. Since the discriminant scale for prohibited images is different in different application scenarios, by adjusting the preset target category probability in different application scenarios, the target category of each target in the image to be recognized can be adjusted, thereby adjusting the selection of prohibited images, which can adapt the model to different applications. Scenes.

Further, different target categories in the same application scenario may correspond to different preset target category probabilities. For the probability that each target in the image to be identified belongs to each image category, the probability of the target belonging to the target category is compared with the preset target category probability of the target category pair in the application scenario to which it belongs, so as to determine whether the target belongs to the target category. falls into this target category.

The above embodiment proposes a method for automatically recognizing prohibited images by a machine. First, the images to be recognized are input into a classification model to obtain a coarsely classified image category of the to-be-recognized image, and candidate prohibited images are screened out according to the coarse classification category, and the target detection model is further utilized. The target in the to-be-recognized image is identified, the target category information of each target is obtained, and whether the to-be-recognized image is a prohibited image is determined according to the target category information of each target. The method of the present disclosure combines the classification model and the target detection model to apply the to-be-recognized image firstly to perform rough classification to recognize the features of the whole image, and then to perform sub-classification to finely recognize the details of the target in the image. Orientation recognition of prohibited images improves the accuracy and efficiency of recognition. In addition, by flexibly configuring preset image categories, preset prohibited target categories, preset image category probabilities, preset target category probabilities, etc. in different application scenarios, the determination of candidate prohibited images and prohibited images can be adjusted, and the above method can be used. It is suitable for image recognition in different application scenarios, making the recognition of prohibited images more accurate and flexible.

The following describes the training process of the classification model and the target detection model in the present disclosure with reference to FIG. 2 .

FIG. 2 is a flowchart of other embodiments of the prohibited image recognition method of the present disclosure. As shown in FIG. 2, the method of this embodiment includes steps S202-S208.

In step S202, a first sample image marked with an image category is acquired as a first training sample set.

As in the foregoing embodiment, image categories that can be classified by the classification model can be set according to actual requirements, and each first sample image is marked with a corresponding image category. In order to improve the accuracy of the classification model, first sample images from various scenes, various media media, and various sources can be obtained. The first sample image may be preprocessed, such as rotation, scaling, color adjustment, etc., to make the first sample image form a uniform specification.

In step S204, the classification model is trained by using the images of the first training sample set to obtain the parameters of the classification model.

In some embodiments, the images in the first training sample set are input into the classification model, the image categories of the output images are obtained, the first loss function value is calculated according to the difference between the image categories of the output images and the marked image categories, and the first loss function value is calculated according to the A loss function adjusts the parameters of the classification model, and the above process is repeated until a preset condition is reached, for example, the value of the first loss function reaches a minimum value or a threshold value.

During each round of training, a preset number of image input classification models in the first training sample set are extracted, and image input classification models belonging to each image category may be extracted according to a first preset ratio. For example, images of four image categories are extracted according to 1:1:1:1.

In some embodiments, the classification model is initially trained by using the images of the first training sample set; the images of the first training sample set are input into the classification model after the initial training to obtain the classification results of the images of the first training sample set; according to The difference between the classification result and the accurate classification result of the images of the output first training sample set is used to determine the difficult sample image; the classification model after the initial training is retrained by using the difficult sample image. Retraining the classification model with difficult sample images can enhance the accuracy and training efficiency of the classification model.

In step S206, a second sample image marked with the target category of the target is acquired as a second training sample set.

In some embodiments, the candidate sample image is input into the classification model to obtain image category information of the output candidate sample image; whether the candidate sample belongs to the candidate prohibited image is determined according to the image category information of the candidate sample image; the candidate sample belonging to the candidate prohibited image is image as a second sample image. After the classification model is trained, some candidate sample images are screened out by using the classification model as second sample images. The candidate sample images may be images in the first training sample set. This can remove a lot of normal pictures, so it will greatly reduce possible confusion and improve training efficiency and accuracy.

Because the content of the prohibited pictures is relatively complex and the features are relatively scattered, there will be more target categories marked and those determined during application in the foregoing embodiment. The target category can be set to include multi-level categories, for example, human body, action, etc. are high-level categories, gender is a middle-level category, and specific parts or action descriptions are low-level categories. Such a target determines the target category through multi-level labels, which is convenient for labeling. Manage and further determine prohibited images based on target categories.

In step S208, the target detection model is trained by using the images of the second training sample set to obtain parameters for target detection.

In some embodiments, the images in the second training sample set are input into the target detection model to obtain the target category of each output target, and the second loss function value is calculated according to the difference between the target category of each output target and the marked target category, and according to The second loss function adjusts the parameters of the target detection model, and the above process is repeated until a preset condition is reached, for example, the value of the second loss function reaches a minimum value or a threshold value. During each round of training, a preset number of image input target detection models in the second training sample set are extracted, and image input target detection models belonging to each target category may be extracted according to a second preset ratio.

The present disclosure also provides a prohibited image recognition device, which will be described below with reference to FIG. 3 .

FIG. 3 is a structural diagram of some embodiments of a prohibited image recognition apparatus of the present disclosure. As shown in FIG. 3 , the apparatus 30 of this embodiment includes: an image category determination module 310 , an image screening module 320 , a target category determination module 330 , and a prohibited image determination module 340 .

The image category determination module 310 is configured to input the image to be recognized into the classification model, and obtain image category information of the output image to be recognized.

The image screening module 320 is configured to determine whether the to-be-recognized image belongs to candidate prohibited images according to the image category information of the to-be-recognized image.

In some embodiments, the image screening module 320 is configured to determine the application scene of the image to be recognized; match the preset image category to which the candidate prohibited images in the application scene belong with the image category information of the image to be recognized; determine the to-be-recognized image according to the matching result Whether the image belongs to the candidate prohibited images; among them, the preset image categories to which the candidate prohibited images belong in different application scenarios are different.

In some embodiments, the image category information of the image to be identified includes: the probability that the image to be identified belongs to each image category; the image screening module 320 is configured to associate the probability of the image to be identified belonging to each image category with the preset image category corresponding to the application scene The probability is compared to determine the image category of the image to be identified; the preset image category to which the candidate prohibited images in the application scenario belong is matched with the image category of the image to be identified; wherein, the preset image category probabilities corresponding to different application scenarios are different.

The target category determination module 330 is configured to input the to-be-recognized image into the target detection model when the to-be-recognized image belongs to candidate prohibited images to obtain target class information of each target in the to-be-recognized image.

The prohibited image determination module 340 is configured to determine whether the image to be recognized is a prohibited image according to the target category information of each target in the image to be recognized.

In some embodiments, the prohibited image determination module 340 is configured to determine the application scene of the image to be recognized; match the preset prohibited target category in the application scene with the target category information of each target in the image to be recognized; determine the target category according to the matching result Identify whether an image is a prohibited image; among them, the preset prohibited target categories in different application scenarios are different.

In some embodiments, the target category information of each target in the to-be-recognized image includes: the probability that each target belongs to each target category; the prohibited image determination module 340 is configured to compare the probability of each target in the to-be-recognized image belonging to each target category with the application The preset target category probabilities corresponding to the scene are compared to determine the target category of each target in the image to be recognized; the preset prohibited target category in the application scenario is matched with the target category of each target in the to-be-recognized image; among them, different applications The preset target category probabilities corresponding to the scene are different.

In some embodiments, the apparatus 30 further includes: a first training module 350, configured to obtain a first sample image marked with an image category as a first training sample set; a classification model using the images of the first training sample set Perform training to get the parameters of the classification model.

In some embodiments, the first training module 350 is configured to perform initial training on the classification model using the images of the first training sample set; input the images of the first training sample set into the classification model after the initial training to obtain the first training sample set According to the difference between the classification results of the images of the first training sample set and the accurate classification results, the difficult sample images are determined; using the difficult sample images, the classification model after the initial training is retrained.

In some embodiments, the apparatus 30 further includes: a second training module 360, configured to acquire a second sample image marked with a target category of the target as a second training sample set; use the images of the second training sample set to pair the target The detection model is trained to obtain the parameters of target detection.

In some embodiments, the second training module 360 is configured to input the candidate sample images into the classification model to obtain image category information of the output candidate sample images; determine whether the candidate samples belong to candidate prohibited images according to the image category information of the candidate sample images; The candidate sample images belonging to the candidate prohibited images are used as the second sample images.

The prohibited image recognition apparatuses in the embodiments of the present disclosure may be implemented by various computing devices or computer systems, which will be described below with reference to FIG. 4 and FIG. 5 .

FIG. 4 is a structural diagram of some embodiments of a prohibited image recognition apparatus of the present disclosure. As shown in FIG. 4 , the apparatus 40 of this embodiment includes a memory 410 and a processor 420 coupled to the memory 410 , the processor 420 is configured to execute any of the implementations of the present disclosure based on instructions stored in the memory 410 Prohibited image recognition method in the example.

The memory 410 may include, for example, a system memory, a fixed non-volatile storage medium, and the like. The system memory stores, for example, an operating system, an application program, a boot loader (Boot Loader), a database, and other programs.

FIG. 5 is a structural diagram of another embodiment of the prohibited image recognition device of the present disclosure. As shown in FIG. 5 , the apparatus 50 in this embodiment includes: a memory 510 and a processor 520 , which are similar to the memory 410 and the processor 420 , respectively. It may also include an input-output interface 530, a network interface 540, a storage interface 550, and the like. These interfaces 530 , 540 , 550 and the memory 510 and the processor 520 can be connected, for example, through a bus 560 . The input and output interface 530 provides a connection interface for input and output devices such as a display, a mouse, a keyboard, and a touch screen. The network interface 540 provides a connection interface for various networked devices, for example, it can be connected to a database server or a cloud storage server. The storage interface 550 provides a connection interface for external storage devices such as SD cards and U disks.

As will be appreciated by one skilled in the art, embodiments of the present disclosure may be provided as a method, system, or computer program product. Accordingly, the present disclosure may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present disclosure may take the form of a computer program product embodied on one or more computer-usable non-transitory storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein .

The present disclosure is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the disclosure. It will be understood that each process and/or block in the flowchart illustrations and/or block diagrams, and combinations of processes and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing device produce Means for implementing the functions specified in a flow or flow of a flowchart and/or a block or blocks of a block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions The apparatus implements the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.

These computer program instructions can also be loaded on a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process such that The instructions provide steps for implementing the functions specified in the flow or blocks of the flowcharts and/or the block or blocks of the block diagrams.

The above descriptions are only preferred embodiments of the present disclosure, and are not intended to limit the present disclosure. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present disclosure shall be included in the protection of the present disclosure. within the range.

Claims (12)

1. A prohibited image recognition method, comprising: Input the image to be recognized into the classification model, and obtain the image category information of the output image to be recognized; Determine whether the to-be-recognized image belongs to a candidate prohibited image according to the image category information of the to-be-recognized image; In the case that the to-be-recognized image belongs to a candidate prohibited image, the to-be-recognized image is input into a target detection model to obtain target category information of each target in the to-be-recognized image; According to the target category information of each target in the to-be-recognized image, it is determined whether the to-be-recognized image is a prohibited image. 2. The prohibited image recognition method according to claim 1, wherein, The determining whether the to-be-recognized image belongs to a candidate prohibited image according to the image category information of the to-be-recognized image includes: determining the application scenario of the to-be-recognized image; Matching the preset image category to which the candidate prohibited image belongs in the application scenario with the image category information of the to-be-recognized image; Determine whether the to-be-identified image belongs to a candidate prohibited image according to the matching result; The preset image categories to which the candidate prohibited images belong in different application scenarios are different. 3. The prohibited image recognition method according to claim 2, wherein, The image category information of the to-be-recognized image includes: the probability that the to-be-recognized image belongs to each image category; The matching of the preset image category to which the candidate prohibited images in the application scenario belongs and the image category information of the to-be-recognized image includes: Comparing the probability that the image to be recognized belongs to each image category and the probability of the preset image category corresponding to the application scenario, to determine the image category of the image to be recognized; Matching the preset image category to which the candidate prohibited image belongs in the application scenario with the image category of the to-be-recognized image; The preset image category probabilities corresponding to different application scenarios are different. 4. The prohibited image recognition method according to claim 1, wherein, The determining whether the to-be-recognized image is a prohibited image according to the target category information of each target in the to-be-recognized image includes: determining the application scenario of the to-be-recognized image; Matching the preset prohibited target category under the application scenario with the target category information of each target in the to-be-recognized image; Determine whether the to-be-identified image is a prohibited image according to the matching result; Among them, the preset prohibited target categories in different application scenarios are different. 5. The prohibited image recognition method according to claim 4, wherein, The target category information of each target in the to-be-recognized image includes: the probability that each target belongs to each target category; The matching of the preset prohibited target category in the application scenario with the target category information of each target in the to-be-recognized image includes: Compare the probability of each target in the to-be-recognized image belonging to each target class with the preset target class probability corresponding to the application scenario, and determine the target class of each target in the to-be-recognized image; Matching the preset prohibited target category under the application scenario with the target category of each target in the to-be-recognized image; The preset target category probabilities corresponding to different application scenarios are different. 6. The prohibited image recognition method according to claim 1, further comprising: Obtain the first sample image marked with the image category as the first training sample set; The classification model is trained by using the images of the first training sample set to obtain the parameters of the classification model. 7. The prohibited image recognition method according to claim 5, wherein, The training of the classification model using the images of the first training sample set includes: Perform initial training on the classification model using the images of the first training sample set; Inputting the image of the first training sample set into the classification model after initial training, to obtain the classification result of the image of the first training sample set; Determine the difficult sample image according to the difference between the output classification result and the accurate classification result of the images of the first training sample set; Using the difficult sample images, the classification model after the initial training is retrained. 8. The prohibited image recognition method according to claim 1, further comprising: Obtain a second sample image marked with the target category of the target as a second training sample set; The target detection model is trained by using the images of the second training sample set to obtain the parameters of the target detection. 9. The prohibited image recognition method according to claim 8, wherein, The acquiring the second sample image of the target category marked with the target includes: Input the candidate sample image into the classification model to obtain image category information of the output candidate sample image; Determine whether the candidate sample belongs to a candidate prohibited image according to the image category information of the candidate sample image; The candidate sample images belonging to the candidate prohibited images are used as the second sample images. 10. A prohibited image recognition device, comprising: an image category determination module, configured to input the image to be recognized into a classification model, and obtain image category information of the output image to be recognized; an image screening module, configured to determine whether the to-be-recognized image belongs to a candidate prohibited image according to the image category information of the to-be-recognized image; a target category determination module, configured to input the to-be-recognized image into a target detection model when the to-be-recognized image belongs to a candidate prohibited image to obtain target class information of each target in the to-be-recognized image; A prohibited image determination module, configured to determine whether the to-be-recognized image is a prohibited image according to target category information of each target in the to-be-recognized image. 11. A prohibited image recognition device, comprising: processor; and A memory coupled to the processor for storing instructions, when executed by the processor, causes the processor to execute the prohibited image recognition method according to any one of claims 1-9. 12. A non-transitory computer-readable storage medium having stored thereon a computer program, wherein the program, when executed by a processor, implements the steps of the method of any one of claims 1-9.

Images