Image recognition method and device, computer equipment, storage medium and program product

Abstract

The invention relates to an image recognition method and device, computer equipment, a storage medium and a program product, and relates to the technical field of image recognition. The method comprises the following steps: acquiring a first comment image, wherein the first comment image is an image contained in a newly generated comment; obtaining a first repetition degree of the first comment image, wherein the first repetition degree is used for indicating a repetition condition between the first comment image and each historical comment image; obtaining a second repetition degree of the first comment image, wherein the second repetition degree is used for indicating a repetition condition between the first comment image and each low-quality comment image; and under the condition that the first repetition degree and the second repetition degree meet a first condition, identifying the first comment image as a new low-quality image. According to the scheme, the accuracy of identifying the low-quality comment image can be improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of image recognition technology, and in particular to an image recognition method, apparatus, computer equipment, storage medium, and program product. Background Art

[0002] With the rapid development of Internet technology, the application of computer technology is becoming more and more extensive, especially in the field of image recognition, where low-quality duplicate images can be identified through computer technology.

[0003] In the related technology, a computer device can obtain multiple low-quality duplicate images, save the multiple low-quality duplicate images in a low-quality duplicate image library, and then calculate the similarity between the image to be detected and the images in the low-quality duplicate image library. When the similarity between the image to be detected and the images in the low-quality duplicate image library reaches a specified threshold, it is determined that the image to be detected is a low-quality duplicate image.

[0004] However, in the solutions shown in the above-mentioned related technologies, the recognition method of low-quality repeated images is too simple and prone to misjudgment, resulting in a low recognition accuracy rate of low-quality repeated images. Summary of the invention

[0005] The embodiments of the present application provide an image recognition method, apparatus, computer device, storage medium and program product, which can improve the accuracy of image recognition. The technical solution is as follows:

[0006] In one aspect, an image recognition method is provided, the method comprising:

[0007] Acquire a first comment image, where the first comment image is an image included in a newly generated comment;

[0008] Acquire a first repetition degree of the first comment image, where the first repetition degree is used to indicate a repetition situation between the first comment image and each historical comment image;

[0009] Acquire a second repetition degree of the first comment image, where the second repetition degree is used to indicate a repetition situation between the first comment image and each low-quality comment image;

[0010] When the first repetition degree and the second repetition degree satisfy a first condition, the first comment image is identified as a new low-quality image.

[0011] In another aspect, an image recognition device is provided, the device comprising:

[0012] A first comment image acquisition module, used to acquire a first comment image, where the first comment image is an image included in a newly generated comment;

[0013] A first repetition degree acquisition module, used for acquiring a first repetition degree of the first comment image, where the first repetition degree is used for indicating a repetition between the first comment image and each historical comment image;

[0014] A second repetition degree acquisition module, used for acquiring a second repetition degree of the first comment image, where the second repetition degree is used for indicating the repetition between the first comment image and each low-quality comment image;

[0015] The low-quality image recognition module is configured to recognize the first comment image as a new low-quality image when the first repetition degree and the second repetition degree satisfy a first condition.

[0016] In some embodiments, the first repetition degree is used to indicate the number of the historical comment images whose similarity with the first comment image is greater than a first similarity threshold among the historical comment images; the second repetition degree is used to indicate the number of the low-quality comment images whose similarity with the first comment image is greater than a second similarity threshold among the low-quality comment images.

[0017] In some embodiments, the first repetition acquisition module is used to acquire feature similarities between image features of each of the historical comment images and image features of the first comment image, as the similarities between each of the historical comment images and the first comment image;

[0018] The number of similarities between each of the historical comment images and the first comment image that is greater than the first similarity threshold is obtained as the first repetition degree.

[0019] In some embodiments, the second repetition acquisition module is used to acquire feature similarities between image features of each of the low-quality comment images and image features of the first comment image, as the similarities between each of the low-quality comment images and the first comment image;

[0020] The number of similarities between each of the low-quality comment images and the first comment image that is greater than the second similarity threshold is obtained as the second repetition degree.

[0021] In some embodiments, the first similarity threshold is positively correlated with the number of each of the historical comment images; and the second similarity threshold is positively correlated with the number of each of the low-quality comment images.

[0022] In some embodiments, the first condition includes: the first repetition degree is greater than a first repetition degree threshold, and the second repetition degree is greater than a second repetition degree threshold.

[0023] In some embodiments, the apparatus further comprises:

[0024] A second comment image acquisition module is used to acquire a second comment image before acquiring the first comment image, where the second comment image is an image included in a comment generated within a first time period before a current time point;

[0025] A first similarity acquisition submodule is used to acquire the similarity between the second comment image and each of the historical comment images before acquiring the first comment image;

[0026] The new historical comment image acquisition module is used to acquire the second comment image as a new historical comment image before acquiring the first comment image if the similarity between the second comment image and each of the historical comment images is not greater than a third similarity threshold.

[0027] In some embodiments, the apparatus further comprises:

[0028] A feedback information acquisition module, used for acquiring feedback information of a third comment image before acquiring the first comment image, the feedback information including the number of negative feedbacks, the third comment image being an image included in a comment generated in a second time period before the current time point;

[0029] A second similarity acquisition submodule is used to acquire the similarity between the third comment image and each of the low-quality comment images before acquiring the first comment image, if the feedback information of the third comment image meets the second condition;

[0030] A new low-quality comment image acquisition module is used to acquire the third comment image as a new low-quality comment image before acquiring the first comment image if the similarity between the third comment image and each of the low-quality comment images is not greater than a fourth similarity threshold.

[0031] In some embodiments, the second condition includes: the amount of the negative feedback is greater than a first amount threshold.

[0032] In some embodiments, the feedback information further includes the number of positive feedbacks; and the second condition includes:

[0033] The number of negative feedbacks is greater than a first number threshold, and the number of positive feedbacks is less than a second number threshold; or, the number of negative feedbacks is greater than the first number threshold, and a ratio between the number of negative feedbacks and the number of positive feedbacks is greater than a ratio threshold.

[0034] In some embodiments, the negative feedback includes at least one of the following: reporting that the image has quality problems, reporting that the image content is illegal, and marking the image as spam.

[0035] On the other hand, a computer device is provided, which includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the image recognition method as described above.

[0036] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction, at least one program, a code set or an instruction set is stored in the storage medium, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by a processor to implement the image recognition method as described above.

[0037] In another aspect, a computer program product is provided, the computer program product comprising a computer program, the computer program being stored in a computer-readable storage medium. A processor of a computer device reads the computer program from the computer-readable storage medium, and the processor executes the computer program, so that the computer device executes the image recognition method provided in the above-mentioned various optional implementations.

[0038] The technical solution provided by this application may have the following beneficial effects:

[0039] After obtaining the image in the newly generated comment (the first comment image), the computer device promptly obtains the first repetition and the second repetition of the first comment image. The first repetition can be used to obtain the repetition between the first comment image and each historical comment image, that is, the repetition degree between the image in the newly generated comment and each historical comment image. If the repetition degree is high, it means that the image in the newly generated comment has the possibility of being a repeated image, that is, the image in the newly generated comment may have appeared in the historical comment; the second repetition can be used to obtain the repetition between the first comment image and each low-quality comment image, that is, the repetition degree between the image in the newly generated comment and each low-quality comment image. If the repetition degree is high, it means that the image in the newly generated comment has the possibility of being a low-quality comment image, that is, the local / all features in the image in the newly generated comment may be the same as the local / all features of the low-quality comment image. When the first repetition and the second repetition meet the first condition, that is, both repetitions meet the specified condition, it means that the first comment image is not only similar to the historical comment image, but also similar to the known low-quality comment image. By combining the two factors to identify the image, the situation of misjudging the high-quality comment image as a low-quality comment image can be reduced, and the accuracy of image recognition can be effectively improved.

[0040] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0042] Figure 1 is a system configuration diagram of an image recognition method involved in one embodiment of the present application;

[0043] Figure 2 is a flowchart of an image recognition method provided by an embodiment of the present application;

[0044] Figure 3 is a flowchart of an image recognition method provided by an embodiment of the present application;

[0045] Figure 4 is a flowchart of an image recognition method provided by an embodiment of the present application;

[0046] Figure 5 It is a framework diagram of a low-quality picture comment detection method based on repetition provided by an embodiment of the present application;

[0047] Figure 6 is a block diagram of an image recognition device provided by an exemplary embodiment of the present application;

[0048] Figure 7 It is a structural diagram of a computer device provided by an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0049] Here, exemplary embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0050] Rather, they are merely examples of apparatus and methods consistent with some aspects of the present application as recited in the appended claims.

[0051] The present application proposes an image recognition scheme, which can identify whether an image is a low-quality image through the repetition between the first comment image and the historical comment images and the repetition between the first comment image and the low-quality comment images, thereby improving the accuracy of image recognition. For ease of understanding, some concepts involved in the present application are explained below.

[0052] 1) Image features refer to unique and identifiable information extracted from an image in digital image processing or computer vision. These features can be used to describe objects, scenes, or patterns in an image and can help algorithms understand and interpret image content.

[0053] 2) Similarity is a concept that measures the degree of similarity between two or more entities. It is widely used in many fields, including but not limited to text analysis, image processing, recommendation systems, bioinformatics, etc. Similarity can be defined and calculated in different ways. Similarity is usually a value between a minimum value (usually 0) and a maximum value (usually 1), where 0 means that the two entities are completely different, 1 means that the two entities are exactly the same, and intermediate values ​​between 0 and 1 represent different degrees of similarity.

[0054] 3) Feature similarity is a measure of the similarity between two or more feature vectors.

[0055] Figure 1 1 is a system configuration diagram of an image recognition method according to an embodiment of the present application. Figure 1 As shown, the system includes a user terminal device 120 for image recognition and a server 130.

[0056] The user terminal device 120 may be a personal computer device such as a desktop computer, a laptop computer, a personal workstation, etc.

[0057] The server 130 may be one server, or a plurality of servers, or a virtualization platform, or a cloud computing service center.

[0058] The user terminal device 120 and the server 130 may be connected via a communication network. Optionally, the communication network is a wired network or a wireless network.

[0059] Optionally, the above-mentioned wireless network or wired network uses standard communication technology and / or protocol. The network is usually the Internet, but it can also be any network, including but not limited to any combination of local area network (LAN), metropolitan area network (MAN), wide area network (WAN), mobile, wired or wireless network, private network or virtual private network. In some embodiments, the data exchanged through the network is represented by technology and / or format including Hyper Text Mark-up Language (HTML), Extensible Markup Language (XML), etc. In addition, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), Internet Protocol Security (IPsec) can also be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technology can also be used to replace or supplement the above-mentioned data communication technology.

[0060] exist Figure 1 In the system shown, the user terminal device 120 can publish network content to the network platform, and make comments on the network content in the network platform. The server 130 can store and manage the comment content. Specifically, it can manage historical comment images and low-quality comment images, and when a new comment image is generated, it can identify whether the new comment image is a low-quality comment image through the repetition between the new comment image and the historical comment images and the low-quality comment images.

[0061] Specifically, the server 130 may include a historical comment image library and a low-quality comment image library, wherein the historical comment image library includes at least one historical comment image, and the low-quality comment image library includes at least one low-quality comment image. The user terminal device 120 may provide the image to be identified (first comment image) to the server 130, or the server 130 may automatically obtain the image to be identified (first comment image) from the network platform. The server 130 obtains the first comment image, obtains a first repetition degree based on the repetition between the first comment image and each historical comment image, obtains a second repetition degree based on the repetition between the first comment image and each low-quality comment image, compares the first repetition degree and the second repetition degree with the first condition respectively, and identifies the first comment image as a new low-quality image when the first repetition degree and the second repetition degree meet the first condition.

[0062] It should be noted that, in the embodiments of the present application, the above solution is implemented by a single-sided device as an example. Optionally, the above solution can also be implemented by the interaction of two-sided devices.

[0063] Figure 2 is a flowchart of an image recognition method provided by an embodiment of the present application. The image recognition can be performed by a computer device, for example, the computer device can be the above-mentioned Figure 1 The server 130 shown, or the computer device mentioned above, may also be other computer devices, such as a server, or a personal computer, etc. The image recognition method mentioned above may include the following steps.

[0064] Step 210: Acquire a first comment image, where the first comment image is an image included in a newly generated comment.

[0065] The above comments may be comments containing images published by users on the network platform; the above newly generated comments are the latest comments containing images published by users on the network platform.

[0066] In an embodiment of the present application, a user can post a comment containing an image on a network platform. After the computer device detects the latest updated comment on the network platform, it obtains the image contained in the comment as the first comment image, and performs subsequent image recognition. It can process the image in the comment posted by the user in a timely manner, thereby achieving timeliness of image recognition.

[0067] For example, a computer device (eg, a server) may obtain an image included in a comment newly posted by a user on a network platform within a time period (eg, 5 minutes) before the current moment as the first comment image.

[0068] Step 220: Obtain a first repetition degree of the first comment image, where the first repetition degree is used to indicate a repetition degree between the first comment image and each historical comment image.

[0069] The first repetition degree can be a single value, indicating the number of historical comment images that are repeated with the first comment image in each historical comment image; or a percentage, indicating the proportion of historical comment images that are repeated with the first comment image in each historical comment image. In some embodiments, the first repetition degree is positively correlated with the repetition situation, that is, the higher the first repetition degree, the higher the repetition degree between the first comment image and each historical comment image; conversely, the lower the first repetition degree, the lower the repetition degree between the first comment image and each historical comment image.

[0070] The above-mentioned historical comment images may be images contained in historical comments posted by users on a network platform, and the historical comments are comments before the newly generated comments.

[0071] In some possible implementations, the above-mentioned repetition situations may include complete repetition (the first comment image is exactly the same as the historical comment image), partial repetition (part of the content in the first comment image is the same as the historical comment image) or no repetition (the first comment image is completely different from the historical comment image).

[0072] In some embodiments, the computer device may include a historical comment image database, which includes at least one historical comment image. In some embodiments, after the computer device detects that a new comment including an image is generated on the network platform, the computer device obtains an image in the previous comment of the newly generated comment as a historical comment image and saves it in the historical comment image database. Alternatively, the computer device may also periodically update the historical comment image database.

[0073] In some embodiments, after the computer device obtains the first comment image, it can obtain historical comment images from a historical comment image database, compare the first comment image with each historical comment image in the historical comment database, and obtain the repetition between the first comment image and each historical comment image.

[0074] Step 230: Obtain a second repetition degree of the first comment image, where the second repetition degree is used to indicate the repetition between the first comment image and each low-quality comment image.

[0075] The second repetition degree can be a single value, indicating the number of low-quality comment images that are repeated with the first comment image in each low-quality comment image; or it can be a percentage, indicating the proportion of low-quality comment images that are repeated with the first comment image in each low-quality comment image. In some embodiments, the second repetition degree is positively correlated with the above-mentioned repetition situation, that is, the higher the second repetition degree, the higher the repetition degree between the first comment image and each low-quality comment image; conversely, the lower the second repetition degree, the lower the repetition degree between the first comment image and each low-quality comment image.

[0076] Among them, the above-mentioned low-quality comment images may be low-quality images contained in comments posted by users on the network platform. The low-quality images may be images that do not meet the standards specified by the network platform (for example, involving violent content), do not meet user expectations (for example, the image subject is different from the image subject expected by the user) or the image itself has quality defects (for example, the image is blurry).

[0077] The above-mentioned duplication may be complete duplication (the first comment image is exactly the same as the low-quality comment), partial duplication (part of the content in the first comment image is the same as the low-quality comment) or non-duplication (the first comment image is completely different from the low-quality comment).

[0078] In an embodiment of the present application, a computer device may include a low-quality comment image database, and the low-quality comment image includes at least one low-quality comment image. In some embodiments, after the computer device detects a low-quality comment containing an image on a network platform, it obtains the image in the low-quality comment and saves it in the low-quality comment database as a low-quality comment image. Optionally, a user may report a comment containing an image, and after the computer device receives the report, it obtains the image in the comment and saves it in the low-quality comment image database; Optionally, a user may report an image in a comment, and after the computer device receives the report, it obtains the image as a low-quality comment image and saves it in the low-quality comment image database. Alternatively, the computer device may also periodically update the low-quality comment image database.

[0079] In some embodiments, after obtaining the first comment image, the computer device may obtain low-quality comment images from a low-quality comment image database, compare the first comment image with each low-quality comment image in the low-quality comment image database, and obtain the repetition between the first comment image and each low-quality comment image.

[0080] Step 240: When the first repetition degree and the second repetition degree satisfy a first condition, identify the first comment image as a new low-quality image.

[0081] The first condition may be a condition preset by a developer.

[0082] For example, the first condition may be a threshold value, indicating a specific threshold value that both the first repetition degree and the second repetition degree can reach. The first repetition degree and the second repetition degree satisfying the first condition indicate that the first repetition degree reaches the threshold value, and the second repetition degree reaches the threshold value, that is, the repetition between the first comment image and each historical comment image reaches a specific threshold value, and the repetition between the first comment image and each low-quality comment image reaches a specific threshold value.

[0083] In the embodiment of the present application, when the first degree of repetition satisfies the first condition and the second degree of repetition satisfies the first condition, the computer device identifies the first comment image as a new low-quality image.

[0084] In the embodiment of the present application, after obtaining the image in the newly generated comment (the first comment image), the computer device promptly obtains the first repetition and the second repetition of the first comment image. The first repetition can be used to obtain the repetition between the first comment image and each historical comment image, that is, the repetition degree between the image in the newly generated comment and the image in each historical comment. If the repetition degree is high, it means that the image in the newly generated comment may be a repeated image, that is, the image in the newly generated comment may have appeared in the historical comments; the second repetition can be used to obtain the repetition between the first comment image and each low-quality comment image, that is, the repetition degree between the image in the newly generated comment and the image in each low-quality comment image. If the repetition degree is high, it means that the image in the newly generated comment may be a low-quality comment image, that is, the local / all features in the image in the newly generated comment may be the same as the local / all features in the low-quality comment image. When the first repetition degree and the second repetition degree satisfy the first condition, that is, both repetition degrees satisfy the specified condition, it means that the first comment image is not only similar to the historical comment image, but also similar to the known low-quality comment image. By combining the two factors to identify the image, the situation of misjudging high-quality comment images as low-quality comment images can be reduced, thereby effectively improving the accuracy of image recognition.

[0085] Based on the schemes shown in any one or more of the above-mentioned embodiments, in some embodiments, the first repetition degree is used to indicate the number of historical comment images among each historical comment image, whose similarity with the first comment image is greater than a first similarity threshold; the second repetition degree is used to indicate the number of low-quality comment images among each low-quality comment image, whose similarity with the first comment image is greater than a second similarity threshold.

[0086] The similarity indicates the degree of closeness between the first comment image and the contents of each historical comment image / low-quality comment image. The greater the similarity, the closer the first comment image is to the contents of each historical comment image / low-quality comment image; conversely, the lower the similarity, the lower the degree of closeness between the first comment image and the contents of each historical comment image / low-quality comment image. In some embodiments, the similarity is a value in the range of [0, 1], where 1 indicates that the two images are exactly the same, and 0 indicates that the two images are completely different.

[0087] In the embodiment of the present application, the above similarity can be obtained by an image similarity calculation method, and the image similarity calculation method includes but is not limited to: one of Euclidean distance, Manhattan distance, Mahalanobis distance, correlation distance and cosine similarity.

[0088] In some embodiments, the computer device can input the first comment image and each historical comment image into a feature extraction network respectively, obtain the feature vectors corresponding to the first comment image and each historical comment image through the feature extraction network, and then use the feature vector of the first comment image and the feature vector of a single historical comment image to obtain the similarity between the first comment image and the single image.

[0089] In some embodiments, the computer device can input the first comment image and each low-quality comment image into a feature extraction network respectively, obtain the feature vectors corresponding to the first comment image and each low-quality comment image through the feature extraction network, and then use the feature vector of the first comment image and the feature vector of the single low-quality comment image to obtain the similarity between the first comment image and the single image.

[0090] The first similarity threshold can be preset by the developer and can be a value in the range of [0, 1], where 1 indicates that the first comment image is exactly the same as the historical comment image, and 0 indicates that the first comment image is completely different from the historical comment image. The larger the similarity threshold, the higher the similarity between the first comment image and the historical comment image; conversely, the lower the similarity threshold, the lower the similarity between the first comment image and the historical comment image.

[0091] The second similarity threshold can be preset by the developer and can be a value in the range of [0, 1], where 1 indicates that the first comment image is exactly the same as the low-quality comment image, and 0 indicates that the first comment image is completely different from the low-quality comment image. The larger the similarity threshold, the closer the first comment image is to the low-quality comment image; conversely, the lower the similarity threshold, the lower the similarity threshold, the closer the first comment image is to the low-quality comment image.

[0092] In an embodiment of the present application, a computer obtains the similarity between the first comment image and each historical comment image, and obtains multiple similarities that are the same as the number of historical comment images, each similarity represents the degree of similarity between the first comment image and a single historical comment image, and compares the multiple similarities with a first similarity threshold respectively, and counts the number of historical comment images in each historical comment image whose similarity with the first comment image is greater than the first similarity threshold as the first repetition degree.

[0093] In an embodiment of the present application, a computer obtains the similarity between the first comment image and each low-quality comment image, and obtains multiple similarities equal to the number of each low-quality comment image, each similarity represents the degree of similarity between the first comment image and a single low-quality comment image, and compares the multiple similarities with the second similarity threshold respectively, and counts the number of historical comment images in each low-quality comment image whose similarity with the first comment image is greater than the second similarity threshold as the second repetition degree.

[0094] In an embodiment of the present application, through the first similarity threshold, the computer device can effectively screen out historical comment images that are highly similar to the first comment image. Through the second similarity threshold, the computer device can further filter out non-low-quality duplicate images among the acquired similar images, thereby improving the accuracy of low-quality image recognition. The above scheme performs image recognition through a double verification mechanism, that is, simultaneously considering the similarity of the first comment image with the historical comment images and the known low-quality comment images, thereby avoiding errors that may be introduced by image recognition based on a single mechanism, ensuring the reliability of image recognition results, and effectively improving the accuracy of image recognition.

[0095] based on Figure 2 , please refer to Figure 3 , Figure 3 This is a flowchart of an image recognition method provided by an embodiment of the present application. Figure 2 Step 220 in the embodiment can be implemented as step 220a and step 220b, as follows:

[0096] Step 220a: Obtain the image features of each historical comment image and the feature similarities between them and the image features of the first comment image, as the similarities between each historical comment image and the first comment image.

[0097] Optionally, the computer device may directly extract image features of the first comment image and each historical comment image through a pre-trained feature extraction network.

[0098] Optionally, the computer device may also extract image features of each historical comment image through a feature extraction network in advance, and store the extracted image features of each historical comment image in an image feature library. When obtaining the similarity between each historical comment image and the first comment image, the image features of each historical comment image may be directly obtained through the image feature library. In this embodiment, an image feature library of historical comment images may be preset in the computer device, and the image feature library contains image features of at least one historical comment image, and different image features correspond to different historical comment images.

[0099] Optionally, the computer device may extract image features from the historical comment images. For example, in an embodiment of the present application, after acquiring the historical comment images, the computer device inputs the historical comment images into a pre-trained feature extraction network, acquires image features corresponding to the historical comment images, and stores the extracted image features of each historical comment image in an image feature library of the historical comment images.

[0100] Step 220b: obtaining the number of similarities between each historical comment image and the first comment image that is greater than a first similarity threshold as a first repetition degree.

[0101] In an embodiment of the present application, a computer obtains the similarity between the first comment image and each historical comment image, and obtains multiple similarities that are the same as the number of historical comment images, each similarity represents the degree of similarity between the first comment image and a single historical comment image, and compares the multiple similarities with a first similarity threshold respectively. When the similarity is greater than the first similarity threshold, the similarities are counted, and one historical comment image is counted as 1. The similarities are accumulated in sequence, and the number of historical comment images whose similarity with the first comment image is greater than the first similarity threshold is counted as the first repetition degree.

[0102] In an embodiment of the present application, the image features can represent the key information in the historical comment images and the first comment images, effectively reflect the main contents of the historical comment images and the first comment images, and calculate the similarity between each historical comment image and the first comment image based on the image features, so as to effectively distinguish the subtle differences between the two images. Even if the two images look very similar visually, slight differences between the two images can be found by comparing the two image features, thereby avoiding misjudgment of similarity, and effectively ensuring the accuracy of the similarity between each historical comment image and the first comment image. Obtaining the first repeatability based on the feature similarity with higher accuracy can effectively ensure the accuracy of the first repeatability, thereby effectively improving the accuracy of subsequent image recognition operations performed based on the first repeatability.

[0103] based on Figure 2 , please refer to Figure 4 , Figure 4 This is a flowchart of an image recognition method provided by an embodiment of the present application. Figure 2 Step 230 in the embodiment can be implemented as step 230a and step 230b, as follows:

[0104] Step 230a: Obtain image features of each low-quality comment image, and obtain feature similarities between the image features of the first comment image, as the similarities between each low-quality comment image and the first comment image.

[0105] Optionally, the computer device may directly extract image features of the first comment image and each low-quality comment image through a pre-trained feature extraction network.

[0106] Optionally, the computer device may also extract image features of each low-quality comment image through a feature extraction network in advance, and store the extracted image features of each low-quality comment image in an image feature library. When obtaining the similarity between each low-quality comment image and the first comment image, the image features of each low-quality comment image may be directly obtained through the image feature library. In this embodiment, an image feature library of low-quality comment images may be preset in the computer device, and the image feature library contains image features of at least one low-quality comment image, and different image features correspond to different low-quality comment images.

[0107] Optionally, the computer device may extract image features from the low-quality comment image. For example, in an embodiment of the present application, after acquiring the low-quality comment image, the computer device inputs the low-quality comment image into a pre-trained feature extraction network to acquire image features corresponding to the low-quality comment image, and stores the extracted image features of each historical comment image into an image feature library of the low-quality comment image.

[0108] Step 230b: obtaining the number of similarities between each low-quality comment image and the first comment image that are greater than a second similarity threshold as a second repetition degree.

[0109] In an embodiment of the present application, a computer obtains the similarity between the first comment image and each low-quality comment image, and obtains multiple similarities that are the same as the number of low-quality comment images, each similarity represents the degree of similarity between the first comment image and a single low-quality comment image, and compares the multiple similarities with the second similarity threshold respectively. When the similarity is greater than the second similarity threshold, one low-quality comment image is counted as 1, and the similarities are accumulated in sequence. The number of low-quality comment images whose similarity with the first comment image is greater than the second similarity threshold is counted as the second repetition degree.

[0110] In an embodiment of the present application, the image features can represent the key information in the low-quality comment image and the first comment image, effectively reflect the main content of the low-quality comment image and the first comment image, and calculate the similarity between each low-quality comment image and the first comment image based on the image features, so as to effectively distinguish the subtle differences between the two images. Even if the two images look very similar visually, by comparing the two image features, it is possible to find slight differences between the two images, thereby avoiding misjudgment of similarity, and effectively ensuring the accuracy of the similarity between each low-quality comment image and the first comment image. Obtaining the second repeatability based on the feature similarity with higher accuracy can effectively ensure the accuracy of the second repeatability, thereby effectively improving the accuracy of subsequent image recognition operations performed based on the second repeatability.

[0111] Based on the solutions shown in any one or more of the above embodiments, in some embodiments, the first similarity threshold is positively correlated with the number of each historical comment image; the second similarity threshold is positively correlated with the number of each low-quality comment image.

[0112] In the embodiment of the present application, a greater total number of historical comment images means a greater base number of repeated images with which the first comment image can be compared, and a greater possibility of repeated or similar images appearing in each historical comment image as well as the first comment image will also increase. The higher the first similarity threshold is set, that is, the higher the requirement for judging the degree of repetition between the first comment image and the historical comment images. The accuracy of the degree of repetition between the identified first comment image and the historical comment images can be ensured by raising the threshold.

[0113] In the embodiment of the present application, a greater total number of low-quality comment images means a greater base number of low-quality images with which the first comment image can be compared, and a higher second similarity threshold is set, that is, a higher requirement for determining the degree of repetition between the first comment image and the low-quality comment image.

[0114] In an embodiment of the present application, the first similarity threshold can be dynamically adjusted as the number of historical comment images increases (the first similarity threshold is set higher), and the second similarity threshold can be dynamically adjusted as the number of low-quality comment images increases (the second similarity threshold is set higher).

[0115] In the embodiment of the present application, the larger the cardinality of the comparison images (historical comment images / low-quality comment images), the larger the similarity threshold. By dynamically adjusting the threshold, it is possible to ensure the stability of image repetition recognition when facing comparison images of different sizes, thereby ensuring the accuracy of the image repetition recognition results.

[0116] Based on the solutions shown in any one or more of the above embodiments, in some embodiments, the first condition includes: the first repetition degree is greater than a first repetition degree threshold, and the second repetition degree is greater than a second repetition degree threshold.

[0117] The first repetition threshold can be preset by the developer. The first repetition threshold can be a single value, indicating the number of historical comment images that are repeated with the first comment image in each historical comment image; or it can be in the form of a percentage, indicating the proportion of historical comment images that are repeated with the first comment image in each historical comment image. In some embodiments, the higher the first repetition threshold, the higher the requirement for judging the degree of repetition between the first comment image and each historical comment image; conversely, the lower the first repetition threshold, the lower the requirement for judging the degree of repetition between the first comment image and each historical comment image.

[0118] The second repetition threshold can be preset by the developer. The second repetition threshold can be a single value, indicating the number of low-quality comment images that are repeated with the first comment image in each low-quality comment image; it can also be in the form of a percentage, indicating the proportion of low-quality comment images that are repeated with the first comment image in each low-quality comment image. In some embodiments, the higher the second repetition threshold, the higher the requirement for judging the degree of repetition between the first comment image and each low-quality comment image; conversely, the lower the second repetition threshold, the lower the requirement for judging the degree of repetition between the first comment image and each low-quality comment image.

[0119] In an embodiment of the present application, by conditionally limiting the degree of repetition between the first comment image and each historical comment image, as well as the degree of repetition between the first comment image and each low-quality comment image, low-repetition situations are filtered out, thereby effectively improving the accuracy of identifying new low-quality images.

[0120] Based on the scheme shown in any one or more of the above-mentioned embodiments, in some embodiments, before the above-mentioned step 210, the computer device obtains a second comment image, which is an image contained in a comment generated within a first time period before the current time point; obtains the similarity between the second comment image and each historical comment image; and when the similarity between the second comment image and each historical comment image is not greater than a third similarity threshold, obtains the second comment image as a new historical comment image.

[0121] The first time period may be a time period preset by a developer, for example, within an hour before the current time point, within three days before the current day, or within a month before the current month. That is, the computer device may periodically acquire new historical review images.

[0122] The third similarity threshold may be preset by a developer and may be a value in the range of [0, 1], where 1 indicates that the first comment image is exactly the same as the historical comment image, and 0 indicates that the first comment image is completely different from the historical comment image.

[0123] After acquiring the second comment image, the computer device calculates the similarity between the second comment image and each historical comment image to obtain multiple similarities, compares each similarity with a third similarity threshold, and when each similarity is not greater than the third similarity threshold, acquires the second comment image as a new historical comment image.

[0124] In an embodiment of the present application, after acquiring a new historical comment image, the computer device may save the new historical comment image in a historical comment image library, where the historical comment image library is used to store historical comment images, and the historical comment image library includes at least one historical comment image library.

[0125] In other embodiments, a computer device may obtain image features of a historical comment image, and save the image features of the historical comment image in an image feature library of historical comment images, wherein the image feature library of historical comment images is used to store image features of images, and the image feature library of historical comment images contains at least one image feature of a historical comment image.

[0126] In an embodiment of the present application, a computer device may periodically acquire images from comments within a specified time period before the current moment, and compare the acquired second comment image with the historical comment image. If the similarity between the second comment image and the historical comment image is not high, the second comment image is acquired as a new historical comment image, thereby effectively ensuring that the new historical comment image has low similarity to the original historical comment image, avoiding the situation where the new historical comment image is repeated with the original historical comment image, expanding the comparison image library of the first comment image, and effectively ensuring the accuracy of subsequent image recognition.

[0127] Based on the scheme shown in any one or more of the above-mentioned embodiments, in some embodiments, before the above-mentioned step 210, the computer device obtains feedback information of the third comment image, the feedback information includes the number of negative feedbacks, and the third comment image is an image included in the comments generated in the second time period before the current time point; when the feedback information of the third comment image satisfies the second condition, the similarity between the third comment image and each low-quality comment image is obtained; when the similarity between the third comment image and each low-quality comment image is not greater than a fourth similarity threshold, the third comment image is obtained as a new low-quality comment image.

[0128] In the embodiment of the present application, the user can give negative feedback to the third comment image, and each third comment image has corresponding feedback information, which includes the number of negative feedbacks received by the third comment image.

[0129] The second time period may be a time period preset by the developer, for example, within one hour before the current time point, within three days before the current day, or within one month before the current month. That is, the computer device may periodically obtain new low-quality review images.

[0130] The second condition may be preset by a developer, for example, a number of negative feedbacks preset by the developer.

[0131] Among them, the above-mentioned fourth similarity threshold can be preset by the developer and can be a value in the range of [0, 1], 1 means that the first comment image is exactly the same as the low-quality comment image, and 0 means that the first comment image is completely different from the low-quality comment image.

[0132] After obtaining the third comment image, the computer device obtains feedback information corresponding to the third comment image. When the feedback information satisfies the second condition, the computer device obtains the similarity between the third comment image and each low-quality comment image, obtains multiple similarities, compares each similarity with a fourth similarity threshold, and when each similarity is not greater than the fourth similarity threshold, obtains the fourth comment image as a new low-quality comment image.

[0133] In an embodiment of the present application, after acquiring a new low-quality comment image, the computer device can save the new low-quality comment image in a low-quality comment image library, which is used to store low-quality comment images, and the low-quality comment image library contains at least one low-quality comment image library.

[0134] In other embodiments, the computer device may obtain image features of the low-quality comment image, and save the image features of the low-quality comment image in an image feature library, where the image feature library is used to store image features of images, and the image feature library contains at least one image feature of the low-quality comment image.

[0135] In an embodiment of the present application, a computer device may periodically obtain images in comments within a specified time period before the current moment, and obtain feedback information of a third comment image. When the feedback information satisfies specified conditions and the similarity between the third comment image and the low-quality comment image is not high, the third comment image is obtained as a new low-quality comment image, thereby effectively ensuring that the new low-quality comment image has a low similarity to the original low-quality comment image, avoiding the situation where the new low-quality comment image is repeated with the original low-quality comment image, expanding the comparison image library of the first comment image, and effectively ensuring the accuracy of subsequent image recognition.

[0136] Based on the solutions shown in any one or more of the above embodiments, in some embodiments, the second condition includes: the number of negative feedbacks is greater than a first number threshold.

[0137] The first quantity threshold may be a value preset by a developer.

[0138] In an embodiment of the present application, after obtaining feedback information of the third comment image, the computer device can obtain the number of negative feedbacks in the feedback information. By comparing the number of negative feedbacks with the first quantity threshold, the number of negative feedbacks can reflect the quality of the third comment image, thereby effectively ensuring the accuracy of the quality judgment of the third comment image.

[0139] Based on the scheme shown in any one or more of the above embodiments, in some embodiments, the feedback information also includes the number of positive feedback; the second condition includes: the number of negative feedback is greater than the first number threshold, and the number of positive feedback is less than the second number threshold; or, the number of negative feedback is greater than the first number threshold, and the ratio between the number of negative feedback and the number of positive feedback is greater than the ratio threshold.

[0140] In the embodiment of the present application, the user can provide positive feedback to the third comment image, and the positive feedback indicates that the quality of the third comment image is relatively high.

[0141] In an embodiment of the present application, after acquiring the third comment image, the computer device reads the feedback information of the third comment image, and obtains the number of positive feedback and the number of negative feedback of the third comment image therefrom; when the number of negative feedback of the third comment image is greater than a first quantity threshold and the number of positive feedback is less than a second quantity threshold, the third comment image is acquired as a new low-quality comment image.

[0142] In an embodiment of the present application, after acquiring the third comment image, the computer device reads the feedback information of the third comment image, and obtains the number of positive feedback and the number of negative feedback of the third comment image therefrom; when the number of negative feedback is greater than a first quantity threshold, and the ratio between the number of negative feedback and the number of positive feedback is greater than a ratio threshold, the third comment image is acquired as a new low-quality comment image.

[0143] In the embodiment of the present application, the computer device is not limited to the single factor of the number of negative feedbacks to determine whether the third comment image is a new low-quality comment image, but also obtains the number of positive feedbacks of the third comment image. By making a comprehensive judgment through the number of positive feedbacks, it is possible to effectively avoid the misjudgment of the originally high-quality third comment image due to malicious negative feedback from the user, thereby effectively improving the accuracy of the judgment of whether the third comment image is a new low-quality comment image, thereby ensuring the accuracy of subsequent image recognition.

[0144] Based on the solutions shown in any one or more of the above embodiments, in some embodiments, negative feedback includes at least one of the following: reporting that the image has quality problems, reporting that the image content is illegal, and marking the image as spam.

[0145] In an embodiment of the present application, a user can provide negative feedback on images in other comments (not comments posted by the user) on the network platform.

[0146] In the embodiment of the present application, the above scheme expands the types of negative feedback, effectively expands the standard range for low-quality images, can more comprehensively obtain low-quality comment images, expand the base of comparison images for subsequent image recognition, and effectively ensure the accuracy of image recognition.

[0147] Based on the above Figures 2 to 6 The steps in the embodiment of the present application illustrate a method for detecting low-quality picture comments based on repetition.

[0148] This embodiment of the scheme provides a method for detecting low-quality comment pictures, which extracts the feature vector of the picture in the target comment picture (equivalent to the first comment image in the above embodiments), matches it in the duplicate picture database and the low-quality duplicate picture database, obtains the picture repetition degree of the target comment picture (equivalent to the first repetition degree in the above embodiments) and the repetition degree of similar low-quality pictures (equivalent to the second repetition degree in the above embodiments), and determines whether the comment picture is of low quality according to the preset similar picture repetition degree threshold (equivalent to the first repetition degree threshold in the above embodiments) and the similar low-quality picture repetition degree threshold (equivalent to the second repetition degree threshold in the above embodiments). The feature vector of the comment picture determined to be a low-quality picture is stored in the duplicate picture database and the low-quality picture database for subsequent detection of new target comment pictures. This scheme can continuously improve the accuracy of detecting low-quality comment pictures.

[0149] Illustratively, this embodiment involves several related variable meanings, please refer to Table 1, which is a definition table of related variable meanings.

[0150] Table 1 Definition of relevant variables

[0151]

[0152]

[0153] For example, please refer to Figure 5 , Figure 5 It is a framework diagram of a method for detecting low-quality picture comments based on repetition provided in one embodiment of the present application.

[0154] This solution is implemented through online services and offline scripts:

[0155] Negative feedback image collection interface: Users view comments through the client. For comment images that users think are of low quality, users can click the feedback control on the client to provide feedback corresponding to the comment image. After the comment server receives the client's feedback request, it counts the number of times the comment has been fed back.

[0156] Comment timing export script: The computer device regularly reads the recently saved comment image information Icmt from the comment database, submits it to the image recognition and retrieval service, extracts its feature vector Xcmt, and stores it in the duplicate image database DBcmt.

[0157] Script for periodically exporting low-quality comment images: The computer device regularly screens recent comment images from the comment database. Comments that have been reported more than a threshold are considered low-quality comments by the computer device. The script extracts the image information Ibad of the comment and submits it to the image recognition and retrieval service, extracts its feature vector Xbad, and stores it in the low-quality image database DBbad.

[0158] Asynchronous detection script for low-quality comment images: After a user posts a new comment with an image, the computer device will asynchronously read the comment image Ic from the comment database and submit it to the image recognition and retrieval service, extract its feature vector Xc, match similar images from the duplicate image database DBcmt and the low-quality duplicate image database DBbad, and obtain the image duplication Scmt and low-quality image duplication Sbad respectively. If Sbad>Tbad and Scmt>Tcmt, the comment image can be considered a low-quality duplicate image, and then the script will mark the comment image as a "low-quality duplicate image" in the comment data.

[0159] Image similarity calculation: Computer equipment can use pre-trained deep learning network models to calculate image similarity, such as ResNet (Residual Network), VGG (Visual Geometry Group), etc. Input image I into the network model, take the hidden layer from the network structure as the feature vector X; the vector features X and Y of the two images, calculate the cosine similarity of the two images, and determine whether the two images are similar. For example, please refer to formula (1), which is the cosine similarity calculation formula:

[0160]

[0161] Among them, X and Y are the hidden layer feature vectors of picture 1 and picture 2 respectively. n is the dimension of the hidden layer in the network model. i ,y i are the i-th components of the feature vectors respectively. cos(θ) is the cosine similarity of feature vectors X and Y, which is used to characterize the similarity between picture 1 and picture 2. X and Y can be expressed as:

[0162] X=(x1, x2, x3,...xn); Y=(y1, y2, y3,...yn).

[0163] is the dot product of vectors X and Y, are the modulus lengths of vectors X and Y respectively. If the similarity is greater than the threshold T, it is determined that there are duplicate images between the retrieved image and the image in the feature library.

[0164] The above is only an embodiment of the present application and should not be regarded as limiting the present application. Professionals should understand that various deformations and modifications can be made to the embodiments to adapt to different application requirements. Therefore, the scope of the present application should be limited by the claims attached to the claims.

[0165] Please refer to Figure 6 , which shows a block diagram of an image recognition device provided by an exemplary embodiment of the present application. The image recognition device can be implemented as all or part of a computer device by hardware or a combination of hardware and software to achieve the above Figures 2 to 4 All or part of the steps in the embodiments shown. Figure 6 As shown, the image recognition device includes:

[0166] A first comment image acquisition module 601 is used to acquire a first comment image, where the first comment image is an image included in a newly generated comment;

[0167] A first repetition degree acquisition module 602, used to acquire a first repetition degree of the first comment image, where the first repetition degree is used to indicate the repetition between the first comment image and each historical comment image;

[0168] A second repetition degree acquisition module 603 is used to acquire a second repetition degree of the first comment image, where the second repetition degree is used to indicate the repetition between the first comment image and each low-quality comment image;

[0169] The low-quality image identification module 604 is configured to identify the first comment image as a new low-quality image when the first repetition degree and the second repetition degree satisfy a first condition.

[0170] In some embodiments, the first repetition degree is used to indicate the number of historical comment images among each historical comment image, whose similarity with the first comment image is greater than a first similarity threshold; the second repetition degree is used to indicate the number of low-quality comment images among each low-quality comment image, whose similarity with the first comment image is greater than a second similarity threshold.

[0171] In some embodiments, the first repetition acquisition module 602 is used to acquire the feature similarity between the image features of each historical comment image and the image features of the first comment image, as the similarity between each historical comment image and the first comment image;

[0172] The number of similarities between each historical comment image and the first comment image that is greater than a first similarity threshold is obtained as a first repetition degree.

[0173] In some embodiments, the second repetition acquisition module 603 is used to acquire the feature similarity between the image features of each low-quality comment image and the image features of the first comment image, as the similarity between each low-quality comment image and the first comment image;

[0174] The number of similarities between each low-quality comment image and the first comment image that is greater than a second similarity threshold is obtained as a second repetition degree.

[0175] In some embodiments, the first similarity threshold is positively correlated with the number of each historical comment image; and the second similarity threshold is positively correlated with the number of each low-quality comment image.

[0176] In some embodiments, the first condition includes: the first repetitiveness is greater than a first repetitiveness threshold, and the second repetitiveness is greater than a second repetitiveness threshold.

[0177] In some embodiments, the apparatus further comprises:

[0178] A second comment image acquisition module is used to acquire a second comment image before acquiring the first comment image, where the second comment image is an image included in a comment generated in a first time period before a current time point;

[0179] A first similarity acquisition submodule is used to acquire the similarity between the second comment image and each historical comment image before acquiring the first comment image;

[0180] The new historical comment image acquisition module is used to acquire the second comment image as a new historical comment image before acquiring the first comment image if the similarity between the second comment image and each historical comment image is not greater than a third similarity threshold.

[0181] In some embodiments, the apparatus further comprises:

[0182] A feedback information acquisition module, used for acquiring feedback information of a third comment image before acquiring the first comment image, the feedback information including the number of negative feedbacks, the third comment image being an image included in the comments generated in a second time period before the current time point;

[0183] A second similarity acquisition submodule is used to acquire the similarity between the third comment image and each low-quality comment image before acquiring the first comment image, if the feedback information of the third comment image meets the second condition;

[0184] The new low-quality comment image acquisition module is used to acquire the third comment image as a new low-quality comment image before acquiring the first comment image if the similarity between the third comment image and each low-quality comment image is not greater than a fourth similarity threshold.

[0185] In some embodiments, the second condition includes: the amount of negative feedback is greater than a first amount threshold.

[0186] In some embodiments, the feedback information further includes the number of positive feedbacks; and the second condition includes:

[0187] The number of negative feedbacks is greater than a first number threshold, and the number of positive feedbacks is less than a second number threshold; or, the number of negative feedbacks is greater than the first number threshold, and a ratio between the number of negative feedbacks and the number of positive feedbacks is greater than a ratio threshold.

[0188] In some embodiments, the negative feedback includes at least one of the following: reporting that the image has quality issues, reporting that the image content is in violation of regulations, and marking the image as spam.

[0189] Please refer to Figure 7 , Figure 7 700 is a schematic diagram of the structure of a computer device provided by an exemplary embodiment of the present application. The computer device 700 includes a central processing unit (CPU) 701, a system memory 704 including a random access memory (RAM) 702 and a read-only memory (ROM) 703, and a system bus 705 connecting the system memory 704 and the central processing unit 701. The computer device 700 also includes a basic input / output system (I / O system) 706 that helps transmit information between various devices in the computer, and a large-capacity storage device 707 for storing an operating system 713, application programs 714 and other program modules 715.

[0190] The basic input / output system 706 includes a display 708 for displaying information and an input device 709 such as a mouse and a keyboard for user inputting information. The display 708 and the input device 709 are connected to the central processing unit 701 through an input / output controller 710 connected to the system bus 705. The basic input / output system 706 may also include an input / output controller 710 for receiving and processing inputs from a plurality of other devices such as a keyboard, a mouse, or an electronic stylus. Similarly, the input / output controller 710 also provides output to a display screen, a printer, or other types of output devices.

[0191] The mass storage device 707 is connected to the central processing unit 701 via a mass storage controller (not shown) connected to the system bus 705. The mass storage device 707 and its associated computer readable medium provide non-volatile storage for the computer device 700. That is, the mass storage device 707 may include a computer readable medium (not shown) such as a hard disk or a CD-ROM (Compact Disc Read-Only Memory) drive.

[0192] Without loss of generality, the computer-readable medium may include computer storage media and communication media. Computer storage media include volatile and non-volatile, removable and non-removable media implemented by any method or technology for storing information such as computer-readable instructions, data structures, program modules or other data. Computer storage media include RAM (Random Access Memory), ROM (Read-Only Memory), EPROM (Erasable Programmable Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory or other solid-state storage technologies, CD-ROM, DVD (Digital Video Disc) or other optical storage, cassettes, tapes, disk storage or other magnetic storage devices. Of course, those skilled in the art will appreciate that the computer storage media is not limited to the above. The above-mentioned system memory 704 and mass storage device 707 can be collectively referred to as memory.

[0193] The computer device 700 can be connected to the Internet or other network devices through a network interface unit 711 connected to the system bus 705 .

[0194] The memory also includes one or more programs, which are stored in the memory. The central processing unit 701 implements the one or more programs by executing the one or more programs. Figures 2 to 4 All or part of the steps in the method shown.

[0195] In an exemplary embodiment, a chip is also provided, which includes a programmable logic circuit and / or program instructions. When the chip runs on a computer device, it is used to implement all or part of the steps of the method shown in the above embodiments of the present application.

[0196] In an exemplary embodiment, a computer program product is also provided, the computer program product includes computer instructions, the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor reads and executes the computer instructions from the computer-readable storage medium to implement all or part of the steps of the method shown in the above-mentioned various embodiments of the present application.

[0197] In an exemplary embodiment, a computer-readable storage medium is further provided, in which a computer program is stored. The computer program is loaded and executed by a processor to implement all or part of the steps of the method shown in the above-mentioned embodiments of the present application.

[0198] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0199] Those skilled in the art should be aware that in one or more of the above examples, the functions described in the embodiments of the present application can be implemented with hardware, software, firmware, or any combination thereof. When implemented using software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein the communication media include any media that facilitates the transmission of a computer program from one place to another. The storage medium can be any available medium that a general or special-purpose computer can access.

[0200] The above description is only an optional embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. An image recognition method, characterized in that: The method comprises: Acquire a first comment image, where the first comment image is an image included in a newly generated comment; Acquire a first repetition degree of the first comment image, where the first repetition degree is used to indicate a repetition situation between the first comment image and each historical comment image; Acquire a second repetition degree of the first comment image, where the second repetition degree is used to indicate a repetition situation between the first comment image and each low-quality comment image; When the first repetition degree and the second repetition degree satisfy a first condition, the first comment image is identified as a new low-quality image.

2. The method according to claim 1, characterized in that: The first repetition degree is used to indicate the number of the historical comment images whose similarity with the first comment image is greater than a first similarity threshold among the historical comment images; The second repetition degree is used to indicate the number of the low-quality comment images whose similarity with the first comment image is greater than a second similarity threshold among the low-quality comment images.

3. The method according to claim 2, characterized in that The obtaining a first repetition degree of the first comment image includes: Obtaining feature similarities between image features of each of the historical comment images and image features of the first comment image, as the similarities between each of the historical comment images and the first comment image; The number of similarities between each of the historical comment images and the first comment image that is greater than the first similarity threshold is obtained as the first repetition degree.

4. The method according to claim 2, characterized in that: The obtaining a second repetition degree of the first comment image includes: Obtaining feature similarities between image features of each of the low-quality comment images and image features of the first comment image, as the similarities between each of the low-quality comment images and the first comment image; The number of similarities between each of the low-quality comment images and the first comment image that is greater than the second similarity threshold is obtained as the second repetition degree.

5. The method according to claim 2, characterized in that: The first similarity threshold is positively correlated with the number of each of the historical comment images; The second similarity threshold is positively correlated with the number of each of the low-quality comment images.

6. The method according to claim 1, characterized in that The first condition includes: The first repetitiveness is greater than a first repetitiveness threshold, and the second repetitiveness is greater than a second repetitiveness threshold.

7. The method according to any one of claims 1 to 6, characterized in that: Before acquiring the first comment image, the method further includes: Acquire a second comment image, where the second comment image is an image included in a comment generated within a first time period before a current time point; Obtaining the similarity between the second comment image and each of the historical comment images; In the case that the similarity between the second comment image and each of the historical comment images is not greater than a third similarity threshold, the second comment image is acquired as a new historical comment image.

8. The method according to any one of claims 1 to 6, characterized in that: Before acquiring the first comment image, the method further includes: Acquire feedback information of a third comment image, the feedback information including the number of negative feedbacks, the third comment image being an image included in a comment generated in a second time period before a current time point; When the feedback information of the third comment image satisfies the second condition, obtaining the similarity between the third comment image and each of the low-quality comment images; In a case where the similarity between the third comment image and each of the low-quality comment images is not greater than a fourth similarity threshold, the third comment image is acquired as a new low-quality comment image.

9. The method according to claim 8, characterized in that The second condition includes: The amount of the negative feedback is greater than a first amount threshold.

10. The method according to claim 8, characterized in that The feedback information also includes the number of positive feedbacks; the second condition includes: The number of negative feedbacks is greater than a first number threshold, and the number of positive feedbacks is less than a second number threshold; or, The number of the negative feedbacks is greater than a first number threshold, and a ratio between the number of the negative feedbacks and the number of the positive feedbacks is greater than a ratio threshold.

11. The method according to claim 8, characterized in that The negative feedback includes at least one of the following: Report image quality issues, report image content violations, and mark images as spam.

12. An image recognition device, characterized in that: The device comprises: A first comment image acquisition module, used to acquire a first comment image, where the first comment image is an image included in a newly generated comment; A first repetition degree acquisition module, used for acquiring a first repetition degree of the first comment image, where the first repetition degree is used for indicating a repetition between the first comment image and each historical comment image; A second repetition degree acquisition module, used for acquiring a second repetition degree of the first comment image, where the second repetition degree is used for indicating the repetition between the first comment image and each low-quality comment image; The low-quality image recognition module is configured to recognize the first comment image as a new low-quality image when the first repetition degree and the second repetition degree satisfy a first condition.

13. A computer device, characterized in that: The computer device includes a processor and a memory, wherein the memory stores instructions, and the instructions are executed by the processor to implement the repeated image recognition method according to any one of claims 1 to 11.

14. A computer-readable storage medium, characterized in that: The storage medium stores instructions, and the instructions are executed by a processor of a computer device to implement the image recognition method according to any one of claims 1 to 11.

15. A computer program product, characterized in that The computer program product includes computer instructions, which are stored in a computer-readable storage medium; the computer instructions are read and executed by a processor of a computer device to implement the image recognition method as described in any one of claims 1 to 11.

Images