Image search method, description information generation method, device and storage medium

Abstract

The embodiments of the present application provide a method for searching for an image, a method for generating description information, a device and a storage medium, wherein the method for searching for an image comprises: obtaining search information; based on the image description information of at least one image, searching for a target image that matches the search information in the at least one image; wherein the image description information is generated based on the image content data; and outputting the target image. Accordingly, in the embodiments of the present application, in the absence of information such as a title and context in the scene in which the image is located, the image description information can be analyzed from the content contained in the image itself. This ensures the feasibility of the image search operation in the above-mentioned specific scenario, and the image description information generated thereby can more accurately represent the image, thereby effectively improving the accuracy of searching for images.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to an image search method, a description information generation method, a device and a storage medium. Background Art

[0002] The technology of searching images by text refers to the technology of using text keywords as query conditions and returning images related to the text keywords.

[0003] At present, the technical principle of searching images by text is usually that the search engine's web crawler crawls images on Internet web pages to obtain information such as the title and context of the image, and then performs image search based on the input text keywords and information such as the title and context of the image.

[0004] However, in many scenarios, such as on private clouds, a large number of images are often isolated and lack information such as titles and context. In these scenarios, image search cannot be completed based on existing text-based image search technology. This seriously affects the adaptability and / or accuracy of text-based image search, bringing difficulties to image search. Summary of the invention

[0005] Multiple aspects of the present application provide an image search method, a description information generation method, a device and a storage medium to improve the applicability and / or accuracy of searching images.

[0006] The present application provides an image search method, including:

[0007] Get search information;

[0008] Based on the picture description information of at least one picture, searching for a target picture matching the search information in the at least one picture;

[0009] Wherein, the picture description information is generated based on the picture content data;

[0010] Output the target image.

[0011] The present application provides a resource search method, including:

[0012] Get search information;

[0013] Based on the description information of at least one resource, searching the at least one resource for a target resource matching the search information;

[0014] Wherein, the description information is generated based on resource content data;

[0015] The target resource is output.

[0016] The present application also provides a method for generating image description information, including:

[0017] Get the image to be processed;

[0018] Extracting picture content data from the picture to be processed;

[0019] Based on the picture content data, picture description information of the picture to be processed is generated.

[0020] The embodiment of the present application also provides a computing device, including a memory and a processor;

[0021] The memory is used to store one or more computer instructions;

[0022] The processor is coupled to the memory and configured to execute the one or more computer instructions for:

[0023] Get search information;

[0024] Based on the picture description information of at least one picture, searching for a target picture matching the search information in the at least one picture;

[0025] Wherein, the picture description information is generated based on the picture content data;

[0026] Output the target image.

[0027] The embodiment of the present application also provides a computing device, including a memory and a processor;

[0028] The memory is used to store one or more computer instructions;

[0029] The processor is coupled to the memory and configured to execute the one or more computer instructions for:

[0030] Get search information;

[0031] Based on the description information of at least one resource, searching the at least one resource for a target resource matching the search information;

[0032] Wherein, the description information is generated based on resource content data;

[0033] The target resource is output.

[0034] The embodiment of the present application also provides a computing device, including a memory and a processor;

[0035] The memory is used to store one or more computer instructions;

[0036] The processor is coupled to the memory and configured to execute the one or more computer instructions for:

[0037] Get the image to be processed;

[0038] Extracting picture content data from the picture to be processed;

[0039] Based on the picture content data, picture description information of the picture to be processed is generated.

[0040] An embodiment of the present application also provides a computer-readable storage medium storing computer instructions. When the computer instructions are executed by one or more processors, the one or more processors are caused to execute the aforementioned image search method or image description information generation method.

[0041] In the embodiment of the present application, the picture description information can be generated based on the picture content data of the picture. Therefore, in the case where the scene in which the picture is located lacks information such as title and context, the picture description information can be analyzed from the content contained in the picture itself. This ensures the feasibility of the picture search operation in the above-mentioned specific scene, and the picture description information generated thereby can more accurately represent the picture, thereby effectively improving the accuracy of the picture search. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0043] Figure 1 A schematic diagram of a flow chart of a picture search method provided in one embodiment of the present application;

[0044] Figure 2 A logical schematic diagram of an image analysis model provided in one embodiment of the present application;

[0045] Figure 3 A schematic diagram of a dedicated cloud scenario provided for an embodiment of the present application;

[0046] Figure 4 A schematic diagram of a flow chart of a method for generating picture description information provided by another embodiment of the present application;

[0047] Figure 5 A schematic diagram of the structure of a computing device provided in yet another embodiment of the present application;

[0048] Figure 6 A schematic diagram of the structure of another computing device provided in yet another embodiment of the present application;

[0049] Figure 7A schematic diagram of a flow chart of a resource search method provided in yet another embodiment of the present application;

[0050] Figure 8 A schematic diagram of the structure of another computing device provided for another embodiment of the present application. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0052] At present, in many scenarios, such as on proprietary clouds, a large number of pictures are often isolated and lack information such as titles and contexts. In these scenarios, image searches cannot be completed based on existing image search technologies. In order to improve the problems existing in the prior art, in some embodiments of the present application: image description information can be generated based on the image content data of the image. Therefore, in the absence of information such as titles and contexts in the scene where the image is located, the image description information can be analyzed from the content contained in the image itself. This ensures the feasibility of the image search operation in the above-mentioned specific scenarios, and the image description information generated thereby can more accurately characterize the image, thereby effectively improving the accuracy of searching for images.

[0053] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0054] Figure 1 A flowchart of an image search method provided in one embodiment of the present application. The image search method provided in this embodiment can be executed by an image search engine, which can be implemented as software or a combination of software and hardware, and can be set in a computing device.

[0055] like Figure 1 As shown, the method includes:

[0056] 100. Obtain search information;

[0057] 101. Based on picture description information of at least one picture, search for a target picture matching the search information in at least one picture; wherein the picture description information is generated based on picture content data;

[0058] 102. Output the target image.

[0059] The image search method provided in this embodiment can be applied to various image search scenarios, especially scenarios where the images are relatively isolated and lack information such as titles and contexts. Of course, in addition to this special scenario, this embodiment can also be used in scenarios with information such as titles and contexts, and this embodiment is not limited to this.

[0060] In step 100, search information may be obtained. The source of the search information may be an individual user, or may be other computing devices, which is not limited in this embodiment.

[0061] In this embodiment, there is at least one picture in the search range. The search range here can be all pictures on the private cloud, or a folder in the user terminal device, or of course, a directory on a server. This embodiment does not limit this.

[0062] In step 101, picture description information of at least one picture may be obtained, and a target picture matching the search information may be searched from the at least one picture.

[0063] The picture description information is generated based on the picture content data, and the picture description information is associated with the picture. The picture description information may include background, color, size, and the relationship between objects, etc., which is not limited in this embodiment.

[0064] Accordingly, when performing an image search, the image description information can be used as a search basis to find target images that match the search information.

[0065] In step 102, the target image may be output as a search result. According to different sources of search information, the target image may be returned to different search parties. In addition, the output form of the target image is not limited in this embodiment.

[0066] In this embodiment, the search information may be a search keyword. In this case, the embodiment may provide a solution for searching for pictures by text. In this embodiment, the search information may also be a picture. In this case, the embodiment may provide a solution for searching for pictures by pictures. Of course, in this embodiment, the search information may also be implemented in other forms, and the embodiment is not limited thereto.

[0067] In this embodiment, the picture description information can be generated based on the picture content data of the picture. Therefore, in the case where the scene where the picture is located lacks information such as title and context, the picture description information can be analyzed from the content contained in the picture itself. This ensures the feasibility of the picture search operation in the above-mentioned specific scene, and the picture description information generated thereby can more accurately represent the picture, thereby effectively improving the accuracy of searching for pictures. Of course, this embodiment can also be applied to the case where the scene where the picture is located has picture description information such as title and context, so as to replace or supplement the existing picture description information, thereby improving the accuracy of searching for pictures.

[0068] In the above or below embodiments, before executing step 100, picture description information may be generated in advance for at least one picture.

[0069] Since the process of generating picture description information for at least one picture is similar, for ease of description, the method of generating picture description information will be described below by taking the first picture as an example, wherein the first picture may be any one of the at least one picture.

[0070] In this embodiment, picture content data may be extracted from the first picture; and picture description information of the first picture may be generated based on the picture content data.

[0071] The picture content data refers to data that can reflect the content contained in the picture. In this embodiment, the association between objects contained in the picture can be used as the picture content data, but it is worth noting that this embodiment is not limited to this, and other content contained in the picture can also be used as the picture content data.

[0072] The following will take the association relationship between objects contained in the picture as the picture content data as an example to illustrate the solution for determining the picture content data.

[0073] In this embodiment, object information contained in the first picture can be detected; and based on the object information contained in the first picture, the association relationship between the objects contained in the first picture is analyzed as the picture content data of the first picture.

[0074] The object information includes the location of the object, the abstract features of the object, and the object identification, etc. The object identification can be a name identification, a category identification, etc., which is not limited in this embodiment. The abstract features of the object refer to the features that are abstracted from the image area where the object is located and can reflect the characteristics of the object.

[0075] In practical applications, a target detection model may be used to detect object information in the first image. The first image may be input into the target detection model, and the position, abstract features and identification of the object contained in the first image may be detected using the target detection model.

[0076] In this case, the abstract features of the object are carried by the feature map of the object. Of course, this embodiment does not limit this.

[0077] Among them, the target detection model can adopt Faster R-CNN model, SSD model, YOLO model or Retinanet model, etc.

[0078] Taking the Faster R-CNN model as an example, after the first image is input into the Faster R-CNN model, it is processed by the pre-trained CNN model in the Faster R-CNN model to obtain a convolutional feature map; select bounding box regions for each pixel; use the region generation network RPN (Region Propose Network) to find regions that may contain objects and their corresponding positions in the first image as the location of the object; classify the objects in the regions to generate object labels; for the aforementioned regions that may contain objects and the convolutional feature map obtained by processing the CNN model, use Rol Pooling processing, and extract the feature map of the relevant object as the abstract feature of the object. Based on this, the object identification, object location and abstract features of the object in the first image can be detected.

[0079] The process of using other target detection models to detect object information will not be described in detail here.

[0080] It is worth noting that, in this embodiment, the object information detection method is not limited to this, and other methods can be used to detect the object information in the first image. In addition, the object information is not limited to the above-mentioned object identification, object location and object abstract features, and the object information can also include other information.

[0081] In this embodiment, the association relationship between objects may include objects and relationships between objects, and the relationship between objects may be a geometric relationship, a possessive relationship, or a semantic relationship, etc. In practical applications, the association relationship between objects may be represented as a triplet of "object, object, relationship".

[0082] In practical applications, a picture analysis model may be used to analyze the association relationship between objects contained in the first picture.

[0083] In this embodiment, the object information contained in the first image can be input into the image analysis model; in the image analysis model, the association relationship between the objects contained in the first image is analyzed according to the object information contained in the first image, and is output as the image content data of the first image. In the above, the object information can include the location of the object, the abstract features of the object, and the object identification, etc.

[0084] In this embodiment, the image analysis model can be connected to the target detection model. The target detection model can input object information such as object identification, object location and object abstract features detected from the first image into the image analysis model. The image analysis model can analyze the relationship between objects contained in the first image based on the obtained object information.

[0085] As mentioned above, the relationship between objects output by the image analysis model can be expressed as a triplet "object, object, relationship". In addition, the object information output by the target detection model can also be expressed as a multi-tuple "object location, object abstract feature, object identification". For the first image, the object information output by the target detection model will be expressed as several multi-tuples "object location, object abstract feature, object identification", and the relationship between objects output by the image analysis model will be expressed as several triplets "object, object, relationship".

[0086] In this embodiment, the image analysis model may adopt an encoder-decoder network structure. Figure 2 A logical schematic diagram of an image analysis model provided for an embodiment of the present application, wherein the image analysis model adopts an encoder-decoder network structure.

[0087] like Figure 2 As shown, for the first picture, multiple multi-tuples corresponding to the first picture (object position B, object abstract feature F, object identifier L) can be used as the input of the picture analysis model, and the encoder part can use the RNN network, such as Figure 2 h1 and h2 in represent the hidden layers of the RNN network. The encoder part can encode these inputs (usually nonlinear transformation) to generate an intermediate semantic representation c1. The decoder part can also use an RNN network to generate the association relationship between the objects to be generated at the current moment based on the association relationship between the intermediate semantic representation c1 and the previously generated objects, where h'1 and h'2 represent the hidden layers of the RNN network.

[0088] Based on this, the image analysis model can output at least one triplet "object O, object O, relationship A" contained in the first image as the image content data of the first image.

[0089] In this embodiment, the image analysis model can also add an attention mechanism, which is similar to Figure 2 The difference between the logical diagram shown in the figure is that the intermediate semantic representation is no longer a fixed c1, but is replaced by c that changes according to the output at different times. i . And each c iIt is generated based on the attention weights of the multiple multi-tuples of the input image analysis model (object position, object abstract features, object identification) output at different times. Among them, for the output at time i, the attention weight can be obtained by the hidden layer node state h' at the previous time i. i-1 The hidden layer node state h corresponding to each multi-tuple in the input j Compare to determine the degree of alignment as the attention weight. Where i represents the number of output triplets and j represents the number of input multi-tuples.

[0090] Of course, the above-mentioned specific processing process is only exemplary, and this embodiment is not limited to these processing details. Each processing detail can be replaced by other processing methods, and this embodiment does not limit this.

[0091] Based on this, the association relationship between the objects contained in the first picture can be obtained as picture content data.

[0092] On this basis, in this embodiment, at least one feature information can be extracted from the picture content data of the first picture as the picture description information of the first picture. Continuing from the above, at least one feature information can be extracted from the association relationship between objects contained in the first picture as the picture description information of the first picture.

[0093] The feature information may be a keyword, or may be a picture region that can reflect the association relationship between objects contained in the first picture. Of course, the feature information in this embodiment is not limited thereto.

[0094] For example, if the association relationship between objects contained in the first picture is "person, horse, person riding a horse", then keywords such as person, horse, person riding a horse, riding, riding a horse, etc. can be extracted as feature information.

[0095] In this embodiment, the target detection model and the image analysis model cooperate with each other, the target detection model is used to extract object information from the image, and the image analysis model is used to analyze the object information in the image to generate the association relationship between the objects contained in the image as the image content data of the image; based on this, feature information can be extracted from the image content data as image description information. This allows the image description information to describe the image more accurately, thereby improving the accuracy of the search.

[0096] In the above or below embodiments, in order to realize the analysis function of the image analysis model, the image analysis model may be pre-trained.

[0097] In this embodiment, sample images may be obtained, and object identifiers, item locations, and associations between objects contained in the sample images may be annotated;

[0098] Use the target detection model to detect the abstract features of objects contained in the sample images;

[0099] The position, abstract features and identification of the objects in the sample images are input into the image analysis model as the object information contained in the sample images, and the image analysis model is trained based on the association relationship between the marked objects in the sample images.

[0100] The source of the sample images is not limited in this embodiment. For example, the sample images may use the public dataset Visual Genome dataset, which has a very dense and complete description of the images, provides detailed labels of the interactions and attributes of objects, and implements the visual concepts at the semantic level. The annotation information of the dataset includes object labels, object positions, and associations between objects on the image, etc. Therefore, in this embodiment, the images and annotation information in the dataset can be used to train the image analysis model.

[0101] In this embodiment, the target detection model can be used to detect the abstract features of the objects contained in the sample image, wherein the processing process of the target detection model can refer to the description in the previous text and will not be repeated here.

[0102] Based on this, the position of the object, the abstract features of the object, and the object identification of the sample image can be input into the image analysis model as the object information contained in the sample image, and the image analysis model will input the association relationship between the objects contained in the sample image. In this embodiment, the image analysis model can be corrected based on the error between the association relationship between the marked objects in the sample image and the result output by the image analysis model. This enables the image analysis model to learn the ability to analyze the association relationship between the objects contained in the image based on the object information such as the position of the object, the abstract features of the object, and the object identification of the image.

[0103] Based on this, the training of the image analysis model can be achieved.

[0104] In this embodiment, a large number of sample images can be used to train a highly accurate image analysis model, thereby ensuring the accuracy of the image description information generated for at least one image within the search range using the image analysis model, thereby improving the accuracy of the search.

[0105] In the above or below embodiments, a target picture matching the search information may be searched for in at least one picture based on the feature information included in the picture description information of at least one picture.

[0106] As mentioned above, the feature information may include keywords or image areas, and the search information may include keywords or images. The following will take the feature information as keywords and the search information as keywords as an example to explain the technical solution.

[0107] In this embodiment, at least one keyword may be extracted from the picture content data of at least one picture as the picture description information of the at least one picture.

[0108] In practical applications, word segmentation technology can be used to extract keywords from image content data. Thus, each image will correspond to at least one keyword.

[0109] In this embodiment, the matching degree between the search keyword and at least one keyword corresponding to each picture may be calculated, and pictures containing keywords whose matching degree meets preset requirements are used as target pictures.

[0110] In actual applications, a keyword index may be established for at least one keyword contained in at least one picture, and each keyword in the keyword index may be completely different.

[0111] The keyword index may be implemented in a matrix form, and a corresponding relationship between keywords and pictures is established in the matrix. For example, for a certain keyword, if the picture contains the keyword, 1 may be marked under the keyword in the matrix, and if the picture does not contain the keyword, 0 may be marked under the keyword in the matrix. Of course, this is only exemplary, and the present embodiment is not limited thereto.

[0112] Based on the keyword index, in this embodiment, the matching degree between the search keyword and at least one keyword in the keyword index can be calculated; and the picture under the keyword whose matching degree meets the preset requirement is determined as the target picture.

[0113] In this way, the matching calculation between the search keyword and the repeated keywords contained in different pictures can be avoided, thereby greatly reducing the amount of calculation and improving the search efficiency.

[0114] There are two implementations of this invention, which can convert the search keyword and at least one keyword in the keyword index into a first word vector and a second word vector respectively; calculate the similarity between the first word vector and the second word vector as the matching degree between the search keyword and at least one keyword in the keyword index.

[0115] The Word2Vec word vector technology, knowledge graph, synonym rewriting or other query similarity calculation technology can be used to calculate the matching degree between the search keyword and at least one keyword in the keyword index, which is not limited in this embodiment.

[0116] In one implementation, the target image may be determined by using a keyword precise matching method, that is, when the search keyword completely matches the keyword contained in the image, the image is determined to be the target image.

[0117] In another implementation, a matching threshold may be set, and when the matching degree between the search keyword and a certain keyword is greater than the matching threshold, the image under the keyword may be determined as the target image.

[0118] For example, when the search keyword is "car", if the matching degree between "sedan" and "car" is greater than the aforementioned matching degree threshold, the picture containing "car" will be determined as the target picture.

[0119] When the feature information is a picture region and the search information is a picture, the degree of match can be determined by comparing the two types of images, which will not be described in detail here.

[0120] Figure 3 A schematic diagram of a dedicated cloud scenario provided in an embodiment of the present application. Figure 3 The professional cloud scenario shown provides a detailed description of the image search method of the present application.

[0121] like Figure 3 As shown, the image search process can be divided into two stages: information preparation stage and actual search stage.

[0122] In the information preparation stage, the target detection model can be used to extract the object information of at least one picture on the private cloud, and the extracted object information can be input into the picture analysis model. In the picture analysis model, the association relationship between the objects contained in each of the at least one picture can be analyzed based on the object information of the at least one picture as the picture content data. Subsequently, at least one keyword can be extracted from the picture content data of at least one picture as the picture description information of each of the at least one picture. For example, the picture description information of a certain picture can include the keywords person, horse, person riding a horse, riding, horse riding, etc.

[0123] On the private cloud, at least one image and its image description information may be associated and saved.

[0124] During the actual search stage, the user can input the search keyword "horse riding" in the search terminal. The search terminal can generate a search request based on the search keyword and send it to the image search engine in this embodiment. Based on the triggering of the search keyword, in this embodiment, the image search engine can obtain the image description information of at least one image on the proprietary cloud, and can find the target image that matches the search keyword "horse riding" from at least one image by calculating the matching degree between the search keyword "horse riding" and at least one keyword contained in the image description information of at least one image.

[0125] Figure 4 A flowchart of a method for generating picture description information provided in another embodiment of the present application.

[0126] like Figure 4 As shown, the method includes:

[0127] 400. Get the image to be processed;

[0128] 401. Extracting picture content data from the picture to be processed;

[0129] 402. Generate picture description information of the picture to be processed based on the picture content data.

[0130] The method for generating picture description information provided in this embodiment can be applied to various scenarios where picture description information needs to be configured for pictures, for example, in the picture search scenario mentioned above, and of course it can also be applied to other scenarios, which is not limited in this embodiment.

[0131] In this embodiment, the image to be processed refers to the image that needs to be configured with image description information. The image to be processed can be the image on the proprietary cloud mentioned above. Of course, it can also be an image in other scenarios, and this embodiment does not limit this.

[0132] In this embodiment, picture content data may be extracted from the picture to be processed, wherein the picture content data refers to data that may reflect the content contained in the picture.

[0133] Based on the image content data of the image to be processed, image description information of the image to be processed can be generated. The image description information is associated with the image. The image description information may include background, color, size, and association between objects, etc., which is not limited in this embodiment.

[0134] In practical applications, object information in the image to be processed can be detected; based on the object information in the image to be processed, the association relationship between the objects contained in the image to be processed is analyzed as the image content data of the image to be processed. The object information may include the location of the object, the abstract features of the object, and the object identification, etc., but this embodiment is not limited thereto.

[0135] The object information extraction process and the relationship analysis process between objects can refer to the relevant description in the previous text, which will not be repeated here.

[0136] Based on the image content data analyzed from the image to be processed, in this embodiment, at least one keyword can be extracted from the image content data of the image to be processed as the image description information of the image to be processed.

[0137] In line with the above, in this embodiment, at least one keyword may be extracted from the association relationship between objects contained in the image to be processed as the image description information of the image to be processed.

[0138] In this embodiment, the image description information is generated based on the image content data of the image to be processed, which can more accurately describe the image to be processed. The image description information generated for the image to be processed can be used in the image search process mentioned above, and of course, can also be used in other processing processes, which is not limited in this embodiment.

[0139] In addition, it is worth noting that in order to reduce the length of the article, the technical details of the process of generating the picture description information are not described in detail in this embodiment. For these technical details, please refer to the relevant description of the process of generating the picture description information in the previous text. However, this should not cause loss of the protection scope of this application.

[0140] It should be noted that the execution subject of each step of the method provided in the above embodiment can be the same device, or the method can be executed by different devices. For example, the execution subject of steps 100 to 101 can be device A; for another example, the execution subject of steps 101 and 102 can be device A, and the execution subject of step 100 can be device B; and so on.

[0141] In addition, in some of the processes described in the above embodiments and the accompanying drawings, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel, and the sequence numbers of the operations, such as 100, 101, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0142] Figure 5 A schematic diagram of a computing device provided in another embodiment of the present application is shown below. Figure 5 As shown, the computing device includes: a memory 50 and a processor 51 .

[0143] The processor 51 is coupled to the memory 50 and is used to execute the computer program in the memory to:

[0144] Get search keywords;

[0145] Based on the picture description information of at least one picture, searching for a target picture matching the search keyword in at least one picture;

[0146] Wherein, the picture description information is generated based on the picture content data;

[0147] Output the target image.

[0148] In an optional embodiment, before searching for a target picture matching the search keyword in at least one picture based on the picture description information of at least one picture, the processor 51 is further configured to:

[0149] For the first picture, extracting picture content data from the first picture;

[0150] Generate picture description information of the first picture based on the picture content data;

[0151] The first picture is any one of the at least one picture.

[0152] In an optional embodiment, when extracting the picture content data from the first picture, the processor 51 is configured to:

[0153] Detecting object information contained in the first image, where the object information includes a position of the object, an abstract feature of the object, and an object identifier;

[0154] According to the positions of the objects, the abstract features of the objects and the object identifiers, the association relationship between the objects contained in the first picture is analyzed as the picture content data of the first picture.

[0155] In an optional embodiment, when detecting object information included in the first picture, the processor 51 is configured to:

[0156] Inputting the first image into the object detection model;

[0157] The target detection model is used to detect the position, abstract features and identification of the object contained in the first image as the object information of the first image.

[0158] In an optional embodiment, the target detection model adopts a Faster R-CNN model, an SSD model, a YOLO model or a Retinanet model.

[0159] In an optional embodiment, when the processor 51 analyzes the association relationship between objects included in the first picture according to the positions of the objects, the abstract features of the objects, and the object identifiers, it is configured to:

[0160] Inputting the position, abstract features and identification of the object contained in the first image into the image analysis model;

[0161] In the picture analysis model, the association relationship between objects contained in the first picture is analyzed according to the positions, abstract features and identifications of objects contained in the first picture, and the relationship is output as picture content data of the first picture.

[0162] In an optional embodiment, before inputting the position of the object, the abstract feature of the object, and the object identification contained in the first image into the image analysis model, the processor 51 is further configured to:

[0163] Obtain sample images and annotate the object identifiers, item locations, and relationships between objects contained in the sample images;

[0164] Use the target detection model to detect the abstract features of objects contained in the sample images;

[0165] The position, abstract features and identification of the objects in the sample images are input into the image analysis model as the object information contained in the sample images, and the image analysis model is trained based on the association relationship between the marked objects in the sample images.

[0166] In an optional embodiment, an attention mechanism is included in the image analysis model.

[0167] In an optional embodiment, when the processor 51 generates the picture description information of the first picture based on the picture content data, it is configured to:

[0168] At least one keyword is extracted from the association relationship between objects included in the first picture as picture description information of the first picture.

[0169] In an optional embodiment, the processor 51 is further configured to:

[0170] Creating a keyword index for at least one keyword contained in at least one image, wherein the keywords in the keyword index are completely different;

[0171] Based on the picture description information of at least one picture, searching for a target picture matching the search keyword in at least one picture includes:

[0172] Calculating a match between the search keyword and at least one keyword in the keyword index;

[0173] The images under the keywords whose matching degree meets the preset requirements are determined as target images.

[0174] In an optional embodiment, when calculating the matching degree between the search keyword and at least one keyword in the keyword index, the processor 51 is configured to:

[0175] Converting the search keyword and at least one keyword in the keyword index into a first word vector and a second word vector respectively;

[0176] The similarity between the first word vector and the second word vector is calculated as the matching degree between the search keyword and at least one keyword in the keyword index.

[0177] Further, if Figure 5 As shown, the computing device also includes: a communication component 52, a power supply component 53 and other components. Figure 5Only some components are shown schematically, and it does not mean that the computing device only includes Figure 5 Components shown.

[0178] It is worth noting that the technical details in the above-mentioned embodiments of the computing device can refer to the technical details recorded in the above-mentioned embodiments related to the image search method. In order to save space, they will not be repeated here, but this should not cause loss of the protection scope of this application.

[0179] Accordingly, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, which, when executed, can implement the steps that can be executed by a computing device in the above method embodiment.

[0180] Figure 6 A schematic diagram of the structure of another computing device provided in yet another embodiment of the present application. Figure 6 As shown, the computing device includes: a memory 60 and a processor 61 .

[0181] The processor 61 is coupled to the memory 60 and is used to execute the computer program in the memory to:

[0182] Get the image to be processed;

[0183] Extracting picture content data from the picture to be processed;

[0184] Based on the image content data, generate image description information of the image to be processed.

[0185] In an optional embodiment, when extracting picture content data from the picture to be processed, the processor 61 is used to:

[0186] Detecting object information contained in the first image, where the object information includes a position of the object, an abstract feature of the object, and an object identifier;

[0187] According to the positions of the objects, the abstract features of the objects and the object identifiers, the association relationship between the objects contained in the first picture is analyzed as the picture content data of the first picture.

[0188] In an optional embodiment, when the processor 61 generates the picture description information of the picture to be processed based on the picture content data, it is used to:

[0189] At least one keyword is extracted from the association relationship between objects contained in the image to be processed as image description information of the image to be processed.

[0190] Further, if Figure 6 As shown, the computing device also includes: a communication component 62, a power supply component 63 and other components. Figure 6 Only some components are shown schematically, and it does not mean that the computing device only includes Figure 6Components shown.

[0191] It is worth noting that the technical details in the above-mentioned embodiments of the computing device can refer to the technical details recorded in the above-mentioned embodiments related to the image description information generation method. In order to save space, they will not be repeated here, but this should not cause loss of protection scope of this application.

[0192] Accordingly, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, which, when executed, can implement the steps that can be executed by a computing device in the above method embodiment.

[0193] Figure 7 A schematic diagram of a resource search method provided in another embodiment of the present application. Figure 7 As shown, the method includes:

[0194] 700. Obtain search information;

[0195] 701. Based on description information of at least one resource, search for a target resource matching the search information in at least one resource;

[0196] Wherein, the description information is generated based on the resource content data;

[0197] 703. Output the target resource.

[0198] The resource search method provided in this embodiment can be applied to various resource search scenarios, especially scenarios where the resources are relatively isolated and lack information such as titles and contexts. Of course, in addition to this special scenario, this embodiment can also be used in scenarios with information such as titles and contexts, and this embodiment is not limited to this.

[0199] In this embodiment, the resources may be pictures, documents, videos, etc., and the storage location of the resources may be a private cloud or a user's personal computer, etc., which is not limited in this embodiment.

[0200] In an optional embodiment, the step, based on the description information of at least one resource, before searching for a target resource matching the search information in at least one resource, further includes:

[0201] For the first resource, extract resource content data from the first resource;

[0202] Generate description information of the first resource based on the resource content data;

[0203] The first resource is any one of the at least one resource.

[0204] In an optional embodiment, the step of extracting resource content data from the first resource includes:

[0205] Detecting object information contained in the first resource, where the object information includes a location of the object, an abstract feature of the object, and an object identifier;

[0206] According to the positions, abstract features and identifiers of the objects, the association relationships among the objects included in the first resource are analyzed as resource content data of the first resource.

[0207] In actual applications, for resources such as pictures and videos, the object may be a person in the resource. For document resources, the object may be a word, formula, etc. in the document. Different types of objects may be concerned with different resources. This embodiment does not limit this.

[0208] In an optional embodiment, the step of detecting the object information contained in the first resource includes:

[0209] Inputting the first resource into the target detection model;

[0210] The target detection model is used to detect the position, abstract features and identification of the object contained in the first resource as the object information of the first resource.

[0211] In an optional embodiment, the target detection model adopts a Faster R-CNN model, an SSD model, a YOLO model or a Retinanet model.

[0212] In an optional embodiment, the step of analyzing the association relationship between objects included in the first resource according to the location of the object, the object abstract feature and the object identifier includes:

[0213] Inputting the location, object abstract features and object identifier of the object contained in the first resource into the resource analysis model;

[0214] In the resource analysis model, the association relationship between objects contained in the first resource is analyzed according to the location, object abstract characteristics and object identification of the objects contained in the first resource, and is output as the resource content data of the first resource.

[0215] In an optional embodiment, before the step of inputting the location, object abstract features and object identifier of the object contained in the first resource into the resource analysis model, the step further includes:

[0216] Obtain sample resources, and annotate the object identifiers, item locations, and relationships between objects contained in the sample resources;

[0217] Utilize the object detection model to detect the abstract features of objects contained in the sample resources;

[0218] The location, abstract features and identification of the objects in the sample resources are input into the resource analysis model as the object information contained in the sample resources, and the resource analysis model is trained based on the association relationship between the marked objects in the sample resources.

[0219] In an optional embodiment, an attention mechanism is included in the resource analysis model.

[0220] In an optional embodiment, the step of generating description information of the first resource based on the resource content data includes:

[0221] At least one feature information is extracted from the association relationship between objects contained in the first resource as description information of the first resource.

[0222] Among them, for different resources, the implementation form of feature information may be different. For resources such as pictures or videos, feature information can be implemented as keywords or picture areas that reflect the association relationship between objects, and for resources such as documents, feature information can also be keywords.

[0223] In an optional embodiment, the method further comprises:

[0224] Establishing an information index for at least one characteristic information contained in at least one resource, wherein each characteristic information in the information index is completely different;

[0225] The step of searching for a target resource matching the search information in at least one resource based on the description information of at least one resource comprises:

[0226] Calculating a matching degree between the search information and at least one feature information in the information index;

[0227] The resources with characteristic information whose matching degree meets the preset requirements are determined as target resources.

[0228] In an optional embodiment, the step of calculating the matching degree between the search information and at least one feature information in the information index includes:

[0229] Converting at least one feature information in the search information and the information index into a first vector and a second vector respectively;

[0230] The similarity between the first vector and the second vector is calculated as the matching degree between the search information and at least one feature information in the information index.

[0231] In an optional embodiment, the feature information includes keywords, and the search information includes search keywords.

[0232] In an optional embodiment, at least one resource is stored on a private cloud.

[0233] It is worth noting that resource is a higher-level concept of picture. Based on this, the technical details in the above-mentioned embodiments of the resource search method can refer to the technical details recorded in the above-mentioned embodiments related to the picture search method. In order to save space, they will not be repeated here, but this should not cause loss of protection scope of this application.

[0234] Figure 8 A schematic diagram of the structure of another computing device provided in another embodiment of the present application. Figure 8 As shown, the computing device includes: a memory 80 and a processor 81 .

[0235] The processor 81 is coupled to the memory 80 and is used to execute the computer program in the memory to:

[0236] Get search information;

[0237] Based on the description information of at least one resource, searching for a target resource matching the search information in at least one resource;

[0238] Wherein, the description information is generated based on the resource content data;

[0239] Output the target resource.

[0240] The computing device provided in this embodiment can be applied to various resource search scenarios, especially scenarios where the resources are relatively isolated and lack information such as titles and contexts. Of course, in addition to this special scenario, this embodiment can also be used in scenarios with information such as titles and contexts, and this embodiment is not limited to this.

[0241] In this embodiment, the resources may be pictures, documents, videos, etc., and the storage location of the resources may be a private cloud or a user's personal computer, etc., which is not limited in this embodiment.

[0242] In an optional embodiment, before searching for a target resource matching the search information in at least one resource based on the description information of at least one resource, the processor 81 is further configured to:

[0243] For the first resource, extract resource content data from the first resource;

[0244] Generate description information of the first resource based on the resource content data;

[0245] The first resource is any one of the at least one resource.

[0246] In an optional embodiment, when extracting resource content data from the first resource, the processor 81 is configured to:

[0247] Detecting object information contained in the first resource, where the object information includes a location of the object, an abstract feature of the object, and an object identifier;

[0248] According to the positions, abstract features and identifiers of the objects, the association relationships among the objects included in the first resource are analyzed as resource content data of the first resource.

[0249] In actual applications, for resources such as pictures and videos, the object may be a person in the resource. For document resources, the object may be a word, formula, etc. in the document. Different types of objects may be concerned with different resources. This embodiment does not limit this.

[0250] In an optional embodiment, when detecting the object information included in the first resource, the processor 81 is configured to:

[0251] Inputting the first resource into the target detection model;

[0252] The target detection model is used to detect the position, abstract features and identification of the object contained in the first resource as the object information of the first resource.

[0253] In an optional embodiment, the processor 81 adopts the Faster R-CNN model, the SSD model, the YOLO model or the Retinanet model in the target detection model.

[0254] In an optional embodiment, when the processor 81 analyzes the association relationship between objects included in the first resource according to the location of the object, the object abstract feature and the object identifier, it is configured to:

[0255] Inputting the location, object abstract features and object identifier of the object contained in the first resource into the resource analysis model;

[0256] In the resource analysis model, the association relationship between objects contained in the first resource is analyzed according to the location, object abstract characteristics and object identification of the objects contained in the first resource, and is output as the resource content data of the first resource.

[0257] In an optional embodiment, before inputting the location, object abstract features, and object identifier of the object included in the first resource into the resource analysis model, the processor 81 is further configured to:

[0258] Obtain sample resources, and annotate the object identifiers, item locations, and relationships between objects contained in the sample resources;

[0259] Utilize the object detection model to detect the abstract features of objects contained in the sample resources;

[0260] The location, abstract features and identification of the objects in the sample resources are input into the resource analysis model as the object information contained in the sample resources, and the resource analysis model is trained based on the association relationship between the marked objects in the sample resources.

[0261] In an optional embodiment, an attention mechanism is included in the resource analysis model.

[0262] In an optional embodiment, when the processor 81 generates the description information of the first resource based on the resource content data, it is configured to:

[0263] At least one feature information is extracted from the association relationship between objects contained in the first resource as description information of the first resource.

[0264] Among them, for different resources, the implementation form of feature information may be different. For resources such as pictures or videos, feature information can be implemented as keywords or picture areas that reflect the association relationship between objects, and for resources such as documents, feature information can also be keywords.

[0265] In an optional embodiment, the processor 81 is further configured to:

[0266] Establishing an information index for at least one characteristic information contained in at least one resource, wherein each characteristic information in the information index is completely different;

[0267] When searching for a target resource matching the search information in at least one resource based on the description information of at least one resource, the processor 81 is configured to:

[0268] Calculating a matching degree between the search information and at least one feature information in the information index;

[0269] The resources with characteristic information whose matching degree meets the preset requirements are determined as target resources.

[0270] In an optional embodiment, when calculating the matching degree between the search information and at least one feature information in the information index, the processor 81 is configured to:

[0271] Converting at least one feature information in the search information and the information index into a first vector and a second vector respectively;

[0272] The similarity between the first vector and the second vector is calculated as the matching degree between the search information and at least one feature information in the information index.

[0273] In an optional embodiment, the feature information includes keywords, and the search information includes search keywords.

[0274] In an optional embodiment, at least one resource is stored on a private cloud.

[0275] Further, if Figure 8 As shown, the computing device also includes: a communication component 82, a power supply component 83 and other components. Figure 8 Only some components are shown schematically, and it does not mean that the computing device only includes Figure 8 Components shown.

[0276] It is worth noting that the technical details in the above-mentioned embodiments of the computing device can refer to the technical details recorded in the above-mentioned resource search method related embodiments. In order to save space, they will not be repeated here, but this should not cause loss of protection scope of this application.

[0277] Accordingly, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, which, when executed, can implement the steps that can be executed by a computing device in the above method embodiment.

[0278] in, Figure 5 , 6 The memory in 8 is used to store computer programs and can be configured to store various other data to support operations on the computing device. Examples of such data include instructions for any application or method operating on the computing device, contact data, phone book data, messages, pictures, videos, etc. The memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0279] in, Figure 5 , 6 The communication component in 8 is configured to facilitate wired or wireless communication between the device where the communication component is located and other devices. The device where the communication component is located can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component can be implemented based on near field communication (NFC) technology, radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology or other technologies to facilitate short-range communication.

[0280] in, Figure 5 , 6The power supply assembly in 8 provides power to various components of the device where the power supply assembly is located. The power supply assembly may include a power management system, one or more power supplies, and other components associated with generating, managing and distributing power to the device where the power supply assembly is located.

[0281] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0282] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0283] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0284] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0285] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0286] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0287] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0288] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0289] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for searching an image, characterized in that: include: Extracting picture content data from a first picture, where the first picture is any one of at least one picture; Extracting at least one feature information from the picture content data of the first picture, wherein the feature information adopts a picture area that can reflect the association relationship between objects contained in the first picture; Creating an information index for at least one feature information contained in each of the at least one picture; Get search information; Convert the search information and at least one feature information in the information index into a first vector and a second vector respectively, and calculate the similarity between the first vector and each second vector as the matching degree between the search information and the at least one feature information in the information index; The image with feature information whose matching degree meets the preset requirements is determined as the target image; Output the target image.

2. The method according to claim 1, characterized in that The step of extracting picture content data from the first picture includes: Detecting object information contained in the first image, where the object information includes a position of the object, an abstract feature of the object, and an object identifier; According to the positions, abstract features and identifiers of the objects, the association relationships between the objects contained in the first picture are analyzed as picture content data of the first picture.

3. The method according to claim 2, characterized in that The detecting the object information contained in the first picture includes: Inputting the first image into a target detection model; The target detection model is used to detect the position, abstract features and identification of the object contained in the first image as the object information of the first image.

4. The method according to claim 3, characterized in that The target detection model adopts a Faster R-CNN model, an SSD model, a YOLO model or a Retinanet model.

5. The method according to claim 2, characterized in that: The analyzing the association relationship between the objects included in the first picture according to the positions, abstract features and identifiers of the objects includes: Inputting the position, abstract features and identification of the object contained in the first image into the image analysis model; In the picture analysis model, the association relationship between objects contained in the first picture is analyzed based on the positions, abstract features and identifications of objects contained in the first picture, and is output as picture content data of the first picture.

6. The method according to claim 5, characterized in that Before inputting the position of the object, the abstract feature of the object and the object identification contained in the first picture into the picture analysis model, the method further includes: Obtain sample images, and mark the object identifiers, item locations, and relationships between objects contained in the sample images; Detecting abstract features of objects contained in the sample images using a target detection model; The position, abstract features and identification of the object in the sample image are input into the image analysis model as the object information contained in the sample image, and the image analysis model is trained based on the association relationship between the marked objects in the sample image.

7. The method according to claim 5 or 6, characterized in that: The image analysis model includes an attention mechanism.

8. The method according to claim 2, characterized in that: The extracting at least one feature information from the picture content data of the first picture includes: At least one feature information is extracted from the association relationship between objects included in the first picture.

9. The method according to claim 1, characterized in that: Each piece of feature information in the information index is completely different.

10. The method according to claim 8, characterized in that The search information includes search keywords.

11. The method according to claim 1, characterized in that: The at least one picture is stored in a dedicated cloud.

12. A resource search method, characterized in that: include: Extracting resource content data from a first resource, where the first resource is any one of at least one resource; Extracting at least one feature information from the resource content data of the first resource, wherein the feature information is a picture area that can reflect the association relationship between objects contained in the first resource; Establishing an information index for at least one characteristic information contained in each of the at least one resource; Get search information; Converting the search information and at least one feature information in the information index into a first vector and a second vector respectively; Calculating the similarity between the first vector and each second vector as the matching degree between the search information and at least one feature information in the information index; Determine the resources with characteristic information that meet the preset requirements as target resources; The target resource is output.

13. A method for generating picture description information, characterized in that: include: Get the image to be processed; Extracting picture content data from the picture to be processed; Extracting at least one feature information from the picture content data as picture description information of the picture to be processed, wherein the feature information adopts a picture region that can reflect the association relationship between objects contained in the first picture; The characteristic information is used as a search basis in the image search method described in any one of claims 1-11.

14. The method according to claim 13, characterized in that The step of extracting picture content data from the picture to be processed includes: Detecting object information contained in the image to be processed, wherein the object information includes the position of the object, the abstract features of the object, and the object identification; According to the positions, abstract features and identifications of the objects, the association relationships among the objects contained in the image to be processed are analyzed as image content data of the image to be processed.

15. The method according to claim 13, characterized in that The step of extracting at least one feature information from the picture content data as picture description information of the picture to be processed includes: At least one keyword is extracted from the association relationship between objects contained in the image to be processed as image description information of the image to be processed.

16. A computing device, characterized in that: including memory and processor; The memory is used to store one or more computer instructions; The processor is coupled to the memory and configured to execute the one or more computer instructions for: Extracting picture content data from a first picture, where the first picture is any one of at least one picture; Extracting at least one feature information from the picture content data of the first picture, wherein the feature information adopts a picture area that can reflect the association relationship between objects contained in the first picture; Creating an information index for at least one feature information contained in each of the at least one picture; Get search information; Converting the search information and at least one feature information in the information index into a first vector and a second vector respectively; Calculating the similarity between the first vector and each second vector as the matching degree between the search information and at least one feature information in the information index; The image with feature information whose matching degree meets the preset requirements is determined as the target image; Output the target image.

17. A computing device, characterized in that including memory and processor; The memory is used to store one or more computer instructions; The processor is coupled to the memory and configured to execute the one or more computer instructions for: Extracting resource content data from a first resource, where the first resource is any one of at least one resource; Extracting at least one feature information from the resource content data of the first resource, wherein the feature information is a picture area that can reflect the association relationship between objects contained in the first resource; Establishing an information index for at least one characteristic information contained in each of the at least one resource; Get search information; Converting the search information and at least one feature information in the information index into a first vector and a second vector respectively; Calculating the similarity between the first vector and each second vector as the matching degree between the search information and at least one feature information in the information index; Determine the resources with characteristic information that meet the preset requirements as target resources; The target resource is output.

18. A computing device, characterized in that including memory and processor; The memory is used to store one or more computer instructions; The processor is coupled to the memory and configured to execute the one or more computer instructions for: Obtaining a picture to be processed; extracting picture content data from the picture to be processed; Extracting at least one feature information from the picture content data as picture description information of the picture to be processed, wherein the feature information adopts a picture region that can reflect the association relationship between objects contained in the first picture; The feature information is used as a search basis when the computing device executes the image search method described in any one of claims 1-11.

19. A computer-readable storage medium storing computer instructions, characterized in that: When the computer instructions are executed by one or more processors, the one or more processors are caused to execute the image search method described in any one of claims 1-11, the resource search method described in claim 12, or the image description information generation method described in any one of claims 13-15.

Images