A method for image classification

Abstract

The present application discloses a method for image classification, which is applied to the cloud. The method includes: receiving an uploaded image collection in response to a user uploading an image through multiple devices, wherein the uploaded image collection includes multiple images uploaded by the user; analyzing and identifying each image in the uploaded image collection to obtain an analysis and identification result corresponding to the image; using the analysis and identification result as a label for the image to obtain a labeled image; determining a preset image classification model that matches the label information based on the label information in the labeled image, wherein the image classification model is trained with training image data corresponding to the label information; finally, calling the image classification model to classify the labeled image, and the labeled image is each image in the uploaded image collection. Obviously, the present application can organize electronic images in multiple devices.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and more specifically, to a method for image classification. Background Art

[0002] With the popularity of smart phones and digital cameras and the improvement of people's living standards, more and more people own multiple mobile phones at the same time, and more and more people use their mobile phones to take pictures to record the beautiful moments of life. Electronic photography has become more and more common.

[0003] In recent years, with the continuous improvement of camera manufacturing technology, the cameras equipped in various smart devices have become increasingly sophisticated, even equipped with high-definition cameras. This has further stimulated users' desire to take photos. A user can also take photos through multiple devices, resulting in a surge in the number of electronic images. How to organize electronic images from multiple devices has become a problem that people are concerned about. Summary of the Invention

[0004] In view of this, the present application provides an image classification method for helping to organize electronic images in multiple devices.

[0005] In order to achieve the above objectives, the following solutions are proposed:

[0006] A method for image classification, comprising:

[0007] In response to an operation of a user uploading pictures through multiple devices, receiving an uploaded picture collection, wherein the uploaded picture collection includes multiple pictures uploaded by the user;

[0008] Analyze and identify each image in the uploaded image set to obtain an analysis and identification result corresponding to the image;

[0009] Using the analysis and recognition result as a label of the image to obtain a labeled image;

[0010] Determining, based on the label information in the labeled image, an image classification model that matches the label information, where the image classification model is trained using training image data corresponding to the label information;

[0011] The image classification model is called to classify the labeled images.

[0012] Optionally, before selecting a corresponding preset image classification model based on the labeled image, the method further includes:

[0013] Obtaining user attributes corresponding to the user;

[0014] Determining, based on the label information in the labeled image, a preset image classification model that matches the label information, including:

[0015] A matching image classification model is selected based on the label information in the labeled image and with reference to the user attributes.

[0016] Optionally, selecting a matching image classification model based on the tag information in the tagged image and referring to the user attributes includes:

[0017] If the labeled image includes multiple types of label information, determining an image classification model corresponding to each of the multiple types of label information;

[0018] The image classification models corresponding to the label information in the labeled images are combined into a first model set;

[0019] Determining a second model set corresponding to the user attribute, the second model set including at least one image classification model;

[0020] The picture classification model that exists in both the first model set and the second model set is used as the matching picture classification model.

[0021] Optionally, selecting a matching image classification model based on the tag information in the tagged image and referring to the user attributes includes:

[0022] If the labeled image includes multiple types of label information, the label information in the labeled image is sorted according to the label sorting rule corresponding to the user attribute to obtain a sorting result;

[0023] The image classification model corresponding to the first label information in the sorting result is selected as the final matching image classification model.

[0024] Optionally, determining an image classification model that matches the label information based on the label information in the labeled image includes:

[0025] If the labels in the labeled image include location labels and date labels, select a holiday strategy model that matches them;

[0026] The holiday policy model is trained using pictures marked with location information and date information as training data.

[0027] Optionally, determining, based on the label information in the labeled image, an image classification model that matches the label information includes:

[0028] If the labels in the labeled image include face labels, location labels, and date labels, selecting a face recognition model that matches them;

[0029] The face recognition model is trained using pictures annotated with face information, location information and date information as training data.

[0030] Optionally, determining, based on the label information in the labeled image, an image classification model that matches the label information includes:

[0031] If the labels in the labeled image include an item label, a location label, and a date label, selecting an item recognition model that matches them;

[0032] The object recognition model is trained using pictures marked with object information, location information and date information as training data.

[0033] Optionally, determining, based on the label information in the labeled image, an image classification model that matches the label information includes:

[0034] If the label in the labeled image includes a photography label, selecting a photography recognition model that matches the label, wherein the photography label includes any one of the following: a label of the device that captured the image, a photography parameter label, a location label, and a scene label;

[0035] The photographic recognition model is trained using pictures annotated with photographic information as training data.

[0036] Optionally, before analyzing and identifying the content of each picture in the uploaded picture set, the method further includes:

[0037] According to the upload method of the uploaded picture set, determine whether each picture in the uploaded picture set needs to be used, and classify according to user-defined rules;

[0038] If the judgment result is no, return to the step of performing analysis and identification on each picture in the uploaded picture set.

[0039] If the judgment result is yes, the image is classified according to the user-defined rules.

[0040] Optionally, in response to the user uploading an image through multiple devices, receiving an uploaded image collection, the uploaded image collection containing multiple images uploaded by the user includes:

[0041] In response to different users of the same family uploading pictures to a family shared directory via multiple devices, receiving an uploaded picture collection, the uploaded picture collection including multiple pictures uploaded by the different users;

[0042] After calling the image classification model to classify the labeled images, the method further includes:

[0043] Obtain the classification results of the labeled pictures, and store the classification results in a home sharing directory.

[0044] From the above technical solution, it can be seen that the picture classification method provided by the embodiment of the present application is applied to the cloud, and by responding to the user's operation of uploading pictures through multiple devices, an uploaded picture collection is received, and the uploaded picture collection contains multiple pictures uploaded by the user. In this way, the electronic pictures that need to be classified in multiple devices can be obtained; then, each picture in the uploaded picture collection can be analyzed and identified to obtain the analysis and identification results corresponding to the picture, so that the information contained in each picture in the uploaded picture collection can be obtained; based on this, the analysis and identification results can be used as the label of the picture to obtain a labeled picture, so that the label information contained in each picture in the uploaded picture collection can be obtained to help select a picture classification model; then, according to the label information in the labeled picture, a preset picture classification model that matches the label information can be determined, and the picture classification model is trained with the training picture data corresponding to the label information; finally, the picture classification model is called to classify the labeled pictures, and the labeled pictures are each picture in the uploaded picture collection. Obviously, the present application can organize electronic pictures in multiple devices.

[0045] In addition, the present application can analyze and identify each picture in the uploaded picture collection, obtain various types of information contained in the picture, and use this as the label of the picture to label the picture. Based on this, the picture classification model can be selected according to the label information in the labeled picture. Since the picture classification model is trained with the training picture data corresponding to the label information, different label information corresponds to different picture classification models, and different picture classification models correspond to different training data. Based on this, the present application can select a matching picture classification model for classification, and thus can better classify each picture. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without any creative work.

[0047] Figure 1This is a flow chart of a picture classification method disclosed in this application;

[0048] Figure 2 This is a flowchart of another image classification method disclosed in this application. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0050] Next, combine Figure 1 The image classification method of this application is introduced in detail, including the following steps:

[0051] Step S110 : In response to the user's operation of uploading pictures through multiple devices, an uploaded picture collection is received, where the uploaded picture collection includes multiple pictures uploaded by the user.

[0052] Specifically, the user may be one user or multiple users, that is, the user may respond to an operation of uploading pictures by the same user through multiple devices, or may respond to an operation of uploading pictures by multiple users through multiple devices.

[0053] This application is applied to the cloud and can support access by multiple users' devices and terminal devices of different systems.

[0054] Users can pre-set multiple devices as usage devices, record the device number corresponding to each of the multiple devices, and establish a corresponding relationship between the device number and the user. Based on this, for each device number that has a corresponding relationship with the user, the pictures uploaded by the matching device can be considered as multiple pictures uploaded by the user. These multiple pictures constitute the uploaded picture collection, and each device that matches the device number can view the uploaded picture collection. Similarly, users can also manage the uploaded picture collection or the album in the cloud through the device that matches the device number.

[0055] The terminal can upload the uploaded picture collection to the cloud in a variety of ways. For example, the terminal can choose to access the Internet via a network cable, or wirelessly access a mobile phone network, the Internet of Things, or local WIFI, and upload the uploaded picture collection to the cloud via the network. The uploaded picture collection can also be transmitted via the DCI service. The DCI service is a multi-cloud network connection service provided for enterprises, cloud service providers (CSPs), data center service providers (IDCs), and managed service providers (MSPs).

[0056] Specifically, the user's image files can be uploaded or downloaded on the edge side through the TCP / IP protocol and stored in the local edge for use by local users or cloud storage; the storage services originally deployed on the cloud are deployed on the edge, which can reduce network latency and increase device access speed.

[0057] Step S120: Analyze and identify each image in the uploaded image set to obtain an analysis and identification result corresponding to the image.

[0058] Specifically, each item of information corresponding to the picture can be analyzed and identified, for example, the corresponding device number, GPS positioning information, aperture information used to take the picture and other photography parameter information and picture formats such as jpg, png, gif, etc. It can also analyze and identify the content contained in each picture in the uploaded picture collection, for example, facial information, object information, landmark information, etc. in the picture.

[0059] Step S130: Use the analysis and recognition result as a label of the image to obtain a labeled image.

[0060] Specifically, the analysis and recognition results corresponding to each image can be used as the label of the image to label the image.

[0061] Taking the target picture as an example, if the analysis and recognition results of the target picture in step S120 are "2021 / 10 / 19" and "jpg", based on this, the "2021 / 10 / 19" and "jpg" can be used as labels of the target picture.

[0062] Step S140: Determine an image classification model that matches the label information based on the label information in the labeled image.

[0063] Specifically, the image classification model is trained using the training image data corresponding to the label information. In other words, different label information corresponds to different image classification models, and different image classification models correspond to different training data.

[0064] This application can select an image classification model corresponding to the label information in different labeled images to classify them.

[0065] Step S150: Call the image classification model to classify the labeled images.

[0066] Specifically, since the image classification model is trained using training image data corresponding to the label information, the image classification model can analyze and identify data information corresponding to the label information in the labeled images, and classify the labeled images based on the obtained data information. After the uploaded image collection is classified, at least one album can be obtained.

[0067] Based on this, the album may also be named, and the name of the album may be composed of the picture classification model and the data information.

[0068] If the image classification model is unable to classify the labeled image, this application can put it into an album to be edited and wait for the user to classify it by himself.

[0069] From the above technical solution, it can be seen that the picture classification method provided by the embodiment of the present application is applied to the cloud, and by responding to the user's operation of uploading pictures through multiple devices, an uploaded picture collection is received, and the uploaded picture collection contains multiple pictures uploaded by the user. In this way, the electronic pictures that need to be classified in multiple devices can be obtained; then, each picture in the uploaded picture collection can be analyzed and identified to obtain the analysis and identification results corresponding to the picture, so that the information contained in each picture in the uploaded picture collection can be obtained; based on this, the analysis and identification results can be used as the label of the picture to obtain a labeled picture, so that the label information contained in each picture in the uploaded picture collection can be obtained to help select a picture classification model; then, according to the label information in the labeled picture, a preset picture classification model that matches the label information can be determined, and the picture classification model is trained with the training picture data corresponding to the label information; finally, the picture classification model is called to classify the labeled pictures, and the labeled pictures are each picture in the uploaded picture collection. Obviously, the present application can organize electronic pictures in multiple devices.

[0070] In addition, the present application can analyze and identify each image in the uploaded image collection, obtain various types of information contained in the image, and use this as the label of the image to label the image. Based on this, the image classification model can be selected according to the label information in the labeled image. Since the image classification model is trained with the training image data corresponding to the label information, different label information corresponds to different image classification models, and different image classification models correspond to different training data. Based on this, the present application can select a matching image classification model for classification, and thus can better classify each image.

[0071] In some embodiments of the present application, step S110, responding to the user's operation of uploading pictures through multiple devices, receiving an uploaded picture collection, wherein the uploaded picture collection includes multiple pictures uploaded by the user, is described in detail.

[0072] Specifically, in response to operations in which different users of the same family upload pictures to a family shared directory through multiple devices, an uploaded picture collection may be received, where the uploaded picture collection includes multiple pictures uploaded by different users.

[0073] The device number corresponding to each user in the same family can be stored in the storage memory corresponding to the family. When the user uploads a picture using the device corresponding to the device number, the picture uploaded by the user can be received.

[0074] On this basis, in step S150, after calling the picture classification model and classifying the labeled pictures, the classification results of the labeled pictures can be obtained, and the classified results are stored in the family sharing directory.

[0075] Specifically, any member of the family can view the photo albums in the family shared directory. After the classified results are stored in the family shared directory, any member can view each picture in the uploaded picture collection.

[0076] As can be seen from the above scheme, this embodiment adds a new implementation scenario compared to the previous embodiment, namely, different users in the same family upload pictures to the family shared directory, and the classification results are stored in the family shared directory. Through the above steps, this application can classify and store pictures uploaded by different users in the same family to the family shared directory, allowing family members to view the classified pictures.

[0077] Reference Figure 2 , which discloses another flowchart of an image classification method.

[0078] In some embodiments of the present application, Figure 2 The image classification method of this application is introduced, which includes the following steps:

[0079] Step S210: In response to the user's operation of uploading pictures through multiple devices, an uploaded picture collection is received, where the uploaded picture collection includes multiple pictures uploaded by the user.

[0080] Step S220: Analyze and identify each image in the uploaded image set to obtain an analysis and identification result corresponding to the image.

[0081] Step S230: Use the analysis and recognition result as a label of the image to obtain a labeled image.

[0082] The above steps S210-S230 correspond one-to-one to steps S110-S130 in the aforementioned embodiment. Please refer to the above introduction for details and will not be repeated here.

[0083] Step S240: Obtain user attributes corresponding to the user.

[0084] Specifically, the present application is applied to the cloud, in which user attributes can be stored. If the user is a new user, the user attributes of the external account can be referenced with the user's authorization.

[0085] In some embodiments, the user attributes may be updated in real time with reference to the total amount of pictures uploaded by the user, the tag information of the pictures, and the total categories of the albums.

[0086] Step S250: Select a matching image classification model based on the label information in the labeled image and with reference to the user attributes.

[0087] Specifically, the tag information in the tagged image and the user attributes corresponding to the user who uploaded the image can be considered at the same time, and an image classification model that matches the tag information and the user attributes can be selected to facilitate classification in subsequent steps.

[0088] The user attributes include various types, such as foodie attributes, family attributes, party attributes, foodie attributes, and photography attributes.

[0089] The user attribute is analyzed based on the number of pictures uploaded by the user. For example, if 70% of the pictures uploaded by the user are of restaurants and the majority are pictures of dishes, then the user attribute of the user can be defined as a foodie.

[0090] Step S260: Call the image classification model to classify the labeled images.

[0091] This step S260 corresponds to step S150 in the aforementioned embodiment. Please refer to the aforementioned introduction for details and will not be repeated here.

[0092] As can be seen from the above technical solution, compared to the previous embodiment, this embodiment adds the process of obtaining user attributes and simultaneously referencing user attributes and tag information to select a matching image classification model. However, the tag information in the labeled image can include multiple data information, which may sometimes correspond to multiple image classification models, making selection difficult. If multiple tag information and user attributes are simultaneously referenced, the most suitable image classification model can be better selected.

[0093] In some embodiments of the present application, the process of step S250, selecting a matching image classification model based on the label information in the labeled image and referring to the user attributes, is described.

[0094] This embodiment provides two different implementation methods. The first method can pre-set the correspondence between user attributes and image classification models, and use the image classification model that corresponds to both label information and user attributes as the final selected image classification model. The second method can pre-set the rules corresponding to user attributes, which set the weights corresponding to each label information, and select the image classification model based on the weights of each label information in the image.

[0095] The first one,

[0096] S10: If the labeled image includes multiple types of label information, determine an image classification model corresponding to each type of label information.

[0097] Specifically, the labeled image may include multiple types of label information. In this case, the image classification model corresponding to each type of label information may be determined.

[0098] Among them, different label information can correspond to the same image classification model.

[0099] S11. The picture classification models corresponding to the label information in the labeled pictures are combined into a first model set.

[0100] Specifically, the image classification model combination confirmed in step S10 is called a first model set. Any image classification model in the first model set has corresponding label information, and the same image classification model can correspond to multiple label information.

[0101] S12. Determine a second model set corresponding to the user attribute, where the second model set includes at least one image classification model.

[0102] Specifically, a second model set corresponding to the user attribute may be determined, and each image classification model in the second model set matches the user attribute.

[0103] The second model set includes at least one image classification model, that is, the second model set may include only one image classification model or multiple image classification models.

[0104] In this application, different user attributes correspond to different image classification models. For example, users with family user attributes can correspond to the holiday strategy model; users with party attributes can correspond to the face recognition model; users with foodie attributes can correspond to the object recognition model; and users with photography attributes can correspond to the photography recognition model.

[0105] S13. Use the picture classification model that exists in both the first model set and the second model set as the matching picture classification model.

[0106] Specifically, the image classification model that matches both the label information and the user attributes can be determined as the final matching image classification model, that is, the image classification model that exists in both the first model set and the second model set can be confirmed as the final image classification model.

[0107] The second type

[0108] S20: If the tagged image includes multiple types of tag information, sort the tag information in the tagged image according to the tag sorting rule corresponding to the user attribute to obtain a sorting result.

[0109] Specifically, the tagged image may include multiple types of tag information. In this case, the multiple types of tag information may be sorted according to the tag sorting rules corresponding to the user attributes.

[0110] For example, if the user attribute is a foodie, the location tag, date tag, and item tags such as food tags and dish tags will be ranked first, and the rest of the tags will be ranked last.

[0111] S21. Select the image classification model corresponding to the first label information in the sorting result as the final matching image classification model.

[0112] Specifically, the image classification model that matches the first label information in the sorting result can be selected as the final image classification model.

[0113] For example, if the tags ranked first are the object tag, the location tag, and the date tag, then the image classification model corresponding to the object tag, the location tag, and the date tag is selected as the final matching image classification model.

[0114] It can be seen from the above technical solution that this embodiment provides two optional methods to realize how to simultaneously refer to the label information in the labeled image and the user attributes to select a matching image classification model. Through the above steps, the image classification model corresponding to the user and the image can be better selected for classification in subsequent steps.

[0115] Next, in some embodiments of the present application, step S140, the process of determining an image classification model that matches the label information based on the label information in the labeled image, is described in detail. The specific process may include four optional implementation methods.

[0116] The first one,

[0117] In some embodiments of the present application, if the labels in the labeled image include a location label and a date label, a holiday policy model matching the location label and the date label is selected.

[0118] The holiday policy model is trained using pictures marked with location information and date information as training data.

[0119] Based on this, the holiday policy model can identify the location information and date information in the labeled image, and identify whether the location information is domestic or foreign, and whether the date information is a holiday. Based on this, the image is classified based on the date information and / or location information.

[0120] The second type

[0121] In some embodiments of the present application, if the labels in the labeled image include face labels, location labels, and date labels, a face recognition model that matches them is selected.

[0122] The face recognition model is trained using pictures annotated with face information, location information and date information as training data.

[0123] Based on this, the face recognition model can identify the location information, face information and date information in the labeled pictures, and can identify the similarity of the face information, location information and date information in multiple labeled pictures classified using the face recognition model, and classify multiple labeled pictures according to the similarity.

[0124] Among them, if the appearance of a certain facial information exceeds a certain threshold, the face recognition model can believe that the user corresponding to the facial information has a close relationship with the user who uploaded the picture using the device.

[0125] The third

[0126] In some embodiments of the present application, if the labels in the labeled image include object labels, location labels, and date labels, an object recognition model that matches them is selected.

[0127] The object recognition model is trained using pictures marked with object information, location information and date information as training data.

[0128] Based on this, the object recognition model can identify the object information, location information and date information in the labeled images. The object information may include food information, dish information, menu information and other information related to the restaurant and food, and can identify the similarity of the object information, location information and date information in multiple labeled images classified using the object recognition model, and classify multiple labeled images according to the similarity.

[0129] The fourth type

[0130] In some embodiments of the present application, if the label in the labeled image includes a photographic label, a photographic recognition model that matches it is selected, and the photographic label includes any one or more of the following: a label of the device that captured the image, a photographic parameter label, a location label, and a scene label.

[0131] The photographic recognition model is trained using pictures annotated with photographic information as training data.

[0132] Based on this, the photographic recognition model can identify the photographic information in the labeled pictures, which may include food information, dish information, menu information and other information related to the restaurant and food, and can identify the similarity of the object information, location information and date information in multiple labeled pictures classified using the object recognition model, and classify multiple labeled pictures according to the similarity.

[0133] As can be seen from the above technical solution, this embodiment provides four optional implementation methods to determine the image classification model that matches the tag information based on the tag information in the tagged image. Obviously, different tagged images can correspond to different tag information, and different tag information can correspond to different image classification models, that is, different tagged images can correspond to different image classification models. The image classification model corresponding to the tag information can be selected to classify the tagged images, which can improve the accuracy of classification and better classify each image in the uploaded image collection.

[0134] Considering that some of the uploaded images may be processed by the user using custom rules, in some embodiments of the present application, it is possible to determine whether the image needs to be processed using user-defined rules, thereby being more in line with actual usage scenarios and meeting user expectations. Next, the process of using user-defined rules will be described in detail.

[0135] Before analyzing and identifying each image in the uploaded image set in step S120, the following steps may be performed:

[0136] S30. According to the upload method of the uploaded picture set, determine whether each picture in the uploaded picture set needs to be used, and classify according to user-defined rules. If the judgment result is no, return to step S120. If the judgment result is yes, execute step S31.

[0137] Specifically, the user can pre-set that pictures uploaded through a specific channel need to be processed using user-defined rules.

[0138] The specific path may include multiple paths, such as pictures uploaded by a specific device or pictures uploaded using a specific network.

[0139] The user-defined rules may include multiple rules. For example, the user may pre-set to directly save the pictures uploaded through a specific channel in the album to be edited, or may pre-set to directly classify the uploaded pictures according to the picture format, for example, pictures in ipg format are one album, and pictures in png format are one album.

[0140] S31. Classify the pictures according to user-defined rules.

[0141] Specifically, the pictures may be classified according to user-defined rules. The specific classification rules are defined by the user.

[0142] As can be seen from the above technical solution, this embodiment, based on the previous embodiment, takes into account the use of user-defined rules to classify images. The specific process is to first determine whether each uploaded image needs to be used, and then classify it according to the user-defined rules. If necessary, classification is only performed according to the customized rules. This can better meet the actual usage scenarios and meet user expectations.

[0143] The following describes the image classification device provided in an embodiment of the present application. The image classification device described below and the image classification method described above can be referenced to each other.

[0144] A response unit, configured to respond to an operation of a user uploading pictures through multiple devices, and receive an uploaded picture set, wherein the uploaded picture set includes multiple pictures uploaded by the user;

[0145] an identification unit, configured to analyze and identify each image in the uploaded image set to obtain an analysis and identification result corresponding to the image;

[0146] a labeling unit, configured to use the analysis and recognition result as a label for the image to obtain a labeled image;

[0147] a determining unit, configured to determine, based on the label information in the labeled image, an image classification model that matches the label information, wherein the image classification model is trained using training image data corresponding to the label information;

[0148] The classification unit is used to call the image classification model to classify the labeled images.

[0149] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0150] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0151] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. The various embodiments of the present application may be combined with each other. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for image classification, characterized in that: Applied to the cloud, including: In response to an operation of a user uploading pictures through multiple devices, receiving an uploaded picture collection, wherein the uploaded picture collection includes multiple pictures uploaded by the user; Analyze and identify each image in the uploaded image collection to obtain an analysis and identification result corresponding to the image, wherein the analysis and identification result includes photographic parameter information and image content information; Obtaining user attributes corresponding to the user; Using the analysis and recognition result as a label of the image to obtain a labeled image; Determining, based on the label information in the labeled image, an image classification model that matches the label information, where the image classification model is trained using training image data corresponding to the label information; Calling the image classification model to classify the labeled images; Determining, based on the label information in the labeled image, an image classification model that matches the label information includes: A matching image classification model is selected based on the label information in the labeled image and with reference to the user attributes.

2. The method according to claim 1, characterized in that Selecting a matching image classification model based on the label information in the labeled image and referring to the user attributes includes: If the labeled image includes multiple types of label information, determining an image classification model corresponding to each of the multiple types of label information; The image classification models corresponding to the label information in the labeled images are combined into a first model set; Determining a second model set corresponding to the user attribute, the second model set including at least one image classification model; The picture classification model that exists in both the first model set and the second model set is used as the matching picture classification model.

3. The method according to claim 1, characterized in that Selecting a matching image classification model based on the label information in the labeled image and referring to the user attributes includes: If the labeled image includes multiple types of label information, the label information in the labeled image is sorted according to the label sorting rule corresponding to the user attribute to obtain a sorting result; The image classification model corresponding to the first label information in the sorting result is selected as the final matching image classification model.

4. The method according to claim 1, wherein Determining, based on the label information in the labeled image, an image classification model that matches the label information includes: If the labels in the labeled image include location labels and date labels, select a holiday strategy model that matches them; The holiday policy model is trained using pictures marked with location information and date information as training data.

5. The method according to claim 1, wherein The determining, based on the label information in the labeled image, an image classification model that matches the label information includes: If the labels in the labeled image include face labels, location labels, and date labels, selecting a face recognition model that matches them; The face recognition model is trained using pictures annotated with face information, location information and date information as training data.

6. The method according to claim 1, characterized in that The determining, based on the label information in the labeled image, an image classification model that matches the label information includes: If the labels in the labeled image include an item label, a location label, and a date label, selecting an item recognition model that matches them; The object recognition model is trained using pictures marked with object information, location information and date information as training data.

7. The method according to claim 1, characterized in that The determining, based on the label information in the labeled image, an image classification model that matches the label information includes: If the label in the labeled image includes a photography label, selecting a photography recognition model that matches the label, wherein the photography label includes any one or more of the following: a label of the device that captured the image, a photography parameter label, a location label, and a scene label; The photographic recognition model is trained using pictures annotated with photographic information as training data.

8. The method according to any one of claims 1 to 7, characterized in that Before analyzing and identifying the content of each picture in the uploaded picture set, the method further includes: According to the upload method of the uploaded picture set, determine whether each picture in the uploaded picture set needs to be used, and classify according to user-defined rules; If the judgment result is no, returning to the step of analyzing and identifying each image in the uploaded image set; If the judgment result is yes, the image is classified according to the user-defined rules.

9. The method according to any one of claims 1 to 7, characterized in that The method responds to a user uploading a picture through multiple devices and receives an uploaded picture collection, wherein the uploaded picture collection includes multiple pictures uploaded by the user, including: In response to different users of the same family uploading pictures to a family shared directory via multiple devices, receiving an uploaded picture collection, the uploaded picture collection including multiple pictures uploaded by the different users; After calling the image classification model to classify the labeled images, the method further includes: Obtain the classification results of the labeled pictures, and store the classification results in a home sharing directory.

Images