Serve images with privacy labels

Abstract

A method for providing an image includes: performing authentication of an entity (1010); obtaining identity information of the entity based on the authentication (1020); obtaining one or more images associated with the entity based at least on the identity information of the entity (1030); and providing the one or more images, wherein each of the one or more images is attached with a privacy label (1040).

Description

Background Art

[0001] There are many images of people on the current Internet. Usually, these images show people's faces. Sometimes, images showing people's faces can be posted on the Internet by people who have the portrait rights of the images; but sometimes, such images may be posted by others without permission. Most of the time, people may not know where their images are uploaded and / or posted on the Internet. In some cases, people's privacy cannot be protected due to images posted on the Internet. Summary of the invention

[0002] This summary is provided to introduce some concepts that will be further described in the following detailed description. It is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.

[0003] Embodiments of the present disclosure provide methods and apparatus for providing images. In some implementations, authentication of an entity may be performed. Identity information of the entity may be obtained based on the authentication. One or more images associated with the entity may be obtained based at least on the identity information of the entity. The one or more images may be provided, wherein each of the one or more images is attached with a privacy label.

[0004] It should be noted that one or more of the above aspects include features that are fully described below and particularly pointed out in the claims. The following description and the accompanying drawings set forth in detail certain illustrative features of one or more aspects. These features are merely indicative of various ways in which the principles of the various aspects may be employed, and the present disclosure is intended to include all such aspects and their equivalents. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] The disclosed aspects will hereinafter be described in conjunction with the accompanying drawings, which are provided for the purpose of illustrating rather than limiting the disclosed aspects.

[0006] Figure 1 An exemplary architecture of a system for providing an image over a network according to an embodiment is shown.

[0007] Figure 2 An exemplary entity index database according to an embodiment is shown.

[0008] Figure 3 An exemplary process for building an entity index database according to an embodiment is shown.

[0009] Figure 4 An exemplary facial recognition model according to an embodiment is shown.

[0010] Figure 5An exemplary system for generating images with privacy labels according to an embodiment is shown.

[0011] Figure 6 An exemplary process for performing authentication on an entity according to an embodiment is shown.

[0012] Figure 7 An exemplary process for providing an image with a privacy label is shown in accordance with an embodiment.

[0013] Figure 8 Another exemplary process for providing an image with a privacy label according to an embodiment is shown.

[0014] Fig. 9 An exemplary user interface for presenting an image with a privacy label is shown in accordance with an embodiment.

[0015] Fig.10 A flow chart of an exemplary method for providing an image according to an embodiment is shown.

[0016] Fig.11 An exemplary apparatus for providing an image according to an embodiment is shown.

[0017] Fig.12 Another exemplary apparatus for providing an image according to an embodiment is shown. DETAILED DESCRIPTION

[0018] The present disclosure will now be discussed with reference to several example implementations. It should be understood that these implementations are discussed only to enable those skilled in the art to better understand and thus implement embodiments of the present disclosure, rather than implying any limitation on the scope of the present disclosure.

[0019] Nowadays, there are a large number of images showing human faces on the Internet. Most images with human faces may not be posted by the people who have the image portrait rights themselves. Sometimes, people may not want some images to be released to the public, for example, images showing their children; images showing their negative image, such as images showing people drunk or smoking or images showing the faces of actresses without makeup; images modified by deep fakes, etc. However, people may not know where these images are uploaded or posted on the Internet. People may spend a lot of time searching all their images and may have difficulty finding images that they may not intend to post on the Internet.

[0020] Embodiments of the present invention propose to provide images with privacy labels associated with entities. In this article, images may include photos, pictures, portraits, etc. Images with privacy labels may be pre-stored in a database, or may be obtained and / or processed from the Internet in real time. In this article, the database may include an entity index database, which includes the ID of each candidate entity, the profile information of each candidate entity, and the image of each candidate entity. The database can be constructed by utilizing any facial recognition technology and search engine to collect and store multiple personal images on the Internet.

[0021] The privacy label may indicate that the image belongs to at least one of the following categories: unauthorized category, sensitive category, modified category, and normal category. As used herein, the unauthorized category may indicate that the image is, for example, not authorized for publication, published outside a specified location, published outside a specified time period, or published outside a specified event. The sensitive category indicates that the image may contain sensitive content, for example, with children, smoking or drunkenness, ugly appearance, etc. The modified category indicates that the image may be generated by modifying the original image, for example, by editing the original image, or may be a forged image. The normal category indicates any other image that is not identified as unauthorized, sensitive, and modified categories.

[0022] Before providing an image with a privacy label associated with an entity, the entity may need to be authenticated. Authentication may be initiated by a user requesting an image associated with an entity or by an application for providing an image. In this article, the user requesting an image associated with an entity may be the entity itself or an agent of the entity. In this article, the agent of the entity may also be referred to as the assistant of the entity and may represent any party with an authorization certificate of the entity. For example, the agent of the entity may be a real person or a virtual person, such as an intelligent robot, a computer device, etc. authorized by the entity. Authentication may be achieved by any appropriate authentication technology, for example, by at least one of the following: a user ID and password pair, a dynamic token or password, a USB key, biometrics, etc. By providing an image with a privacy label to an associated entity or its agent via the present method or system, the image of the associated entity may be checked and / or managed. In some examples, an entity or its agent may request an application to delete some images with certain privacy labels from an image source. In some other examples, an entity or its agent may request an application to provide a link to an image source of one or more images, and the entity and / or its agent may contact the image source through the corresponding link to delete certain images from the image source if necessary. Through an embodiment of the present disclosure, the privacy of an entity may be protected.

[0023] Figure 1 An exemplary architecture of a system 100 for providing images over a network according to an embodiment is shown.

[0024] exist Figure 1 In the example, the network 110 is used for interconnection between the terminal device 120 and the server 130.

[0025] The network 110 may be any type of network capable of interconnecting network entities. The network 110 may be a single network or a combination of various networks. In terms of coverage, the network 110 may be a local area network (LAN), a wide area network (WAN), etc. In terms of carrier media, the network 110 may be a wired network, a wireless network, etc. In terms of data exchange technology, the network 110 may be a circuit switching network, a packet switching network, etc.

[0026] The terminal device 120 may be any type of electronic computing device that can connect to the network 110, access a server or website on the network 110, process data or signals, etc. For example, the terminal device 120 may be a desktop computer, a laptop computer, a tablet computer, a smart phone, an AI terminal, etc. Figure 1 Only one terminal device is shown in FIG. 1 , but it should be understood that a different number of terminal devices may be connected to the network 110 .

[0027] In one implementation, the terminal device 120 may include an application (APP) 121 , a camera 122 , a user interface 123 , a memory 124 , a processor 125 , and one or more communication connections 126 for connecting to the network 110 .

[0028] Application 121 may include any application that can receive input from a user and provide an image associated with an entity. Application 121 may be a dedicated application for providing images with privacy labels, or a public third-party application that can provide images with privacy labels.

[0029] The camera 122 may capture images from the environment, including images of a user of the terminal device 120 .

[0030] User interface (UI) 123 can be used to present images. Images can be attached with privacy labels and presented to users. UI 123 can be specific to application 121 in terminal device 120, or can be incorporated into third-party applications (such as Facebook, Instagram, Linkedin, Twitter, or any other social network application).

[0031] The memory 124 may be one or more devices for storing data, which may include read-only memory (ROM), random access memory (RAM), magnetic RAM, core memory, magnetic disk storage media, optical storage media, flash memory, and / or other machine-readable media for storing information. The term "machine-readable medium" may include, but is not limited to, portable or fixed storage, optical storage, wireless channels, and various other media for storing, containing, or carrying instructions and / or data.

[0032] Processor 125 may include any number of processing units and is programmed to execute computer executable instructions for implementing various aspects of the present disclosure. In some examples, processor 125 is programmed to execute computer executable instructions stored in memory 124 to implement functions such as the attached Fig.10 The method shown in .

[0033] The communication connection 126 may include a wireless connection or a wired connection to connect the terminal device 120 to the network 110 or any other device.

[0034] Server 130 may be connected to database 140. Figure 1 Although not shown in the figure, the server 130 may include a database 140. The server 130 may provide data or applications to the terminal device 120. As an example, the server 130 may be a network server that provides data over a network. The server 130 may provide data to the terminal device 120 over a network through the network 110.

[0035] The database 140 may include an entity index database, which includes the ID of each candidate entity, the profile information of each candidate entity, and the image of each candidate entity. The data in the entity index database 140 may be collected or obtained from the image source 150 and / or the knowledge graph 160.

[0036] Image source 150 may include any source that may provide images, such as a social networking site, a search site, public and / or personal websites and / or applications, and the like.

[0037] In this article, the knowledge graph 160 may refer to a single knowledge graph or multiple knowledge graphs in various domains. The knowledge graph 160 may be established from various websites on the Internet, such as Facebook, Linkedin, Twitter, Wikipedia, etc. The knowledge graph 160 may include any possible knowledge graph, such as Satori, Knowledge Vault, Schema, Graph Search, etc.

[0038] It should be understood that Figure 1All components shown in the figure may be exemplary and may be used in Figure 1 Any component in the terminal device 120 may be added, omitted or replaced.

[0039] Figure 2 An exemplary entity index database 200 according to an embodiment is shown.

[0040] The entity index database 200 may include several entity index items, for example, Figure 2 , which includes a candidate entity ID, a candidate entity profile, and a candidate entity image. In this article, a candidate entity may include any entity collected for building an entity index database 200, which can be used to compare with the requested entity to find a matching candidate entity corresponding to the requested entity.

[0041] like Figure 2 As shown, the candidate entity IDs may be randomly or sequentially assigned to the candidate entities. It should be understood that although the candidate entity IDs are shown in the form of ID_0, ID_1, ID_2, ..., ID_n, they may be any suitable form for indicating the corresponding candidate entity.

[0042] The candidate entity profile may include any profile information collected or generated based at least on a knowledge graph (e.g., knowledge graph 160). It may include any personal information of the candidate entity. For example, the candidate entity profile may include, but is not limited to, the entity's name, gender, age, profile photo, home address, employer information, occupation, degree of fame (e.g., public figure or ordinary person), etc. In some examples, this entity profile information may be represented in the form of a feature vector.

[0043] The candidate entity images in each entity index item of the entity index database 200 may include multiple images associated with the candidate entity having a unique entity ID and a privacy label and / or feature vector for each image. Figure 2 As shown, candidate entity images of a candidate entity assigned with ID_0 may be displayed as images each having a privacy label and a feature vector V_nm, where V_nm represents facial features of each image, such as {image, unauthorized, V_01}, {image, sensitive, V_02}, {image, normal, V_03}.

[0044] It should be understood that although the candidate entity image is shown in the form of a group {image, privacy label, feature vector}, it can be shown as an image with a privacy label, a feature vector attached, and any other features (e.g., a view count and / or a share count, a link to the image source, and / or a publication chain of each image) may also be attached or provided with any other features. The view count, share count, and / or a link to the image source may be obtained from the image source of each image by any extraction model and attached to the candidate entity image or provided with the candidate entity image. The publication chain of the image may be represented as an ordered list of sites for publishing and broadcasting the image. In some examples, the site may also be referred to as a web page. For example, the publication chain of the image may be obtained or determined via a search engine by obtaining the "source" of the image from an associated site and cascading the link or site name to form an ordered list of sites. The publication chain of the image may be attached to the image or provided with the image. For example, image i may be first published on site A, and then referenced and / or copied by site B and forwarded to one or more other sites, such as site C, site D... The publishing chain of the image can be expressed as: site A-site B-site C-site D.......

[0045] In some examples, images in such a group may not be stored in the entity index database, but in other databases or the cloud, and represented by corresponding feature vectors and links to the images. In some examples, the privacy label may be attached to the image in any suitable manner, such as by watermarking the image, attaching it as an annotation around the image, etc. In some other examples, the privacy label may be displayed as A, B, C, D levels, each level indicating what category the image belongs to, such as A for unauthorized category, B for sensitive category, C for modified category, and D for normal category.

[0046] It should be understood that Figure 2 The structure of the entity index database 200 shown in FIG. 2 is merely exemplary, and any other structure for embodying entity profile or identity information and entity image may be possible for the present disclosure.

[0047] Figure 3 An exemplary process 300 for building an entity index database according to an embodiment is shown.

[0048] At 302, an image of an entity may be obtained from one or more image sources, such as captured from the Internet, social networks, and any site where images may be published, or provided by one or more image sources. The image of the entity may include a face of the entity, which may indicate the identity of the entity. In implementation, as many images as possible may be obtained through one or more search engines from the Internet, and the images may be processed by a face detection model. The face detection model may be implemented by an object detection model, such as a regional convolutional neural network (RCNN), a fast RCNN, a faster RCNN, or a single multi-box detector (SSD). For example, it may add 5 landmarks to the face recognition model to predict the coordinates of two eyes, a nose, and the left and right sides of the mouth to recognize the face of the entity.

[0049] At 304, a privacy label for the image may be determined based on at least one rule and attached to the image to generate an image with the privacy label. For example, the privacy label may be determined by a privacy label model (e.g., a multi-label classifier using machine learning (e.g., supervised learning)) based at least on the image and at least one rule.

[0050] In some examples, the privacy label may identify the image as at least one of the following: an unauthorized category, a sensitive category, a modified category, and a normal category. In some examples, the unauthorized category may indicate that the image is not authorized to be published, published outside a specified location, published outside a specified time period, or published outside a specified event. In some examples, the sensitive category indicates that the image includes sensitive content. In some examples, the modified category indicates that the image is generated by modifying the original image. In some examples, the normal category indicates that the image is an image other than images identified as unauthorized, sensitive, and modified.

[0051] The at least one rule may include: rules associated with the authorization status of each image, rules associated with the sensitive content contained in each image, and rules associated with the modification status of each image. For example, the rules associated with the authorization status of each image may include at least one of the following rules: whether the image is from a source listed in a whitelist provided by the user; whether the image is allowed to be published to a specified site location, event, or time period. In some examples, if the user does not provide a whitelist, any image may be considered unauthorized. In some other examples, if any image may be published outside of a specified site location, event, or time period, it may be considered unauthorized. For example, if the specified event is the Oscars, then images that are not published for the Oscars may be considered unauthorized. In implementation, the rules associated with the sensitive content contained in each image may define a list of sensitive types or labels through a classification model, such as images showing a family, images showing a child, images showing an ugly entity, images showing an entity drunk and / or smoking, and the like. The rules associated with the modification status of each image may include at least one of the following rules: whether the current image is modified compared to the original image; whether the current image is maliciously modified by the deep learning-based model through malicious modification detection; whether the current image is edited by the deep learning-based model through editing detection. In some examples, the original image can be provided by the user or the entity itself. In some examples, the deep learning-based model can be trained to predict whether the image is maliciously modified by collecting unmodified images as positive examples and modified pictures as negative examples. In some examples, the deep learning-based model can be trained based on the original pixels to understand whether the current image has been edited.

[0052] At 306, facial recognition can be performed on the image. For example, facial recognition can be performed by a facial recognition model. For example, when an image is received, a feature vector can be extracted by a facial embedding model implemented in the facial recognition model, and the feature vector can be projected to a probability of an entity name by a softmax layer, as described later in Figure 4 As shown in . The entity name with the highest probability can be selected as the facial recognition result for the face in the image. In other examples, other entity names with probabilities other than the highest probability can be selected based on different settings. In some examples, the facial recognition model can include a recognition model for a specific group, for example, by utilizing a celebrity recognition model for celebrities such as MS-Celeb-1M, Rekognition, etc.

[0053] At 308, if an entity name can be generated through facial recognition on the image at 306, the process can move to 310. If an entity name cannot be generated, the process can move to 314.

[0054] At 310, entity names and facial features of faces in the image may be generated. In some examples, the facial features may be represented by a feature vector or a number of feature vectors.

[0055] At 312 , a search may be performed on an entity index database based at least on the generated entity name and / or facial features to find a candidate entity that matches the entity in the obtained image at 302 .

[0056] At 314, if the entity name cannot be generated by applying the facial recognition model on the image, a named entity extraction model may be performed on the image to extract the entity name from the image source, site, or web page displaying the image. In some examples, the named entity extraction model may be implemented by natural language processing (NLP) techniques, such as bidirectional long short-term memory units (Bi-LSTM), bidirectional encoder representations from transformer models (BERT), and the like.

[0057] At 316, if an entity name can be generated through named entity extraction at 314, the process can move to step 310 to generate entity names and facial features for the image. If an entity name cannot be generated through named entity extraction, the process can move to step 318 to generate facial features. In some examples, if an entity name cannot be generated, a face grouping model can be utilized to collect images with the same or similar facial features together as images without entity names or entity profiles (e.g., Figure 2 , and an entity index item assigned with ID_2). Herein, each entity index item may represent an entity. In some examples, a face grouping model may be implemented, for example, by using a density peak clustering algorithm, which may cluster several images into a group. In some examples, those entity images whose difference may be equal to or less than a threshold may be clustered into a group corresponding to an entity, which means that those entity images may be associated with the same entity.

[0058] At 320 , a search may be performed on an entity index database (eg, entity index database 200 ) based on the generated facial features to find candidate entities that match the entity in the obtained image at 302 .

[0059] At 322, if there is a matching candidate entity, the process can move to step 324. At 324, the obtained image with the determined privacy label and the generated facial features can be added to the "Candidate Entity Image" section in the entity index item of the matching candidate entity.

[0060] If there is no matching candidate entity, the process can move to step 326. At 326, a new entity index item for the new candidate entity can be generated by causing the obtained image to include the privacy label, facial features, and possible entity names.

[0061] It should be understood that while entity names may be generated and included in entity index terms, any other entity profile information may also be generated and included in entity index terms through any facial recognition and / or information extraction techniques (e.g., from a user profile and / or from a knowledge graph).

[0062] Figure 4 An exemplary facial recognition model 400 according to an embodiment is shown. The facial recognition model 400 can be implemented by a neural network structure (e.g., ResNet, DenseNet, etc.). In this article, a 34-layer ResNet for encoding an image of an entity of a neural network structure can be shown as an example, which can provide relatively good accuracy and fast training / testing.

[0063] For example, in Figure 4 In the example, "3*3conv,64" means that there are 64 filters, each with a convolution kernel or function and a size of 3×3 pixels. " / 2" means double stride. "avg pool" means average pooling operation. The output of the average pooling operation is an entity vector, which is a dense vector representation of the input image.

[0064] The entity vectors can also be provided to the softmax layer. The softmax layer can output entity names according to different entity name classification or selection strategies. For example, the entity names with the highest probability or the highest value or the values ​​that meet the conditions or thresholds can be selected for output. The entity names can be represented in the form of keywords, feature vectors, or some feature vectors.

[0065] The facial recognition model 400 may perform facial recognition on the image. For example, when an image is obtained, facial recognition may be performed on the image by the facial recognition model 400 to obtain the entity name of the entity shown in the image.

[0066] It should be understood that ResNet is an exemplary technology that can be used in image recognition, and any other technology, such as DenseNet, AlexNet, GoogleNet, VGG-Net, etc., can be used in image recognition.

[0067] In some embodiments, a region of interest (RoI) of an image may be identified, and the facial recognition model 400 may also perform facial recognition on the RoI to obtain entity names of entities shown in the RoI. In this document, a "RoI" may refer to a facial region in an image showing more than one person, each person having a face shown in the image. For example, suppose there are two different people in the image. The facial regions of these two different people may represent different entities with different names. If these two facial regions can be identified and the name of each entity represented by each facial region can be determined, then it would be helpful to understand which name the entity on the image has.

[0068] Figure 5 An exemplary system 500 for generating an image with a privacy label according to an embodiment is shown. The system 500 may implement Figure 3 Operation 304 in .

[0069] like Figure 5 As shown, an image 501 is provided to a privacy label model 510 and a generation model 520. The privacy label model 510 may determine a privacy label of the image 501 based on at least one rule 502. The rule 502 may be provided to the privacy label model 510 from other components, or may be pre-stored in the privacy label model 510. The privacy label model 510 may provide the generated privacy label to the generation model 520. The generation model 520 may attach the provided privacy label to the image 510 to generate an image with a privacy label 503.

[0070] Figure 6 An exemplary process 600 for performing authentication on an entity is shown in accordance with an embodiment.

[0071] At 602, authentication information of an entity may be received. In this context, the authentication information may be provided by a user, who is the entity or an agent of the entity, or may be obtained from a third-party application by a user successfully logging into the third-party application. The authentication information may include at least one of the following: a user name and a static password pair, a dynamic password, a token, a USB key, a digital signature, biometric information, etc.

[0072] At 604, authentication can be performed based on the authentication information. In some examples, when the authentication information of the entity is the biometric identification information of the entity (for example, a facial image of the entity obtained by a camera), performing authentication may include: performing facial liveness detection to determine whether the obtained image is a real-time photo. In some examples, when the authentication information is provided by a user who is an agent of the entity and the agent cannot be authenticated through biometrics, authorization information for the agent from the entity may be required, such as an authorization certificate, a power of attorney, a notarial certificate, etc. In such examples, authentication can be performed by verifying at least the authorization information. In some examples, when the authentication information is obtained from a third-party application, authentication can be performed based on the third-party authentication. For example, if the authentication of the third-party application of the entity is successful, the authentication of the entity in this article can be considered to be successful.

[0073] At 606, identity information of the entity can be provided based on the authentication. For example, the identity information can include at least one of an identity description of the entity and an identity image of the entity. In some examples, when the authentication information includes a facial image, the identity information can be obtained directly from the facial image through a facial recognition model, such as an identity name, an identity image, which can be just an image or can be represented by image features and / or both. In this document, image features may also be referred to as facial features. In some other examples, when the authentication information includes other information in addition to the facial image, the identity information can be obtained from a user profile, such as a user profile in a knowledge graph or a third-party application, etc.

[0074] It should be understood that the above Figure 6 The authentication process and identity information shown may be merely exemplary, and any other authentication methods and identity information are possible for the present disclosure.

[0075] Figure 7 An exemplary process 700 for providing an image with a privacy label according to an embodiment is shown. The exemplary process 700 may be based on, for example, Figure 6 The identity information provided in operation 606 of step 100 may be performed by an independent application, an application that cooperates with a third-party application, or a third-party application. For example, an independent application may include an application dedicated to providing images with at least a privacy label. A third-party application may include a social application, a search application, and the like, such as Bing, Facebook, Linkedin, Instagram, Twitter, and the like.

[0076] At 702, an identity description of an entity may be obtained. Herein, the identity description of an entity may be a text description of an entity profile, such as, but not limited to, the entity's name, gender, age, home address, employer information, occupation, degree of fame (e.g., public figure or ordinary person), and the like.

[0077] At 704, entity names may be extracted from the identity description. The extraction operation may be performed by any possible feature extraction model or technique. The entity names may be extracted in the form of keywords or feature vectors.

[0078] At 706, matching candidate entities may be determined based on the extracted entity names. In some examples, the determination may include: selecting a candidate entity name from the extracted entity names by a keyword matching algorithm; determining a candidate entity corresponding to the selected candidate entity name as a matching candidate entity. In some other examples, the determination may include: calculating a feature vector difference between the extracted entity name and the candidate entity name in the entity index item; comparing the feature vector difference with a threshold; determining one or more candidate entity names whose feature vector difference is less than the threshold; selecting at least one candidate entity name from the determined one or more candidate entity names based on a ranking rule; selecting at least one candidate entity corresponding to at least one selected candidate entity name as at least one matching candidate entity.

[0079] At 708, one or more images associated with the candidate entity may be obtained from the entity index database. Specifically, the images may be obtained from a "candidate entity image" section in the entity index database.

[0080] At 710, each image with a privacy label associated with a matching candidate entity may be provided from a "candidate entity image" section in a corresponding entity index item to be presented to a user through a user interface, as follows: Fig. 9 shown.

[0081] In some examples, the image may be provided with at least one of the following: a view count, a share count, a link to the image source, and a publishing chain. In this article, the view count, the share count, and the link to the image source for each image may be obtained from the image source of the image. The publishing chain of the image may be obtained as an ordered list of sites that published the image.

[0082] Optionally, at 712, instructions for managing one or more images may be received. In some examples, the instructions for managing one or more images may include at least one of the following: instructions for deleting at least one of the one or more images from a specified image source; instructions for monitoring a view count and / or a share count of at least one of the one or more images; instructions for changing a privacy label of at least one of the one or more images; and instructions for obtaining an image source of at least one of the one or more images. In this context, the instructions for deleting may include instructions for deleting the at least one image from all sites displaying the at least one image, or instructions for deleting the at least one image from an unauthorized site displaying the at least one image.

[0083] Figure 8 Another exemplary process 800 for providing an image with a privacy label according to an embodiment is shown. The exemplary process 800 can be based on, for example, Figure 6 The privacy protection may be performed by using the identity image in the identity information provided in operation 606 of step 1, and may be implemented by an independent application, an application that cooperates with a third-party application, or a third-party application. For example, an independent application may include an application dedicated to providing an image with at least a privacy label. A third-party application may include a social application, a search application, etc., such as Bing, Facebook, Linkedin, Twitter, Instagram, etc.

[0084] At 802, an identity image having a face of an entity may be obtained. In this context, the identity image may be, for example but not limited to, a profile photo stored in a user profile for the entity, or a profile photo currently captured from a user via a camera. In the latter case, the user may be the entity itself.

[0085] At 804, a facial recognition model may be used to perform facial recognition on the image to generate an entity name. The facial recognition model herein may be combined with Figure 4 The facial recognition model 400 described is implemented.

[0086] At 806, if an entity name can be generated through facial recognition on the image at 804, the process can move to 808. Otherwise, the process can move to 810.

[0087] At 808, entity names and facial features for faces in the image may be generated. In some examples, entity names may be represented in the form of keywords or feature vectors. Facial features may be represented in the form of feature vectors.

[0088] At 810, if an entity name cannot be generated, facial features of the faces in the image may be generated.

[0089] In some examples, entity name generation and / or facial feature generation may include: compressing feature vectors into a result set through one or more ranking layers. The ranking layers may be implemented, for example, by a product quantization algorithm.

[0090] At 812, matching candidate entities can be determined based on the generated entity names and / or facial features. In some examples, the determination can be performed based on the entity names, similar to Figure 7 In some other examples, the determination may be performed based on facial features, which may include: calculating a feature vector difference between the facial features and a candidate entity image or a candidate profile photo in the entity index item; comparing the feature vector difference with a threshold; determining a candidate profile photo or one or more candidate entity images whose feature vector difference is less than the threshold; and determining at least one candidate entity corresponding to the candidate profile photo or the candidate entity image as at least one matching candidate entity.

[0091] At 814, one or more images associated with the candidate entity may be obtained from the entity index database. Specifically, the image may be obtained from a "candidate entity image" section in the entity index database.

[0092] At 816, one or more images each having a privacy label associated with the matching candidate entity may be provided from the "Candidate Entity Images" section of the corresponding entity index item to be presented to the user through the user interface, as follows: Fig. 9 shown.

[0093] In some examples, the image may be provided with at least one of the following: a view count, a share count, a link to the image source, and a publication chain. In this article, the view count, the share count, and the link to the image source for each image may be obtained from the image source of the image. The publication chain of the image may be obtained as an ordered list of sites that published the image.

[0094] Optionally, at 818 (which may be similar to Figure 7At operation 712 in ), an instruction for managing one or more images may be received. In some examples, the instruction for managing one or more images may include at least one of the following: an instruction for deleting at least one of the one or more images from a specified image source; an instruction for monitoring a view count and / or a share count of at least one of the one or more images; an instruction for changing a privacy label of at least one of the one or more images; and an instruction for obtaining an image source of at least one of the one or more images. In this article, the instruction for deleting may include an instruction for deleting the at least one image from all sites displaying the at least one image, or an instruction for deleting the at least one image from an unauthorized site displaying the at least one image.

[0095] It should be understood that although the entity name shown in this article is extracted from the identity description or from the obtained image through facial recognition, it is merely exemplary profile information, and any other profile information described above may also be extracted and processed with or without the entity name.

[0096] It should be understood that although it is described above that an image with a privacy label can be provided based on the identity description and the identity image of the entity respectively, an image can also be provided based on a combination of them.

[0097] Fig. 9 An exemplary user interface 900 for presenting an image with a privacy label according to an embodiment is shown. Although the user interface 900 is embodied in a smartphone as an example herein, it can be embodied in any other terminal device, such as a desktop computer, a laptop computer, a tablet computer, an AI terminal, etc.

[0098] The images can be presented as an entity image set 810 along with privacy labels, such as images marked as unauthorized, images marked as sensitive, images marked as modified, and images marked as normal. The images can be displayed one by one, or can be displayed in groups, such as unauthorized group {image 1...}, sensitive group {image 1...}, normal group {image 1, image 2, image 3...}, etc.

[0099] Fig.10 A flow chart of an exemplary method 1000 for providing images according to an embodiment is shown. When a user wants to search or examine images associated with an entity, the method 1000 can be implemented. In some examples, the user can be the entity. In some other examples, the user can be an agent of the entity.

[0100] At 1010, authentication of an entity may be performed.

[0101] At 1020, identity information of the entity can be obtained based on the authentication.

[0102] At 1030, one or more images associated with the entity can be obtained based at least on the identity information of the entity.

[0103] At 1040, one or more images may be provided, wherein each of the one or more images has a privacy tag attached thereto.

[0104] In one implementation, the privacy label identifies the image as at least one of: an unauthorized category, a sensitive category, a modified category, and a normal category.

[0105] In one implementation, the unauthorized category may indicate that the image is: not authorized for publication, published outside of a specified location, published outside of a specified time period, or published outside of a specified event. In other implementations, the sensitive category may indicate that the image includes sensitive content. In other implementations, the modified category may indicate that the image is generated by modifying the original image. In other implementations, the normal category may indicate that the image is other than images identified as unauthorized, sensitive, and modified.

[0106] In one implementation, the privacy label of each image may be determined based on at least one of: rules associated with an authorization status of each image, rules associated with sensitive content contained in each image, and rules associated with a modification status of each image.

[0107] In one implementation, obtaining one or more images associated with the entity may also include: determining a candidate entity matching the entity in an entity index database based on the identity information of the entity; and obtaining one or more images associated with the candidate entity from the entity index database.

[0108] In one implementation, the entity index database may include a plurality of entity index items. In another implementation, each entity index item may include profile information and an image associated with the corresponding candidate entity. In another implementation, the profile information may be created based at least on the knowledge graph, and the image associated with the corresponding candidate entity may be obtained from one or more image sources.

[0109] In one implementation, the identity information may include at least one of an identity description of the entity and an identity image of the entity.

[0110] In one implementation, authentication may be performed by at least one of: facial liveness detection, authorization information verification, and third-party authentication.

[0111] In one implementation, method 1000 may also include receiving an indication for managing one or more images.

[0112] In one implementation, instructions for managing one or more images include at least one of the following: instructions for deleting at least one of the one or more images from a specified image source; instructions for monitoring a view count and / or a share count of at least one of the one or more images; instructions for changing a privacy label of at least one of the one or more images; and instructions for obtaining an image source of at least one of the one or more images.

[0113] In one implementation, the method 1000 may further include: obtaining a view count and / or a share count for each of the one or more images; and providing the view count and / or the share count together with the each image.

[0114] In one implementation, method 1000 may also include: obtaining a link to an image source for each of the one or more images; and providing the link to the image source along with the each image.

[0115] In one implementation, the method 1000 may also include: obtaining a publication chain for each of the one or more images; and providing the publication chain along with the each image.

[0116] In one implementation, method 1000 may be implemented by a standalone application, an application coordinated with a third party application, or a third party application.

[0117] It should be understood that method 1000 may also include any steps / processes for providing an image according to the above-mentioned embodiments of the present disclosure.

[0118] Fig.11 An exemplary apparatus 1100 for providing an image according to an embodiment is shown.

[0119] The device 1100 may include: an authentication execution module 1110, which is used to perform authentication of an entity; an identity information acquisition module 1120, which is used to obtain identity information of the entity based on the authentication; an image acquisition module 1130, which is used to obtain one or more images associated with the entity based at least on the identity information of the entity; and an image providing module 1140, which is used to provide the one or more images, wherein each of the one or more images is attached with a privacy label.

[0120] In one implementation, the privacy label identifies the image as at least one of: an unauthorized category, a sensitive category, a modified category, and a normal category.

[0121] In one implementation, the privacy label for each image is determined based on at least one of: rules associated with an authorization status of each image, rules associated with sensitive content contained in each image, and rules associated with a modification status of each image.

[0122] In one implementation, the image obtaining module 1120 is further used to: determine a candidate entity matching the entity in an entity index database based on the identity information of the entity; and obtain one or more images associated with the candidate entity from the entity index database.

[0123] In one implementation, the apparatus 1100 may further include an instruction receiving module, which is configured to receive an instruction for managing one or more images.

[0124] In addition, the apparatus 1100 may further include any other modules configured to provide an image according to the above-mentioned embodiments of the present disclosure.

[0125] Fig.12 Another exemplary apparatus 1200 for providing an image according to an embodiment is shown.

[0126] The apparatus 1200 may include at least one processor 1210. The apparatus 1200 may further include a memory 1220 connected to the processor 1210. The memory 1220 may store computer executable instructions that, when executed, cause the processor 1210 to perform any operation of the method for providing an image according to the above-described embodiments of the present disclosure.

[0127] Embodiments of the present disclosure may be embodied in a non-transitory computer-readable medium. The non-transitory computer-readable medium may include instructions that, when executed, cause one or more processors to perform any operation of the method for providing an image according to the embodiments of the present disclosure described above.

[0128] It should be understood that all operations in the above methods are merely exemplary, and the present disclosure is not limited to any operations in these methods or the sequential order of these operations, and should cover all other equivalents under the same or similar concepts.

[0129] It should also be understood that all modules in the above-mentioned device can be implemented in various ways. These modules can be implemented as hardware, software or a combination thereof. In addition, any one of these modules can also be functionally divided into submodules or combined together.

[0130] Processors have been described in conjunction with various devices and methods. These processors can be implemented using electronic hardware, computer software, or any combination thereof. Whether these processors are implemented as hardware or software will depend on the specific application and the overall design constraints imposed on the system. For example, the processors presented in this disclosure, any part of the processors, or any combination of processors can be implemented with microprocessors, microcontrollers, digital signal processors (DSPs), field programmable gate arrays (FPGAs), programmable logic devices (PLDs), state machines, gated logic, discrete hardware circuits, and other suitable processing components configured to perform the various functions described throughout this disclosure. The functions of the processors presented in this disclosure, any part of the processors, or any combination of processors can be implemented with software executed by microprocessors, microcontrollers, DSPs, or other suitable platforms.

[0131] Software should be broadly interpreted as instructions, instruction sets, codes, code segments, program codes, programs, subroutines, software modules, applications, software applications, software packages, routines, subroutines, objects, execution threads, processes, functions, etc. Software can reside on a computer-readable medium. For example, a computer-readable medium can include a memory, such as a magnetic storage device (e.g., a hard disk, a floppy disk, a magnetic stripe), an optical disk, a smart card, a flash memory device, a random access memory (RAM), a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), a register, or a removable disk. Although the memory is shown as being separate from the processor in various aspects presented throughout the present disclosure, the memory can be inside the processor (e.g., a cache or a register).

[0132] The previous description is provided to enable those skilled in the art to implement the various aspects described herein. Various modifications to these aspects will be apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects. Therefore, the claims are not intended to be limited to the aspects shown herein. All structural and functional equivalents of the elements throughout the various aspects described in this disclosure that are known or will later be known to those of ordinary skill in the art are expressly incorporated herein by reference and are intended to be covered by the claims.

Claims

1. A method for providing an image, comprising: Perform authentication of entities; obtaining identity information of the entity based on the authentication; obtaining one or more images associated with the entity based at least on the identity information of the entity; determining a category of the one or more images, wherein the category comprises an unauthorized category, a modified category, a sensitive category, and a normal category; appending a privacy label to each of the one or more images based on the determined category of the one or more images; obtaining a publication chain comprising, for each of the one or more images, a source and a cascading link from the source, the cascading link indicating a path of the image from the source; and The one or more images are provided with the attached privacy label and the publication chain.

2. The method according to claim 1, wherein: The privacy label identifies the image as at least one of: an unauthorized category, a sensitive category, a modified category, and a normal category.

3. The method according to claim 2, wherein: The unauthorized category indication image is: unauthorized publication, publication outside a designated location, publication outside a designated time period, or publication outside a designated event; The sensitive category indicates that the image includes sensitive content; The modified category indicates that the image is generated by modifying an original image; as well as The normal category indicates that the image is an image other than images identified as unauthorized, sensitive, and modified.

4. The method according to claim 1, wherein: The privacy label for each image is determined based on at least one of: rules associated with an authorization status of each image, rules associated with sensitive content contained in each image, and rules associated with a modification status of each image.

5. The method according to claim 1, wherein: The obtaining of one or more images associated with the entity further comprises: Determining, based on the identity information of the entity, a candidate entity in an entity index database that matches the entity; and One or more images associated with the candidate entity are obtained from the entity index database.

6. The method according to claim 5, wherein: The entity index database includes multiple entity index items, each entity index item includes profile information and an image associated with a corresponding candidate entity, and wherein the profile information is created based on at least a knowledge graph, and the image associated with the corresponding candidate entity is obtained from one or more image sources.

7. The method according to claim 1, wherein: The identity information includes at least one of the following: an identity description of the entity and an identity image of the entity.

8. The method according to claim 1, wherein: The authentication is performed by at least one of: facial liveness detection, authorization information verification, and third-party authentication.

9. The method according to claim 1, further comprising: An indication for managing the one or more images is received.

10. The method according to claim 9, wherein: The instructions for managing the one or more images include at least one of: an indication for deleting at least one of the one or more images from a specified image source; an indication for monitoring a view count and / or a share count of at least one of the one or more images; an instruction to change a privacy tag of at least one of the one or more images; as well as An indication of an image source used to obtain at least one of the one or more images.

11. The method according to claim 1, further comprising: obtaining a view count and / or a share count of each of the one or more images; as well as The view count and / or share count is provided along with the respective image.

12. The method according to claim 1, further comprising: obtaining a link to an image source for each of the one or more images; as well as The link to the image source is provided along with the respective image.

13. The method according to claim 1, wherein: The publication chain also includes the order in which images will be published successively on different sites.

14. The method according to claim 1, wherein: The method is implemented by an independent application, an application coordinated with a third-party application, or a third-party application.

15. An apparatus for providing an image, comprising: An authentication execution module, used to execute authentication on an entity; An identity information obtaining module, used to obtain the identity information of the entity based on the authentication; an image acquisition module, configured to acquire one or more images associated with the entity based at least on the identity information of the entity; means for determining a category of the one or more images, wherein the category comprises an unauthorized category, a modified category, a sensitive category, and a normal category; means for attaching a privacy label to each of the one or more images based on the determined category of the one or more images; means for obtaining a publication chain comprising, for each of the one or more images, a source and a cascading link from the source, the cascading link indicating a path of the image from the source; and The image providing module is used to provide the one or more images with the attached privacy label and the publication chain.

16. The device according to claim 15, wherein: The privacy label identifies the image as at least one of: an unauthorized category, a sensitive category, a modified category, and a normal category.

17. The device according to claim 15, wherein: The privacy label for each image is determined based on at least one of: rules associated with an authorization status of each image, rules associated with sensitive content contained in each image, and rules associated with a modification status of each image.

18. The device according to claim 15, wherein: The image acquisition module is also used for: Determining, based on the identity information of the entity, a candidate entity in an entity index database that matches the entity; and One or more images associated with the candidate entity are obtained from the entity index database.

19. The apparatus according to claim 15, further comprising: The instruction receiving module is used to receive an instruction for managing the one or more images.

20. An apparatus for providing an image, comprising: one or more processors; as well as a memory storing computer executable instructions that, when executed, cause the one or more processors to: Perform authentication of entities; obtaining identity information of the entity based on the authentication; obtaining one or more images associated with the entity based at least on the identity information of the entity; determining a category of the one or more images, wherein the category comprises an unauthorized category, a modified category, a sensitive category, and a normal category; appending a privacy label to each of the one or more images based on the determined category of the one or more images; obtaining a publication chain comprising, for each of the one or more images, a source and a cascading link from the source, the cascading link indicating a path of the image from the source; and The one or more images are provided with the attached privacy label and the publication chain.

Images