A method for generating expression, computing device and storage medium

Abstract

The embodiment of the present application provides a method for generating an expression, a computing device, and a storage medium. In the embodiment of the present application, a fusion model of multiple groups of faces and expression data corresponding to the fusion model of multiple groups of faces are obtained; a fusion model of a target face is obtained, and the corresponding relationship between the fusion model of the target face and the fusion model of multiple groups of faces is determined; based on the corresponding relationship and the expression data corresponding to the fusion model of multiple groups of faces obtained, target expression data corresponding to the fusion model of the target face is generated. Thus, the expression data of the fusion model of the target face can be automatically determined, which saves manpower and improves efficiency. At the same time, since the target expression data is determined by the fusion model of multiple groups of faces and the expression data corresponding to the fusion model of multiple groups of faces, the accuracy of the target expression data can be guaranteed to a certain extent, and the accuracy is also improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method for generating an expression, a method for generating an expression model, a method for virtual live broadcasting, a method for generating virtual customer service, a computing device, and a storage medium. Background Art

[0002] Currently, the design and production of the avatar characters' avatars and expression animations require experienced designers. It takes an experienced designer 3-4 days to produce a set of expressions that match a certain avatar. In order to enrich the diversity of avatars, the avatar's face can be modified to produce a variety of different avatars. However, after changing the avatar's face, the new avatar also needs to be re-adapted to the corresponding expression, which consumes more manpower to design the corresponding expression. Summary of the invention

[0003] Multiple aspects of the present application provide a method for generating expressions, a method for generating expression models, a method for virtual live broadcasting, a method for generating virtual customer service, a computing device, and a storage medium, which are used to quickly generate expression data and save a lot of manpower.

[0004] An embodiment of the present application provides a method for generating facial expressions, comprising: obtaining fusion models of multiple groups of faces and expression data corresponding to the fusion models of the multiple groups of faces; obtaining the fusion model of a target face, and determining the correspondence between the fusion model of the target face and the fusion models of the multiple groups of faces; and generating target expression data corresponding to the fusion model of the target face based on the correspondence and the acquired expression data.

[0005] The embodiment of the present application also provides a method for generating an expression model, including: obtaining fusion models of multiple target faces and corresponding target expression data; for the fusion model of the target face, obtaining a first offset of a preset sampling vertex corresponding to the fusion model of the target face relative to a preset sampling vertex corresponding to a basic face model; for the target expression data, obtaining a second offset of a preset sampling vertex corresponding to the target expression data relative to a preset sampling vertex corresponding to basic expression data, wherein the basic expression data corresponds to the basic face model; for different basic expression data, using the corresponding first offset and the corresponding second offset as training data, training a preset initial expression model, and obtaining expression models for different expressions, so as to generate expression data based on the expression model.

[0006] An embodiment of the present application also provides a method for generating an expression, comprising: acquiring a facial fusion model and a basic facial model to obtain a first offset of a preset sampling vertex; inputting the first offset into a corresponding expression model to obtain a second offset of the preset sampling vertex output by the expression model for corresponding basic expression data; determining position data corresponding to the preset sampling vertex based on the second offset and the corresponding basic expression data to obtain expression data corresponding to the facial fusion model.

[0007] An embodiment of the present application also provides a method for virtual live broadcast, including: obtaining a basic face model and a fusion model of the face provided by a user to obtain a first offset of a preset sampling vertex; inputting the first offset into a corresponding expression model to obtain a second offset of the preset sampling vertex output by the expression model for corresponding basic expression data; determining, based on the second offset and the corresponding basic expression data, position data corresponding to the preset sampling vertex to obtain expression data corresponding to the fusion model of the face; generating a video screen with a corresponding virtual anchor based on the fusion model of the face and the corresponding expression data, and performing video live broadcast.

[0008] An embodiment of the present application also provides a method for virtual live broadcast, including: obtaining fusion models of multiple groups of faces and expression data corresponding to the fusion models of the multiple groups of faces; obtaining a fusion model of a target face provided by a user, and determining the correspondence between the fusion model of the target face and the fusion models of the multiple groups of faces; generating target expression data corresponding to the fusion model of the target face based on the correspondence and the acquired expression data; generating a video screen with a corresponding virtual anchor according to the fusion model of the target face and the target expression data, and performing video live broadcast.

[0009] An embodiment of the present application also provides a method for generating a virtual customer service, including: obtaining a fusion model of multiple groups of faces and expression data corresponding to the fusion model of the multiple groups of faces; obtaining a fusion model of a target face provided by a user, and determining the correspondence between the fusion model of the target face and the fusion models of the multiple groups of faces; based on the correspondence and the acquired expression data, generating target expression data corresponding to the fusion model of the target face; and generating a corresponding image of the virtual customer service for display according to the fusion model of the target face and the target expression data.

[0010] An embodiment of the present application also provides a computing device, including: a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program, so as to: obtain a fusion model of multiple groups of faces and expression data corresponding to the fusion model of the multiple groups of faces; obtain a fusion model of a target face provided by a user, and determine the correspondence between the fusion model of the target face and the fusion models of the multiple groups of faces; based on the correspondence and the acquired expression data, generate target expression data corresponding to the fusion model of the target face; according to the fusion model of the target face and the target expression data, generate a corresponding virtual customer service image for display.

[0011] An embodiment of the present application also provides a computing device, including: a memory and a processor; the memory is used to store computer programs; the processor is used to obtain fusion models of multiple groups of faces and expression data corresponding to the fusion models of the multiple groups of faces; obtain the fusion model of a target face, and determine the correspondence between the fusion model of the target face and the fusion models of the multiple groups of faces; based on the correspondence and the acquired expression data, generate target expression data corresponding to the fusion model of the target face.

[0012] An embodiment of the present application also provides a computing device, including: a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program, so as to: obtain fusion models of multiple target faces and corresponding target expression data; for the fusion model of the target face, obtain a first offset of a preset sampling vertex corresponding to the fusion model of the target face relative to a preset sampling vertex corresponding to a basic face model; for the target expression data, obtain a second offset of a preset sampling vertex corresponding to the target expression data relative to a preset sampling vertex corresponding to basic expression data, wherein the basic expression data corresponds to the basic face model; for different basic expression data, use the corresponding first offset and the corresponding second offset as training data to train a preset initial expression model, and obtain expression models for different expressions, so as to generate expression data based on the expression model.

[0013] An embodiment of the present application also provides a computing device, including: a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program to: obtain a facial fusion model and a basic face model to obtain a first offset of a preset sampling vertex; input the first offset into a corresponding expression model to obtain a second offset of the preset sampling vertex for the corresponding basic expression data output by the expression model; determine the position data corresponding to the preset sampling vertex according to the second offset and the corresponding basic expression data to obtain the expression data corresponding to the facial fusion model.

[0014] An embodiment of the present application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by one or more processors, the one or more processors implement the steps in the above method.

[0015] In the embodiment of the present application, a fusion model of multiple groups of faces and expression data corresponding to the fusion model of multiple groups of faces are obtained; a fusion model of a target face is obtained, and the corresponding relationship between the fusion model of the target face and the fusion model of multiple groups of faces is determined; based on the corresponding relationship and the expression data corresponding to the fusion model of multiple groups of faces, target expression data corresponding to the fusion model of the target face is generated. Thus, the expression data of the fusion model of the target face can be automatically determined, which saves manpower and improves the determination efficiency. At the same time, since the target expression data is determined by the fusion model of multiple groups of faces and the expression data corresponding to the fusion model of multiple groups of faces, the accuracy of the target expression data can be guaranteed to a certain extent, and the accuracy is also improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0017] Figure 1 A schematic diagram of the structure of an expression generation system according to an exemplary embodiment of the present application;

[0018] Figure 2 A schematic flow chart of a method for generating an expression according to an exemplary embodiment of the present application;

[0019] Figure 3 A schematic diagram of model training according to an exemplary embodiment of the present application;

[0020] Figure 4 A schematic diagram of a model application of an exemplary embodiment of the present application;

[0021] Figure 5 A schematic flow chart of a method for generating an expression model according to an exemplary embodiment of the present application;

[0022] Figure 6 A schematic flow chart of a method for generating an expression according to an exemplary embodiment of the present application;

[0023] Figure 7 A flowchart of a method for virtual live broadcast of data according to an exemplary embodiment of the present application;

[0024] Figure 8 A flowchart of a method for virtual live broadcast of data according to an exemplary embodiment of the present application;

[0025] Fig. 9 A schematic diagram of the structure of an expression generation device provided by an exemplary embodiment of the present application;

[0026] Fig.10 A schematic diagram of the structure of a device for generating an expression model provided by an exemplary embodiment of the present application;

[0027] Fig.11 A schematic diagram of the structure of an expression generation device provided by an exemplary embodiment of the present application;

[0028] Fig.12 A schematic diagram of the structure of a computing device provided for an exemplary embodiment of the present application;

[0029] Fig.13 A schematic diagram of the structure of a computing device provided for an exemplary embodiment of the present application;

[0030] Fig.14 A schematic diagram of the structure of a computing device provided for an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0031] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0032] As mentioned above, in order to enrich the virtual image, it can be achieved by modifying the face of the virtual image, but it is also necessary to re-adapt the corresponding expressions to the new virtual image, which will consume a lot of time and manpower.

[0033] Therefore, the embodiment of the present application provides a method for generating expressions, which can automatically and quickly generate expression data, saving a lot of time and manpower.

[0034] Figure 1 The following is a schematic diagram of a system for generating expressions provided by an exemplary embodiment of the present application. Figure 1 As shown, the system 100 may include: a first device 101 and a second device 102 .

[0035] Among them, the first device 101 can be a device with certain computing capabilities, which can realize the function of sending data to the second device 102, can receive data sent by the second device 102, or can not receive data sent by the second device 102. The basic structure of the first device 101 may include: at least one processor. The number of processors may depend on the configuration and type of the device with certain computing capabilities. The device with certain computing capabilities may also include a memory, which may be volatile, such as RAM, or non-volatile, such as read-only memory (ROM), flash memory, etc., or may include both types at the same time. The memory usually stores an operating system (OS), one or more application programs, and may also store program data, etc. In addition to the processing unit and the memory, the device with certain computing capabilities also includes some basic configurations, such as a network card chip, an IO bus, a display component, and some peripheral devices, etc. Optionally, some peripheral devices may include, for example, a keyboard, an input pen, etc. Other peripheral devices are well known in the art and will not be described in detail here. Optionally, the first device 101 may be an intelligent terminal, such as a mobile phone, a desktop computer, a notebook, a tablet computer, etc.

[0036] The second device 102 refers to a device that can provide computing and processing services in a network virtual environment, and can refer to a device that uses the network to generate expression data. In physical implementation, the second device 102 can be any device that can provide computing services, respond to service requests, and perform data processing, such as a cloud server, cloud host, virtual center, conventional server, etc. The second device 102 mainly includes a processor, a hard disk, a memory, a system bus, etc., which is similar to a general computer architecture.

[0037] Specifically, the second device 102 obtains fusion models of multiple groups of faces and expression data corresponding to the fusion models of the multiple groups of faces; obtains the fusion model of the target face, and determines the correspondence between the fusion model of the target face and the fusion models of the multiple groups of faces; based on the correspondence and the obtained expression data, generates target expression data corresponding to the fusion model of the target face.

[0038] In addition, the second device 102 may obtain model data corresponding to the fusion model of the target face corresponding to the determination operation based on the user's determination operation. For example, the first device 101 may send the fusion model of the target face provided by the user to the second device 102, that is, Figure 1 After receiving the fusion model of the target face, the second device 102 may determine the corresponding expression data based on the above method.

[0039] The second device 102 can send the determined expression data to the first device 101, so that the first device 101 displays the expression corresponding to the expression data according to the expression data for display.

[0040] The second device 102 may also store the expression data for future use. For example, when a live video is subsequently broadcast, the virtual image corresponding to the fusion model of the target face provided by the user is used as the video anchor. When the video anchor shows the corresponding expression, the expression data can be obtained to generate the corresponding expression, and then sent to the corresponding smart terminal for display and viewing by the user.

[0041] Specifically, the second device 102 determines a fitting relationship between the fusion model of the target face and the fusion models of the multiple groups of faces, so as to generate target expression data corresponding to the fusion model of the target face based on the fitting relationship and the acquired expression data.

[0042] Specifically, the second device 102 determines the fitting coefficient corresponding to the fusion model of each face according to the fusion model of the target face, the fusion models of multiple faces and the preset weight coefficient corresponding to the fusion model of each face, when the fusion model of the target face is the same as or similar to the fusion model of the fitted face fitted by the fusion models of multiple groups of faces within a preset similarity.

[0043] Specifically, the second device 102 determines the fitting coefficient of the expression data corresponding to the fusion model of each face according to the fitting coefficient corresponding to the fusion model of each face; fits the expression data according to the fitting coefficient of the corresponding expression data to generate target expression data.

[0044] In addition, the second device 102 corrects the target expression data based on the expression data, the fusion model of the target face, and the fusion model of the corresponding face to obtain a corrected target data expression.

[0045] Specifically, the second device 102 determines the difference data between the fusion model of the target face and the fusion model of the corresponding face, and modifies the target expression data according to the difference data to obtain the modified target expression data.

[0046] In addition to the above methods, expressions can also be generated based on models.

[0047] Specifically, the second device 102 obtains a facial fusion model and a basic facial model to obtain a first offset of a preset sampling vertex; inputs the first offset into a corresponding expression model to obtain a second offset of a preset sampling vertex output by the expression model for corresponding basic expression data; determines position data corresponding to the preset sampling vertex based on the second offset and the corresponding basic expression data to obtain expression data corresponding to the facial fusion model.

[0048] The preset sampling vertices are evenly distributed on the face model or the face fusion model.

[0049] In addition, the position data corresponds to part of the expression data, and the second device 102 fills the complete expression data according to the position data to obtain the expression data.

[0050] The second device 102 obtains fusion models of multiple target faces and corresponding target expression data; for each fusion model of the target face, obtains a first offset of a preset sampling vertex corresponding to the fusion model of the target face relative to a preset sampling vertex corresponding to the basic face model; for each target expression data, obtains a second offset of a preset sampling vertex corresponding to the target expression data relative to a preset sampling vertex corresponding to the corresponding basic expression data; for different basic expression data, uses the corresponding first offset and the corresponding second offset as training data, trains a preset initial expression model, and obtains expression models for different expressions, so as to generate expression data based on the expression model.

[0051] In the virtual anchor scene of expression generation in the embodiment of the present application, the user can deform the basic face model 104 on the display interface 103 provided by the first device 101, such as a computer, such as pinching the face, obtain the fusion model 105 of the face, and send the fusion model 105 of the face to the second device 102, such as a server. After the server receives the fusion model 105 of the face provided by the user, it needs to generate expressions for the fusion model 105 of the face. The generation method may include the following two methods:

[0052] After receiving the face fusion model 105, the server determines the fitting coefficients corresponding to the multiple groups of face fusion models according to the received face fusion model 105, so that the fitted face fusion model is the same or similar to the received face fusion model 105. Then, according to the determined fitting coefficient, the fitting coefficient of the corresponding expression data is determined, and according to the fitting coefficient of the corresponding expression data, the expression data is fitted, so that the target expression data of the provided face fusion model 105 can be generated, so that the corresponding expression can be determined.

[0053] In addition, after receiving the fusion model 105 of the face, the server can first determine the first offset of the preset sampling vertex corresponding to the fusion model 105 of the face relative to the preset sampling vertex corresponding to the basic face model 104, and then input the first offset into the preset expression model, so as to obtain the second offset output by the expression model, which refers to the offset of the preset sampling vertex for the corresponding basic expression data, such as the smiling expression data. According to the second offset and the corresponding basic expression data, the position data of the preset sampling vertex, such as the three-dimensional coordinates in space, can be determined. If the preset sampling vertex is all the vertices of the face model (or the fusion model of the face), the position data is the expression data corresponding to the fusion model 105 of the face. If the preset sampling vertex is part of the vertices of the face model (or the fusion model of the face), the position data is part of the expression data of the fusion model 105 of the face. At this time, it is necessary to continue to fill the part of the expression data according to the position data to obtain the complete expression data, that is, the expression data of the fusion model 105 of the face. The server can return the expression data to the computer for the user to watch. It can also be directly saved, or the expression data and the fusion model 105 of the face can be provided to other third-party servers, and other third-party servers can provide expression data and the fusion model 105 of the face to the corresponding third parties. So that when the application is carried out later, the expression data can be directly obtained from the corresponding server to generate expressions for viewing. For example, the user can initiate a live video broadcast, and use the virtual image or virtual character corresponding to the fusion model 105 of the face as a virtual anchor. When the virtual anchor needs to generate a corresponding expression, the server can directly obtain the corresponding expression data and generate the corresponding expression, and send the expression to each smart terminal, such as a mobile phone, to display the expression. Alternatively, the expression data is directly sent by the server to each smart terminal, such as a mobile phone, and the mobile phone generates an expression based on the expression data for display.

[0054] In addition, the application scenario of the embodiment of the present application can also be in the application scenario of virtual customer service. Virtual customer service can be used in industries such as e-commerce, finance, and medical care. For example, according to the above generation method, the server can generate the final expression data of the facial fusion model 105, and can return the expression data to the computer for the user to view. It can also be saved directly, or the expression data and the facial fusion model 105 can be provided to other third-party servers, and the other third-party servers can provide the expression data and the facial fusion model 105 to the corresponding third parties.

[0055] When the subsequent application scenario of e-commerce virtual customer service is being carried out, for example, a user can send a message to the customer service of the merchant M of the e-commerce platform through the instant messaging service of the e-commerce platform application of an intelligent terminal, such as a mobile phone or a computer. The user's mobile phone can provide an interface for the instant messaging service, and the user can input on the interface and send the message to the server corresponding to the customer service by the mobile phone. The server generates a corresponding reply message based on the message (the reply message can also be selected or determined manually. When the message is manually determined, the server can send the message to the mobile phone or computer of the corresponding manual customer service for display, so that the manual customer service replies and can send the reply message to the server), such as selecting the corresponding reply message from multiple preset reply messages. Then the server generates an expression corresponding to the reply message based on the reply message, the preset expression data (that is, the final expression data generated by the above method) and the fusion model 105 of the face, and the corresponding expression data can be selected from multiple expression data. Alternatively, based on each preset reply message, the corresponding expression data is preset, and the corresponding expression data is directly selected when the reply message is selected. The server can send the reply message, the corresponding expression data and the fusion model 105 of the face to the user's mobile phone and display them through the interface. The user can not only see the reply message on the interface, but also see the image or image of the corresponding virtual customer service (i.e., the fusion model 105 of the face) and the corresponding expression on the interface. The image or image of the virtual customer service can be a simple image or animation, or a short video.

[0056] When the subsequent application scenario of financial virtual customer service is being carried out, for example, a user can send a message to the customer service of the financial platform through the instant messaging service of the financial platform application of the smart terminal, such as a mobile phone or a computer. The user's mobile phone can provide an interface for the instant messaging service, and the user can input on the interface and send the message to the server corresponding to the customer service by the mobile phone. The server generates a corresponding reply message based on the message (the reply message can also be selected or determined manually. When the message is manually determined, the server can send the message to the mobile phone or computer of the corresponding manual customer service for display. Thus, the manual customer service replies and can send the reply message to the server), such as selecting the corresponding reply message from multiple preset reply messages. Then the server generates an expression corresponding to the reply message based on the reply message, the preset expression data (that is, the final expression data generated by the above method) and the fusion model 105 of the face, and the corresponding expression data can be selected from multiple expression data. Alternatively, based on each preset reply message, the corresponding expression data is preset, and the corresponding expression data is directly selected when the reply message is selected. The server can send the reply message, the corresponding expression data and the fusion model 105 of the face to the user's mobile phone and display them through the interface. The user can not only see the reply message on the interface, but also see the image or image of the corresponding virtual customer service (i.e., the fusion model 105 of the face) and the corresponding expression on the interface. The image or image of the virtual customer service can be a simple image or animation, or a short video.

[0057] When performing subsequent application scenarios of online medical consultation virtual customer service, for example, users can send messages to online doctors through the instant messaging service of the disease consultation platform application program of a smart terminal, such as a mobile phone or a computer. The user's mobile phone can provide an interface for the instant messaging service, on which the user can input, and send messages to the server corresponding to the doctor by the mobile phone, and then the server sends the message to the corresponding doctor's mobile phone for display, so that the online doctor replies based on the message and can send the reply message to the server. Alternatively, the server can also generate a corresponding reply message based on the message, such as selecting a corresponding reply message from a plurality of preset reply messages. Then the server generates an expression corresponding to the reply message based on the reply message, the preset expression data (i.e., the final expression data generated by the above method) and the fusion model 105 of the face, and the corresponding expression data can be selected from a plurality of expression data. Alternatively, based on each preset reply message, the corresponding expression data is preset, and the corresponding expression data is directly selected when the reply message is selected. The server can send the reply message, the corresponding expression data and the facial fusion model 105 to the user's mobile phone, and display them through the interface. The user can not only see the doctor's reply message on the interface, but also see the corresponding virtual doctor's image or image (i.e., the facial fusion model 105, which can be a fusion model of the doctor's face) and the corresponding expression on the interface. The image or image of the virtual doctor can be a simple image or animation, or a short video. It should be noted that for the above-mentioned different application scenarios, the implementation method of the specific steps involved in generating the target expression data can be introduced through the detailed description of the subsequent method, which will not be elaborated here.

[0058] In the above embodiment, the first device 101 and the second device 102 are connected to a network, and the network connection may be a wireless connection. If the first device 101 and the second device 102 are connected by communication, the network standard of the mobile network may be any one of 2G (GSM), 2.5G (GPRS), 3G (WCDMA, TD-SCDMA, CDMA2000, UTMS), 4G (LTE), 4G+ (LTE+), WiMax, 5G, etc.

[0059] The following is a detailed description of the expression generation process in conjunction with the method embodiment.

[0060] Figure 2 The flowchart of a method for generating an expression according to an exemplary embodiment of the present application is shown in FIG. The method 200 provided in the embodiment of the present application is executed by a computing device, such as a server. The method 200 includes the following steps:

[0061] 201: Obtain fusion models of multiple groups of faces and expression data corresponding to the fusion models of multiple groups of faces.

[0062] 202: Obtain a fusion model of the target face, and determine a correspondence between the fusion model of the target face and fusion models of multiple groups of faces.

[0063] 203: Generate target expression data corresponding to the fusion model of the target face based on the corresponding relationship and the acquired expression data.

[0064] It should be noted that the method 200 can also be implemented through an intelligent terminal, such as a computer.

[0065] The following is a detailed description of the above steps:

[0066] 201: Obtain fusion models of multiple groups of faces and expression data corresponding to the fusion models of multiple groups of faces.

[0067] The fusion model of the face refers to a deformation based on the basic face model. The deformation method refers to a method for changing the facial features and / or face shape in the face model, which may include but is not limited to the face pinching method. The basic face model is preset, and may be set by presetting the face shape and presetting the facial features. The face shape and facial features may be of different types, such as the types of face shapes may include: round face, long face, oval face, etc. The types of facial features, wherein the types of eyes may include: round eyes, phoenix eyes, etc., will not be described one by one for the facial features here. A basic face model may be preset based on this. It may also be redesigned based on this, such as a preset standard face shape, standard eyes, standard nose and standard mouth, and this standard facial features and face shape may be set according to a standard ratio, such as the golden ratio, etc. The fusion model of the face may change the face shape and / or facial features of the basic face model, such as replacing or modifying the standard eyes in the basic face model to phoenix eyes, etc.

[0068] The expression data is the expression data corresponding to the fusion model of the face, and the expression data can generate the corresponding visual expression. The fusion model of the face and the expression data can be obtained by artificial design, or they can be pre-set models and data, and stored in the corresponding device for acquisition. Correspondingly, the basic face model also has corresponding basic expression data, which will not be described here.

[0069] Specifically, obtaining fusion models of multiple groups of faces and expression data corresponding to the fusion models of multiple groups of faces includes: obtaining a basic face model and a fusion model of multiple groups of faces obtained by deforming the basic face model; obtaining basic expression data corresponding to the basic face model and expression data obtained by the fusion model of the corresponding face.

[0070] For example, according to the above, a designer user can design and create a basic face model and corresponding basic expressions (corresponding to basic expression data) on a design software interface provided by a smart terminal, such as a computer. The basic face model can correspond to 54 basic expressions, such as a smiling expression, a laughing expression, a closed-eye expression, etc. The interface can provide images of the face shape and facial features required for designing and creating models and expressions. The designer user can design and create on the interface. The computer responds to the designer user's creation operation to generate a basic face model and corresponding basic expressions, such as Figure 3 As shown, the basic face model 301 and the corresponding basic expression 302. It should be understood that both the basic face model and the basic expression are supported and implemented by data, so obtaining the model and the expression can also be regarded as obtaining the model data (i.e., the model) and the expression data. The computer can continue to respond to the sending operation of the designer user and send the generated basic face model and the corresponding basic expression expression data to the server. So that the server can store it. The designer user can also continue to pinch the basic face model on this interface, and can obtain 10 groups of face models after pinching, that is, the fusion model of multiple groups of faces, such as 1 group: long face + phoenix eyes + lion nose + thick lips, 2 groups: round face + almond eyes + standard nose + thin lips, etc., a total of 10 groups. After the designer user has produced the fusion model of 10 groups of faces, he continues to produce corresponding expression data, i.e., expression data, for the fusion model of these 10 groups of faces. The fusion model of each group of faces also has 54 expression data. Since the face fusion model is deformed based on the basic face model, the basic expression data can also be used as a reference to create expression data, that is, the expression type corresponding to each group of face fusion models is the same as the expression type corresponding to the basic expression data, such as smiling expression, laughing expression, closed eyes expression, etc. The computer can respond to the production operation of the designer user to generate 10 groups of face fusion models and corresponding deformed expressions (i.e. corresponding to the expression data), such as Figure 3 As shown in FIG. 3 , there are a plurality of fusion models 303 of faces and corresponding deformed expressions 304 (where each row is a group of fusion models and deformed expressions of faces). The computer can also respond to the sending operation of the designer user and send the fusion models of 10 groups of faces and the corresponding deformed expressions (i.e., the corresponding expression data) to the server. The server can obtain 10*54 expressions (i.e., the corresponding expression data) and the corresponding fusion models of 10 faces, as shown in FIG. Figure 3 As shown, the obtained data can be used as a database 305, and the database 305 has the above data, such as a fusion model 306 of the face and a corresponding deformed expression 307.

[0071] In addition, according to the previous discussion, in addition to designer users designing these expressions and facial fusion models, they can also obtain preset facial fusion models and corresponding expressions from other devices.

[0072] 202: Obtain a fusion model of the target face, and determine a correspondence between the fusion model of the target face and fusion models of multiple groups of faces.

[0073] Wherein, obtaining the fusion model of the target face includes: based on a determination operation of the user, obtaining model data corresponding to the fusion model of the target face corresponding to the determination operation.

[0074] 1) It may be to obtain model data corresponding to the fusion model of the target face provided by the user.

[0075] For example, according to the above, the user of the live video broadcasting party can pinch the basic face model on the live broadcast client on the computer, such as the interface provided by the web client, to obtain the fusion model of the target face. The computer can respond to the user's creation operation and obtain the fusion model of the target face, such as long face + round eyes + standard nose + cherry lips. And in response to the user's sending operation, the fusion model of the target face is sent to the server so that the server can obtain and store it.

[0076] 2) It may be possible to provide a plurality of preset face fusion models to the user, and determine the model data corresponding to the corresponding preset face fusion model according to the user's determination operation (or selection operation).

[0077] For example, according to the above, the server can provide multiple preset face fusion models to the live client on the computer of the user of the live video broadcast party, such as the interface provided by the web client. The user selects or determines the multiple preset face fusion models to determine the corresponding preset face fusion model. In response to the operation, the web client can send the number of the determined preset face fusion model to the server, so that the server determines the corresponding preset face fusion model.

[0078] The step of obtaining the fusion model of the target face includes obtaining a deformation coefficient and generating the fusion model of the target face according to the deformation coefficient and the basic face model.

[0079] Among them, the deformation coefficient can determine the final target face fusion model, that is, the deformation coefficient can determine the deformation trend of the face model. The coefficient can refer to the face pinching coefficient. The basic face model can be used as a benchmark to set its face pinching coefficient to change the basic face model, that is, to pinch the face. Among them, the face pinching coefficient can include one or more, each face pinching coefficient can determine at least part of the area change of the basic face model, such as changes in the eyes, changes in the nose, changes in the position of the temples, etc., and different deformation coefficients, such as different face pinching coefficients, can be used according to needs to determine the appearance of different areas or different positions, and each area or each position corresponds to a coefficient.

[0080] For example, according to the above, the user of the live video broadcasting party can directly select the corresponding face pinching coefficient on the interface of the above client. In response to the user's selection operation, the computer determines the face pinching coefficient and sends the face pinching coefficient to the server, which pinches the basic face model according to the face pinching coefficient to obtain a fusion model of the target face.

[0081] In addition, the computer can also perform face pinching according to the face pinching coefficient to obtain a fusion model of the target face and send it to the server. In addition, both the computer and the server can directly modify the data corresponding to the basic face model to achieve the purpose of modifying the basic face model.

[0082] In addition to determining the fusion model of the target face based on the above-mentioned deformation coefficients, the user can also directly perform editing operations on the displayed basic face model, such as stretching and modifying the lines of the model, to obtain the model data of the corresponding fusion model of the target face, and send the model data to the server so that the server obtains the model data of the fusion model of the target face.

[0083] The basic face model can also be modified based on the user's requirements for the fusion model of the target face. The user can record the description information of the fusion model of the target face in text or voice. For example, the fusion model of the target face is the description information of big eyes, standard nose and long face. The client sends the description information to the server, so that the server can modify the basic face model according to the description information.

[0084] Specifically, determining the correspondence between the fusion model of the target face and the fusion models of the multiple groups of faces includes: determining the fitting relationship between the fusion model of the target face and the fusion models of the multiple groups of faces, so as to generate target expression data corresponding to the fusion model of the target face based on the fitting relationship and the acquired expression data.

[0085] The corresponding relationship may be a fitting relationship.

[0086] It should be noted that, in addition to determining the target expression data through the fitting relationship, it can also be determined through other corresponding relationships, such as mapping relationships, or determining the difference data between the fusion model of the target face and the fusion models of multiple groups of faces, which can be the difference data between the fusion model of any group of faces, and then correcting the expression data corresponding to the fusion model of the group of faces based on the difference data, so as to obtain the target expression data corresponding to the fusion model of the target face, and the data can have multiple, such as 54.

[0087] Specifically, determining the fitting relationship between the fusion model of the target face and the fusion models of the multiple groups of faces includes: when the fusion model of the target face is the same as the fusion model of the fitted face fitted by the fusion models of the multiple groups of faces or is similar within a preset similarity, determining the fitting coefficient corresponding to the fusion model of each face according to the fusion model of the target face, the fusion models of the multiple faces and the preset weight coefficient corresponding to the fusion model of each face.

[0088] For example, according to the foregoing, the server may determine the fitting coefficient according to the following formula 1):

[0089]

[0090] Among them, E represents the energy function that is optimized and solved to minimize it. When it is 0, the fusion model of the target face is the same as the fusion model of the fitted face fitted by the fusion model of multiple groups of faces. When it is not 0 but the minimum value, the fusion model of the target face is

[0091] The fusion model of the fitted face obtained by fitting the model and the fusion model of multiple groups of faces is similar within a preset similarity; v Indicates the preset weights of different vertices v on the face model (it can be a fusion model of the face, expression data. No matter which fusion model of the face is used, its topological structure is the same, it is composed of polygonal surfaces, and the composition method is also the same). The greater the change in the action, the higher the corresponding vertex weight; a i represents the fitting coefficient of the solution; Represents the value of vertex v of the fusion model (position 1) of the face corresponding to the i-th group of expression data in the data, Indicates M deform The value of the vertex position v, n g is the number of fusion models of multiple groups of faces. And,

[0092] It should be understood that for w vFor E, it is not necessary to determine E. It is just that when the expression changes, the parts of the face model with larger changes in movement can affect the accuracy of the expression data more. In order to improve this accuracy, this weight is introduced. Even if this weight is not introduced, the target expression data can still be determined, that is, the fitting coefficient can be determined.

[0093] In addition, the fusion model M of the fitted face D As shown in the following formula 2):

[0094]

[0095] 203: Generate target expression data corresponding to the fusion model of the target face based on the corresponding relationship and the acquired expression data.

[0096] Specifically, based on the corresponding relationship and the acquired expression data, target expression data corresponding to the fusion model of the target face is generated, including: determining the fitting coefficient of the expression data corresponding to the fusion model of each face according to the fitting coefficient corresponding to the fusion model of each face; fitting the expression data according to the fitting coefficient of the corresponding expression data to generate the target expression data.

[0097] For example, according to the above, the server can determine the target expression data according to the following formula 3):

[0098]

[0099] in, represents the target expression data labeled j, That is M D It should be noted that it is necessary to make n g The fusion model of each group of faces, the different expression data contained in the fusion model of each group of faces, and the fusion model of the faces in the group, the total number is n exp Indicates that the total number of n g *n exp data (including expression data and the corresponding facial fusion model), the data is arranged as follows: i，j , i=1…n g j=1…n exp , D i， : means taking the i-th row in the data (i.e., it is necessary to make the fusion model of the i-th group of faces and all expression data), D ：，j It means taking the jth column of expression data, that is, all the combinations of deformed expression j (j can be a smiling expression); Indicates D i，j The vertex numbered v in the data.

[0100] After receiving the fusion model of the target face (i.e., Mdeform ), 54 target expression data corresponding to the fusion model of the target face can be obtained from the fusion model of the 10*54 deformed expressions and the face corresponding to each group of 54 deformed expressions.

[0101] However, due to the limited number of expression data and the corresponding facial fusion models, there may still be a deviation between the final fitted expression data and the actual expression data corresponding to the fusion model of the target face. Therefore, in order to obtain the final expression data more accurately, the fitted target expression data can be corrected.

[0102] Specifically, the method 200 further includes: based on the expression data, the fusion model of the corresponding face and the fusion model of the target face, correcting the target expression data to obtain a corrected target data expression.

[0103] Among them, based on the expression data, the fusion model of the target face and the fusion model of the corresponding face, the target expression data is corrected, including: determining the difference data between the fusion model of the target face and the fusion model of the corresponding face; correcting the target expression data according to the difference data to obtain the corrected target expression data.

[0104] For example, according to the above, after obtaining the fitted target expression data, the server can modify or detect the target expression data. According to the above, the server can fit the fusion models of the 54 faces corresponding to the 54 sets of expression data according to the above fitting method. Thus, the fitted face fusion model M can be obtained. D The fitted face fusion model M D Fusion model M with the target face deform (i.e., the fusion model of the target face) to compare the data and determine the difference data between the two. The difference data is used to fit the target expression data Exp D ,like Make corrections and get the corrected target expression data Exp deform .

[0105] The difference data may refer to the change data between the corresponding polygonal faces in the models of the two faces, such as the fusion model M of the fitted face. D Fusion model M with the target face deform The topological structure is the same, with a total of n f triangular faces, for M D Transform to M deform , for example, the i-th triangle face F i (v1, v2, v3) produces the following spatial transformation and in, The spatial rotation matrix, The spatial translation matrix makes F i (v1, v2, v3) transform to position Then set the matrix parameters V and V d ,in,

[0106] V = [v2-v1, v3-v1, v4-v1]

[0107]

[0108]

[0109] Among them, v4 is the vertex, which is the inserted virtual vertex in the fusion model M of the face. D On the triangle face F i (v1, v2, v3), we can calculate:

[0110]

[0111] Similarly, for the target expression data Transform to the corrected target expression data Exp deform , the rotation matrix of the i-th triangular face is expressed as The translation matrix of the i-th triangular face is expressed as The following energy equation E needs to be optimized st :

[0112]

[0113] And satisfy the constraints:

[0114]

[0115] Among them, i, j and k represent the shared vertex v j The set of all polygonal faces.

[0116]

[0117] Among them, E p Represents the vertex pair constraint energy, which is used to constrain the selected vertex group (c i , d i ) are at the same distance in space; (c i , d i) is a group of points, which are manually selected and are different points on the same face model (e.g., the fusion model of the face); they can be the vertices of the eyes and eyelashes on the face model. This group of points should keep the same relative position (e.g., eyelashes are attached to eyelids). In the program, the distance between this group of points after expression is limited to 0.

[0118] By optimizing the energy term:

[0119] E=E p +E st

[0120] The energy E is the smallest, and we can get

[0121] For example, the fusion model M of the face D is a round face, and the fusion model M of the target face deform If it is a long face, then according to the above method, for the fitted target expression data Exp D Correction of face shape.

[0122] In addition, it is also possible to decide whether to make corrections based on the degree of difference or similarity.

[0123] Specifically, the method 200 further includes: determining the difference between the fusion model of the target face and the fusion model of the corresponding face, and when the difference is greater than a threshold, executing the step of correcting the target expression data according to the difference data.

[0124] For example, according to the above, the fusion model M of the face is determined D The difference between the target face model (i.e., the fusion model of the target face) and the target face model can be determined by whether the position coordinates of the corresponding vertices in the two fusion models are the same. For example, the difference can be determined by the ratio of the number of corresponding vertices with the same position coordinates to the total number of vertices in the fusion model. If it is greater than a threshold, correction can be made, otherwise no correction is required.

[0125] In addition, the vector features of the fusion model of the two faces can be determined first, and the difference can be determined by determining the similarity between the vector features. The similarity can be determined by a similarity algorithm, such as Euclidean distance. When the similarity is less than a certain threshold, or the difference is greater than a certain threshold, correction is required, otherwise no correction is required.

[0126] It should be noted that when the amount of expression data is large enough, if a fusion model of a face that is similar to or the same as the fusion model of the target face can be directly found from the fusion model, the expression data corresponding to the fusion model of the face can be directly used as the target expression data. In addition, in order to save the storage capacity of the server, the data can be firstly processed before fitting the target expression data, such as a total of n g *n exp For example, PCA (principal components analysis) compresses the above-mentioned facial fusion model and the corresponding expression data, and the compressed data

[0127] The database is then formed. Then the subsequent target expression data is generated. PCA can extract the main components from the redundant features of the data, which improves the model training speed without sacrificing the quality of the model after subsequent training.

[0128] In addition, in order to reduce the amount of data in the above database, reduce the size of the data packet and speed up the operation, the expression data can also be generated by training the model on this basis. The specific implementation method is as follows;

[0129] The model is trained as follows:

[0130] Specifically, the method 200 also includes: obtaining fusion models of multiple target faces and corresponding target expression data; for each fusion model of the target face, obtaining a first offset of a preset sampling vertex corresponding to the fusion model of the target face relative to a preset sampling vertex corresponding to the basic face model; for each target expression data, obtaining a second offset of a preset sampling vertex corresponding to the target expression data relative to a preset sampling vertex corresponding to the corresponding basic expression data; for different basic expression data, using the corresponding first offset and the corresponding second offset as training data, training the preset initial expression model, and obtaining expression models for different expressions, so as to generate expression data based on the expression model.

[0131] The preset initial expression model may be a neural network model, such as a CNN (Convolutional Neural Networks) model. Figure 3 As shown, the CNN model 308 may include two fully connected layers + ReLU linear rectified function (Rectified Linear Unit).

[0132] The preset sampling vertices refer to the sampling vertices for the face model (such as the basic face model or the fusion model of the face). It should be noted that the face model is a 3D face model, which can be composed of multiple polygonal faces, such as multiple triangular faces, quadrilateral faces, etc., as well as a mixture of triangular faces and quadrilateral faces. The vertices of each polygonal face can be the vertices of the face model. Multiple vertices can be randomly selected from the face model as preset sampling vertices. It should be understood that for the embodiments of the present application, the composition form of the polygonal faces of the face models involved is the same, and the sizes of the corresponding polygonal faces may be different.

[0133] In order to save the running time of the server and improve its running performance, a preset number (eg, 300) of vertices may be uniformly selected from the face model as preset sampling vertices, that is, the preset sampling vertices are uniformly distributed on the face model.

[0134] In addition, each type of expression corresponds to a preset initial expression model, such as a smile expression corresponds to a preset initial expression model, and a closed-eye expression corresponds to a preset initial expression model, but the training method of each model is similar. However, due to the different types of expressions, the target expression data is also different. For example, for the expression model corresponding to the smile expression, its target expression data is the expression data for the smile expression, and the second offset is the offset of the preset sampling vertex in the expression data of the smile expression relative to the preset sampling vertex in the basic smile expression data.

[0135] For example, according to the foregoing, the server can obtain the fusion models of multiple target faces and the corresponding target expression data, that is, the fitted expression data, in the above manner, and can have 54 target expression data. Then, for each target face fusion model, the server obtains the data corresponding to the preset sampling vertices from the target face fusion model based on the preset sampling vertices, and then obtains the data corresponding to the preset sampling vertices from the basic face model. Based on the data corresponding to the preset sampling vertices of the two face models, the server determines the offset corresponding to the preset sampling vertex as the first deviation. Correspondingly, the server obtains the second offset based on the same method. Among them, for different types of expressions, that is, each target expression data corresponding to the fusion model of the same target face, and different basic expression data corresponding to the basic face model. Obtain the corresponding second offset, and for each second offset, the expression type of the target expression data and the basic expression data is the same, such as a smile expression.

[0136] Take a fusion model of a target face as an example: each target expression data of the fusion model of the target face will correspond to a set of first offsets and second offsets, so the same expression type of the fusion model of multiple target faces, such as a smiling expression, will have multiple sets of first offsets and second offsets for the corresponding target expression data, such as smiling expression data. The multiple sets of first offsets and second offsets of the same expression type are used as the preset initial expression model, such as the training data of the initial CNN model, to perform model training to obtain the final trained CNN model. The specific model training process is existing technology and will not be repeated here. The training method of the preset initial expression model corresponding to other expression types is also as described above and will not be repeated here.

[0137] It should be noted that the first offset is used as input training data of the initial CNN model. The second offset is used as output training data of the initial CNN model. Both the first offset and the second offset are offsets corresponding to multiple preset sampling vertices.

[0138] In addition, in order to more directly and conveniently determine the target expression data corresponding to the fusion model of the target face, the sampling vertices can be all vertices of the face model (such as the basic face model or the fusion model of the face). That is, the preset sampling vertices include all vertices on the face model. The corresponding model training process is also similar to the training process described above, and will not be repeated here.

[0139] After the model is trained, the target expression data corresponding to the fusion model of the target face can be obtained through the model.

[0140] Specifically, the method 200 further includes: obtaining expression data corresponding to the fusion model of the target face according to the fusion model of the target face and the corresponding expression model, as target expression data.

[0141] When the server subsequently applies the trained model, it can obtain the first offset of the preset sampling vertices in the fusion model of the target face relative to the preset sampling vertices in the basic face model according to the aforementioned processing method, and then input the first offset into the trained model to obtain the second offset, so as to obtain the final expression data.

[0142] Specifically, expression data corresponding to the fusion model of the target face is obtained according to the fusion model of the target face and the corresponding expression model, including: obtaining the offset of the preset sampling vertex according to the fusion model of the target face and the basic face model provided by the user, and inputting it into the corresponding expression model to obtain the offset of the preset sampling vertex output by the expression model for the corresponding basic expression data; determining the position data corresponding to the preset sampling vertex according to the obtained offset and the corresponding basic expression data, the position data corresponding to the partial expression data of the fusion model of the target face; filling the complete expression data according to the position data to obtain the expression data.

[0143] Among them, the offset of the preset sampling vertex obtained according to the fusion model of the target face and the basic face model provided by the user can be regarded as the first offset, which is input into the corresponding expression model, and the offset of the preset sampling vertex for the corresponding basic expression data output by the expression model can be regarded as the second offset.

[0144] For example, according to the above, after the server has trained the CNN model, in the subsequent model application process, the server can receive the fusion model of the target face sent by the computer, and the fusion model is generated by the user of the live video broadcast party through the computer. Figure 4 As shown, the server can directly input the fusion model 401 of the target face into the trained CNN model 402. After receiving the fusion model of the target face, the CNN model first determines the first offset of the preset sampling vertices in the fusion model of the target face relative to the preset sampling vertices in the basic face model. Then the first offset is input into the CNN model to obtain the second offset, that is, the offset of the preset sampling vertices in the basic expression data. After obtaining the second offset, the position data of the final preset sampling vertices can be determined according to the basic expression data and the second offset. When the preset sampling vertices belong to all vertices of the face model (such as the basic face model or the fusion model of the face), the expression data 403 can be directly determined according to the final position data. When the preset sampling vertices belong to some vertices of the face model, such as some vertices uniformly obtained, after determining the position data, the data needs to be filled in, and the remaining expression data is filled in to obtain the final expression data. For example, Laplacian deformation is applied to the above 300 vertex positions, and the Laplacian coordinates of the vertices before deformation are equal to the Laplacian coordinates of the vertices after deformation. The Laplacian coordinates can contain detailed information in the face model, and the detailed information is not desired to be changed after deformation. Other vertex positions are generated thereby.

[0145] After obtaining the expression data, the server can store the expression data for direct retrieval when it is used later.

[0146] Specifically, the method 200 further includes: generating corresponding expressions according to the target expression data for display.

[0147] For example, as described above, after the server receives a request from a user of the live video broadcasting party to broadcast live, the server can broadcast live video of the virtual anchor according to the fusion model of the target face and the target expression data stored in the server. The corresponding virtual anchor image is generated according to the fusion model of the target face and the target expression data, so as to broadcast live video.

[0148] Based on the above embodiment, the user can log in to the cloud (or cloud server) interface through the web client or application of the smart terminal, and upload the fusion model of the target face from the local through the web client or application, or send information that can generate the fusion model of the target face to the cloud through other methods provided by the cloud interface of the web client or application (please refer to the implementation method described above). The cloud can generate the target expression data in the above manner according to the determined fusion model of the target face. The cloud then returns the target expression data to the user's cloud interface to show it to the user. It can also be stored in the cloud (i.e., stored on the cloud server). The generation process of the expression data has been explained above and will not be repeated here.

[0149] After obtaining the fusion model of the target face and the target expression data, the user can display the fusion model of the target face and the target expression data received through an intelligent terminal, such as a computer or a mobile phone. When the user modifies the fusion model of the target face, the fusion model of the target face can be modified and adjusted by the above-mentioned modification or adjustment method, and the modified fusion model of the target face is sent to the server. When the server receives the fusion model of the target face modified by the user, it modifies the corresponding target expression data according to the modified fusion model of the target face (it can also be modified directly by the intelligent terminal). The modification method can also be modified in the manner described above, which will not be repeated here.

[0150] Alternatively, when the user modifies the fusion model of the target face through the smart terminal, the smart terminal can display a reminder message to the user in response to the user's modification operation to remind the user whether to modify the target expression data corresponding to the fusion model of the target face, and when the user determines to synchronously modify the target expression data, the smart terminal can directly modify the corresponding target expression data in response to the confirmation operation, or send a modification instruction to the server in response to the confirmation operation, and the modification instruction can carry the modified fusion model of the target face, so that the server modifies the corresponding target expression data according to the modified fusion model of the target face.

[0151] In addition, based on the generation method in the embodiment of the present application, after determining the fusion model of the target face and the target expression data, the server can use the fusion model of the target face and the target expression data to create a virtual person. For example, the server can generate a corresponding virtual person based on the fusion model of the target face and the target expression data for display, bringing fun to the user. It can introduce information to the user, answer questions, and help the user with emotional healing, etc.

[0152] Or it can be used for audio and video production of images and video content. For example, the server can add the corresponding fusion model of the target face and the expression of the target expression data to the preset video according to the fusion model of the target face and the target expression data, so that an image or animation of a virtual person can be formed on the preset video, and dubbing can also be provided to vividly express the virtual image. It is also possible to directly form an image or animation of a virtual person according to the fusion model of the target face and the expression of the target expression data, and dubbing can also be provided to generate the first short video, etc.

[0153] In addition, it can also help users generate expressions under different expressions to increase the entertainment of social scenes. For example, in the application scenario of instant messaging, the user's smart terminal, such as a mobile phone, can generate expressions such as haha, hehe, hehe, etc. according to the fusion model of the target face and the target expression data (here, the fusion model of the target face and the target expression data can be stored in the smart terminal, such as a mobile phone), and these expression data can be used for reply. The user can input information on the instant messaging interface, and the mobile phone can match the corresponding target expression data based on the keywords or emotional words of the input information, such as happy, sad, etc. (the target expression data can have a preset corresponding relationship with these words). Then, the mobile phone can display the corresponding expression based on the corresponding target expression data. At the same time, the corresponding animated expression can also be generated based on the fusion model of the target face and the corresponding target expression data for display. When the user likes the expression or animation, he can directly click to send. The mobile phone can also directly generate corresponding expressions based on different expression data for display for users to choose for message reply.

[0154] It should be noted that the embodiment of the present application can not only acquire expression data from a face model (such as a fusion model of a target face), but can also modify the body of the virtual image, such as modifying the height, fatness, and other details of the virtual image. The modification method can be similar to the method of pinching the face to modify or replace.

[0155] Then, the body shape or the body movements and behaviors can be determined in a similar manner as described above, and corresponding data can be determined, such as adjusting the vertex positions of the body model according to the body skeleton, thereby adjusting the body movements and behaviors. This will not be described in detail here.

[0156] It should also be noted that the specific implementation steps of the target expression data generation process in the above-mentioned application scenarios, such as the virtual anchor scenario and the virtual customer service scenario, have been described in detail in the method 200.

[0157] Based on the above similar inventive concept, Figure 5 A flow chart of a method for generating an expression model provided by another exemplary embodiment of the present application is shown. The method 500 provided by the embodiment of the present application is executed by a server, and the method 500 includes the following steps:

[0158] 501: Obtain fusion models of multiple target faces and corresponding target expression data.

[0159] 502: For the fusion model of the target face, obtain a first offset of a preset sampling vertex corresponding to the fusion model of the target face relative to a preset sampling vertex corresponding to the basic face model.

[0160] 503: For the target expression data, obtain a second offset of a preset sampling vertex corresponding to the target expression data relative to a preset sampling vertex corresponding to the basic expression data, where the basic expression data corresponds to the basic face model.

[0161] 504: For different basic expression data, the corresponding first offset and the corresponding second offset are used as training data to train the preset initial expression model to obtain expression models for different expressions, so as to generate expression data based on the expression model.

[0162] Since the specific implementation of steps 501-504 has been described in detail above, they will not be repeated here. It is only explained that the method 500 can also be applied to smart terminals, such as computers.

[0163] In addition, the method 500 further includes: based on the basic face model, randomly generating a corresponding random face fusion model as the target face fusion model.

[0164] Since it has been explained in the previous article, it will not be repeated here. It is only explained that the method of randomly generating the corresponding random face fusion model can be generated by the deformation coefficient mentioned in the previous article. The deformation coefficient can be randomly selected to achieve the purpose of randomly generating the random face fusion model, which will not be repeated.

[0165] In addition, the method 500 further includes: obtaining a fusion model of the face provided by the user, and obtaining expression data corresponding to the fusion model of the provided face according to the fusion model of the provided face and the corresponding expression model. As it has been explained above, it will not be repeated here. It is only explained that the user can generate a fusion model of the face that meets his needs in the manner described above, so as to subsequently determine the expression data corresponding to the fusion model of the provided face.

[0166] Specifically, based on the provided face fusion model and the corresponding expression model, expression data corresponding to the provided face fusion model is obtained, including: based on the provided face fusion model and the basic face model, a first offset of a preset sampling vertex is obtained, and input into the corresponding expression model, to obtain a second offset of the preset sampling vertex output by the expression model for the corresponding basic expression data; based on the second offset and the corresponding basic expression data, position data corresponding to the preset sampling vertex is determined, and the position data corresponds to part of the expression data of the provided face fusion model; based on the position data, the complete expression data is filled to obtain the expression data. As it has been explained in the previous article, it will not be repeated here.

[0167] In addition, the preset sampling vertices are evenly distributed on the face model (or the fused model of the face).

[0168] In addition, for the contents not described in detail in the present method 500, reference may also be made to the various steps in the above-mentioned method 200.

[0169] Based on the above similar inventive concept, Figure 6 A flow chart of a method for generating an expression provided by another exemplary embodiment of the present application is shown. The method 600 provided by the embodiment of the present application is executed by a server, and the method 600 includes the following steps:

[0170] 601: Acquire the fusion model of the face and the basic face model, and obtain the first offset of the preset sampling vertex.

[0171] 602: Input the first offset into the corresponding expression model to obtain a second offset output by the expression model for a preset sampling vertex corresponding to the basic expression data.

[0172] 603: Determine the position data corresponding to the preset sampling vertex according to the second offset and the corresponding basic expression data, so as to obtain the expression data corresponding to the fusion model of the face.

[0173] Since the specific implementation of steps 601-603 has been described in detail above, they will not be repeated here. It is only to illustrate that the method 600 can also be applied to smart terminals, such as computers.

[0174] In addition, the position data corresponds to part of the expression data, and the method 600 further includes: filling the complete expression data according to the position data to obtain the expression data.

[0175] The preset sampling vertices are evenly distributed on the face model (such as a fusion model of the face).

[0176] In addition, for the contents not described in detail in the method 600 , reference can also be made to the various steps in the above method 200 .

[0177] Based on the above similar inventive concept, Figure 7 A flowchart of a method for virtual live broadcasting provided by another exemplary embodiment of the present application is shown. The method 700 provided by the embodiment of the present application is executed by a server, and the method 700 includes the following steps:

[0178] 701: Obtain a basic face model and a fusion model of the face provided by the user, and obtain a first offset of a preset sampling vertex.

[0179] 702: Input the first offset into the corresponding expression model to obtain the second offset of the preset sampling vertex corresponding to the basic expression data output by the expression model.

[0180] 703: Determine the position data corresponding to the preset sampling vertex according to the second offset and the corresponding basic expression data, so as to obtain the expression data corresponding to the fusion model of the face.

[0181] 704: Generate a video screen with a corresponding virtual anchor according to the facial fusion model and the corresponding expression data, and broadcast the video live.

[0182] Since the specific implementation of steps 701-703 has been described in detail above, they will not be repeated here. It is only explained that method 700 can also be applied to intelligent terminals, such as computers. In addition, for step 704, taking the server as an example, after obtaining the corresponding expression data, the server can respond to the live broadcast request of the intelligent terminal, such as a computer, and generate the corresponding virtual anchor image according to the fusion model of the face, and at the same time generate the corresponding expression of the virtual anchor according to the corresponding expression data, thereby completing a series of expression actions of the virtual anchor. The server can also broadcast the video according to the preset live broadcast background image or live broadcast background video.

[0183] The above-mentioned embodiments of the present application can be applied to not only live video broadcasting, but also games, such as image settings of game players, etc. It can also be applied to other scenarios with virtual image settings.

[0184] In addition, for the contents not described in detail in the method 700, reference may also be made to the various steps in the above method 200.

[0185] Based on the above similar inventive concept, Figure 8 A flowchart of a method for virtual live broadcasting provided by another exemplary embodiment of the present application is shown. The method 800 provided by the embodiment of the present application is executed by a server, and the method 800 includes the following steps:

[0186] 801: Obtain fusion models of multiple groups of faces and expression data corresponding to the fusion models of multiple groups of faces.

[0187] 802: Obtain a fusion model of a target face provided by a user, and determine a correspondence between the fusion model of the target face and fusion models of multiple groups of faces.

[0188] 803: Generate target expression data corresponding to the fusion model of the target face based on the corresponding relationship and the acquired expression data;

[0189] 804: Generate a video screen with a corresponding virtual anchor according to the fusion model of the target face and the target expression data, and broadcast the video live.

[0190] Since the specific implementation of steps 801-804 has been described in detail above, they will not be repeated here. It is only to illustrate that the method 800 can also be applied to smart terminals, such as computers.

[0191] In addition, for the contents not described in detail in the method 800 , reference can also be made to the various steps in the above method 200 .

[0192] Fig. 9 A schematic diagram of the structural framework of an expression generation device provided by an exemplary embodiment of the present application. The device 900 can be applied to a server. The device 900 includes: an acquisition module 901, a determination module 902, and a generation module 903; the functions of each module are described in detail below:

[0193] The acquisition module 901 is used to acquire fusion models of multiple groups of faces and expression data corresponding to the fusion models of multiple groups of faces.

[0194] The determination module 902 is used to obtain a fusion model of a target face and determine a correspondence between the fusion model of the target face and fusion models of multiple groups of faces.

[0195] The generating module 903 is used to generate target expression data corresponding to the fusion model of the target face based on the corresponding relationship and the acquired expression data.

[0196] Specifically, the acquisition module 901 includes: an acquisition unit for acquiring a basic face model and a fusion model of multiple groups of faces obtained by deforming the basic face model; an acquisition unit for acquiring basic expression data corresponding to the basic face model and expression data obtained by the corresponding face fusion model.

[0197] Specifically, the determination module 902 is used to: based on a determination operation of the user, obtain model data corresponding to the fusion model of the target face corresponding to the determination operation.

[0198] Specifically, the determination module 902 obtains the deformation coefficient, and generates a fusion model of the target face according to the deformation coefficient and the basic face model.

[0199] Specifically, the determination module 902 is used to determine the fitting relationship between the fusion model of the target face and the fusion models of multiple groups of faces, so as to generate target expression data corresponding to the fusion model of the target face based on the fitting relationship and the acquired expression data.

[0200] Specifically, the determination module 902 is used to: when the fusion model of the target face is the same as the fusion model of the fitted face fitted by the fusion models of multiple groups of faces or is similar within a preset similarity, determine the fitting coefficient corresponding to the fusion model of each face according to the fusion model of the target face, the fusion models of multiple faces and the preset weight coefficient corresponding to the fusion model of each face.

[0201] Specifically, the generation module 903 includes: a first determination unit, which is used to determine the fitting coefficient of the expression data corresponding to the fusion model of each face according to the fitting coefficient corresponding to the fusion model of each face; a fitting unit, which is used to fit the expression data according to the fitting coefficient of the corresponding expression data to generate target expression data.

[0202] In addition, the device 900 further includes: a correction module, which is used to correct the target expression data based on the expression data, the fusion model of the corresponding face and the fusion model of the target face to obtain the corrected target data expression.

[0203] Specifically, the correction module includes: determining difference data between the fusion model of the target face and the fusion model of the corresponding face; and correcting the target expression data according to the difference data to obtain the corrected target expression data.

[0204] In addition, the determination module 902 is further used to: determine the difference between the fusion model of the target face and the fusion model of the corresponding face, and when the difference is greater than a threshold, perform a step of correcting the target expression data according to the difference data.

[0205] In addition, the acquisition module 901 is used to: acquire fusion models of multiple target faces and corresponding target expression data; the acquisition module 901 is used to: for each fusion model of the target face, acquire the first offset of the preset sampling vertex corresponding to the fusion model of the target face relative to the preset sampling vertex corresponding to the basic face model; the acquisition module 901 is used to: for each target expression data, acquire the second offset of the preset sampling vertex corresponding to the target expression data relative to the preset sampling vertex corresponding to the corresponding basic expression data; the device 900 also includes: a training module, which is used to train the preset initial expression model with the corresponding first offset and the corresponding second offset as training data for different basic expression data, and obtain expression models for different expressions, so as to generate expression data based on the expression model.

[0206] In addition, the device 900 further includes: an obtaining module, which is used to obtain expression data corresponding to the fusion model of the target face according to the fusion model of the target face and the corresponding expression model, as target expression data.

[0207] Specifically, the module includes: an output unit, which is used to obtain the offset of the preset sampling vertex according to the fusion model of the target face and the basic face model provided by the user, and input it into the corresponding expression model to obtain the offset of the preset sampling vertex for the corresponding basic expression data output by the expression model; a second determination unit, which is used to determine the position data corresponding to the preset sampling vertex according to the obtained offset and the corresponding basic expression data, and the position data corresponds to part of the expression data of the fusion model of the target face; a filling unit, which is used to fill the complete expression data according to the position data to obtain the expression data.

[0208] The preset sampling vertices are evenly distributed on the face model (or the fusion model of the face).

[0209] The preset sampling vertices include all vertices on the face model (or the fusion model of the face).

[0210] In addition, the generating module 903 generates corresponding expressions according to the target expression data for display.

[0211] Fig.10 A schematic diagram of the structural framework of a device for generating an expression model provided by another exemplary embodiment of the present application is shown. The device 1000 can be applied to a server. The device 1000 includes: an acquisition module 1001 and a training module 1002. The functions of each module are described in detail below:

[0212] The acquisition module 1001 is used to acquire fusion models of multiple target faces and corresponding target expression data.

[0213] The acquisition module 1001 is used to acquire, for the fusion model of the target face, a first offset of a preset sampling vertex corresponding to the fusion model of the target face relative to a preset sampling vertex corresponding to the basic face model.

[0214] The acquisition module 1001 is used to acquire, for target expression data, a second offset of a preset sampling vertex corresponding to the target expression data relative to a preset sampling vertex corresponding to basic expression data, wherein the basic expression data corresponds to a basic face model.

[0215] The training module 1002 is used to train the preset initial expression model with the corresponding first offset and the corresponding second offset as training data for different basic expression data, and obtain expression models for different expressions, so as to generate expression data based on the expression model.

[0216] In addition, the device 1000 further includes: a generation module, which is used to randomly generate a corresponding random face fusion model based on the basic face model as the target face fusion model.

[0217] In addition, the device 1000 also includes: an obtaining module, which is used to obtain a facial fusion model provided by a user, and obtain expression data corresponding to the provided facial fusion model based on the provided facial fusion model and the corresponding expression model.

[0218] Specifically, the obtaining module includes: an obtaining unit, which is used to obtain a first offset of a preset sampling vertex according to a provided facial fusion model and a basic face model, and input it into a corresponding expression model to obtain a second offset of the preset sampling vertex output by the expression model for the corresponding basic expression data; a determining unit, which is used to determine position data corresponding to the preset sampling vertex according to the second offset and the corresponding basic expression data, and the position data corresponds to partial expression data of the provided facial fusion model; and a filling unit, which is used to fill the complete expression data according to the position data to obtain the expression data.

[0219] The preset sampling vertices are evenly distributed on the face model (or the fusion model of the face).

[0220] It should be noted that for some contents not mentioned in the device 1000 , reference may be made to the contents of the above-mentioned device 900 .

[0221] Fig.11 A schematic diagram of the structural framework of an expression generation device provided by another exemplary embodiment of the present application is shown. The device 1100 can be applied to a server. The device 1100 includes: an acquisition module 1101, an acquisition module 1102 and a determination module 1103. The functions of each module are described in detail below:

[0222] The acquisition module 1101 is used to acquire the fusion model of the face and the basic face model, and obtain the first offset of the preset sampling vertex.

[0223] The obtaining module 1102 is used to input the first offset into the corresponding expression model to obtain the second offset of the preset sampling vertex corresponding to the basic expression data output by the expression model.

[0224] The determination module 1103 is used to determine the position data corresponding to the preset sampling vertex according to the second offset and the corresponding basic expression data, so as to obtain the expression data corresponding to the fusion model of the face.

[0225] In addition, the location data corresponds to part of the expression data, and the module 1102 is further used to fill the complete expression data according to the location data to obtain the expression data.

[0226] The preset sampling vertices are evenly distributed on the face model (or the fusion model of the face).

[0227] It should be noted that for some contents not mentioned in the device 1100 , reference may be made to the contents of the above-mentioned device 900 .

[0228] The above describes Fig. 9 The internal functions and structure of the device 900 shown, in one possible design, Fig. 9 The structure of the device 900 shown can be implemented as a computing device, such as a cloud server. Fig.12 As shown, the device 1200 may include: a memory 1201, a processor 1202;

[0229] The memory 1201 is used to store computer programs.

[0230] Processor 1202 is used to execute a computer program to: obtain fusion models of multiple groups of faces and expression data corresponding to the fusion models of the multiple groups of faces; obtain the fusion model of a target face, and determine the correspondence between the fusion model of the target face and the fusion models of the multiple groups of faces; and generate target expression data corresponding to the fusion model of the target face based on the correspondence and the obtained expression data.

[0231] Specifically, the processor 1202 is specifically used to: obtain a basic face model and a fusion model of multiple groups of faces obtained by deforming the basic face model; obtain basic expression data corresponding to the basic face model and expression data obtained by the corresponding face fusion model.

[0232] Specifically, the processor 1202 is specifically configured to: based on a determination operation of the user, obtain model data corresponding to the fusion model of the target face corresponding to the determination operation.

[0233] Specifically, the processor 1202 is specifically configured to: obtain a deformation coefficient, and generate a fusion model of the target face according to the deformation coefficient and the basic face model.

[0234] Specifically, the processor 1202 is specifically used to: determine the fitting relationship between the fusion model of the target face and the fusion models of multiple groups of faces, so as to generate target expression data corresponding to the fusion model of the target face based on the fitting relationship and the acquired expression data.

[0235] Specifically, the processor 1202 is specifically used to: when the fusion model of the target face is the same as the fusion model of the fitted face fitted by the fusion models of multiple groups of faces or is similar within a preset similarity, determine the fitting coefficient corresponding to the fusion model of each face according to the fusion model of the target face, the fusion models of multiple faces and the preset weight coefficient corresponding to the fusion model of each face.

[0236] Specifically, the processor 1202 is specifically used to: determine the fitting coefficient of the expression data corresponding to the fusion model of each face according to the fitting coefficient corresponding to the fusion model of each face; fit the expression data according to the fitting coefficient of the corresponding expression data to generate target expression data.

[0237] In addition, the processor 1202 is further configured to: modify the target expression data based on the expression data, the fusion model of the corresponding face, and the fusion model of the target face to obtain a modified target data expression.

[0238] Specifically, the processor 1202 is specifically configured to: determine difference data between the fusion model of the target face and the fusion model of the corresponding face; and correct the target expression data according to the difference data to obtain corrected target expression data.

[0239] In addition, the processor 1202 is further configured to: determine the difference between the fusion model of the target face and the fusion model of the corresponding face, and when the difference is greater than a threshold, execute a step of correcting the target expression data according to the difference data.

[0240] In addition, the processor 1202 is also used to: obtain fusion models of multiple target faces and corresponding target expression data; for each fusion model of the target face, obtain a first offset of a preset sampling vertex corresponding to the fusion model of the target face relative to a preset sampling vertex corresponding to the basic face model; for each target expression data, obtain a second offset of a preset sampling vertex corresponding to the target expression data relative to a preset sampling vertex corresponding to the corresponding basic expression data; for different basic expression data, use the corresponding first offset and the corresponding second offset as training data to train the preset initial expression model to obtain expression models for different expressions, so as to generate expression data based on the expression model.

[0241] In addition, the processor 1202 is further configured to obtain, according to the fusion model of the target face and the corresponding expression model, expression data corresponding to the fusion model of the target face as target expression data.

[0242] Specifically, processor 1202 is specifically used to: obtain the offset of the preset sampling vertex according to the fusion model of the target face and the basic face model provided by the user, and input it into the corresponding expression model to obtain the offset of the preset sampling vertex for the corresponding basic expression data output by the expression model; determine the position data corresponding to the preset sampling vertex according to the obtained offset and the corresponding basic expression data, the position data corresponding to part of the expression data of the fusion model of the target face; fill the complete expression data according to the position data to obtain the expression data.

[0243] The preset sampling vertices are evenly distributed on the face model (or the fusion model of the face).

[0244] The preset sampling vertices include all vertices on the face model (or the fusion model of the face).

[0245] In addition, the processor 1202 is further configured to generate corresponding expressions according to the target expression data for display.

[0246] In addition, an embodiment of the present invention provides a computer storage medium, when the computer program is executed by one or more processors, causing the one or more processors to implement Figure 1-Figure 4 The steps of a method for generating an expression in a method embodiment.

[0247] The above describes Fig.10 The internal functions and structure of the device 1000 shown, in one possible design, Fig.10 The structure of the device 1000 shown can be implemented as a computing device, such as a cloud server. Fig.13 As shown, the device 1300 may include: a memory 1301, a processor 1302;

[0248] The memory 1301 is used to store computer programs.

[0249] Processor 1302 is used to execute the computer program to: obtain fusion models of multiple target faces and corresponding target expression data; for the fusion model of the target face, obtain a first offset of a preset sampling vertex corresponding to the fusion model of the target face relative to a preset sampling vertex corresponding to the basic face model; for the target expression data, obtain a second offset of a preset sampling vertex corresponding to the target expression data relative to a preset sampling vertex corresponding to the basic expression data, the basic expression data corresponding to the basic face model; for different basic expression data, use the corresponding first offset and the corresponding second offset as training data to train the preset initial expression model to obtain expression models for different expressions, so as to generate expression data based on the expression model.

[0250] In addition, the processor 1302 is further configured to: based on the basic face model, randomly generate a corresponding random face fusion model as a target face fusion model.

[0251] In addition, the processor 1302 is further configured to: obtain a facial fusion model provided by a user, and obtain expression data corresponding to the provided facial fusion model based on the provided facial fusion model and the corresponding expression model.

[0252] Specifically, processor 1302 is specifically used to: obtain a first offset of a preset sampling vertex based on a provided facial fusion model and a basic facial model, and input the first offset into a corresponding expression model to obtain a second offset of the preset sampling vertex output by the expression model for corresponding basic expression data; determine position data corresponding to the preset sampling vertex based on the second offset and the corresponding basic expression data, the position data corresponding to partial expression data of the provided facial fusion model; fill in the complete expression data based on the position data to obtain the expression data.

[0253] The preset sampling vertices are evenly distributed on the face model (or the fusion model of the face).

[0254] It should be noted that for some contents not mentioned in the device 1300 , reference may be made to the contents of the above-mentioned device 1200 .

[0255] In addition, an embodiment of the present invention provides a computer storage medium, when the computer program is executed by one or more processors, causing the one or more processors to implement Figure 5 The method embodiment provides steps of a method for generating an expression model.

[0256] The above describes Fig.11 The internal functions and structure of the device 1100 shown, in one possible design, Fig.11The structure of the device 1100 shown can be implemented as a computing device, such as a cloud server. Fig.14 As shown, the device 1400 may include: a memory 1401, a processor 1402;

[0257] The memory 1401 is used to store computer programs.

[0258] Processor 1402 is used to execute the computer program to: obtain a facial fusion model and a basic facial model to obtain a first offset of a preset sampling vertex; input the first offset into a corresponding expression model to obtain a second offset of a preset sampling vertex output by the expression model for corresponding basic expression data; determine position data corresponding to the preset sampling vertex according to the second offset and the corresponding basic expression data to obtain expression data corresponding to the facial fusion model.

[0259] In addition, the position data corresponds to part of the expression data, and the processor 1402 is further used to fill the complete expression data according to the position data to obtain the expression data.

[0260] The preset sampling vertices are evenly distributed on the face model (or the fusion model of the face).

[0261] It should be noted that for some contents not mentioned in the device 1400 , reference may be made to the contents of the above-mentioned device 1200 .

[0262] In addition, an embodiment of the present invention provides a computer storage medium, when the computer program is executed by one or more processors, causing the one or more processors to implement Figure 6 The steps of a method for generating an expression in a method embodiment.

[0263] Another exemplary embodiment of the present application provides a virtual live broadcast device. The device 1100 can be applied to a server. The device 1500 includes: an acquisition module 1501, a obtaining module 1502, a determination module 1503 and a generation module 1504. The functions of each module are described in detail below:

[0264] The acquisition module 1501 is used to acquire the basic face model and the fusion model of the face provided by the user, and obtain the first offset of the preset sampling vertex.

[0265] The obtaining module 1502 is used to input the first offset into the corresponding expression model to obtain the second offset of the preset sampling vertex corresponding to the basic expression data output by the expression model.

[0266] The determination module 1503 is used to determine the position data corresponding to the preset sampling vertex according to the second offset and the corresponding basic expression data, so as to obtain the expression data corresponding to the fusion model of the face.

[0267] The generation module 1504 is used to generate a video screen with a corresponding virtual anchor according to the facial fusion model and the corresponding expression data, and to broadcast the video live.

[0268] It should be noted that for some contents not mentioned in the device 1500 , reference may be made to the contents of the above-mentioned device 900 .

[0269] The internal functions and structure of the device 1500 are described above. In a possible design, the structure of the device 1500 can be implemented as a computing device, such as a cloud server. The device 1600 may include: a memory 1601, a processor 1602;

[0270] The memory 1601 is used to store computer programs.

[0271] Processor 1602 is used to execute the computer program to: obtain a basic face model and a fusion model of the face provided by a user to obtain a first offset of a preset sampling vertex; input the first offset into a corresponding expression model to obtain a second offset of a preset sampling vertex output by the expression model for corresponding basic expression data; determine position data corresponding to the preset sampling vertex according to the second offset and the corresponding basic expression data to obtain expression data corresponding to the fusion model of the face; generate a video screen with a corresponding virtual anchor according to the fusion model of the face and the corresponding expression data, and perform live video broadcast.

[0272] It should be noted that for some contents not mentioned in the device 1600 , reference may be made to the contents of the above-mentioned device 1200 .

[0273] In addition, an embodiment of the present invention provides a computer storage medium, when the computer program is executed by one or more processors, causing the one or more processors to implement Figure 7 The steps of a method for virtual live broadcast in a method embodiment.

[0274] Another exemplary embodiment of the present application provides a virtual live broadcast device. The device 1700 can be applied to a server. The device 1700 includes: an acquisition module 1701, a determination module 1702, and a generation module 1703. The functions of each module are described in detail below:

[0275] The acquisition module 1701 is used to acquire fusion models of multiple groups of faces and expression data corresponding to the fusion models of multiple groups of faces.

[0276] The determination module 1702 is used to obtain the fusion model of the target face provided by the user, and determine the correspondence between the fusion model of the target face and the fusion models of multiple groups of faces.

[0277] The generating module 1703 is used to generate target expression data corresponding to the fusion model of the target face based on the corresponding relationship and the acquired expression data.

[0278] The generation module 1703 is used to generate a video screen with a corresponding virtual anchor according to the fusion model of the target face and the target expression data, and to broadcast the video live.

[0279] It should be noted that for some contents not mentioned in the device 1100 , reference may be made to the contents of the above-mentioned device 900 .

[0280] The internal functions and structure of the device 1700 are described above. In a possible design, the structure of the device 1700 can be implemented as a computing device, such as a cloud server. The device 1800 may include: a memory 1801, a processor 1802;

[0281] The memory 1801 is used to store computer programs.

[0282] Processor 1802 is used to execute the computer program to: obtain fusion models of multiple groups of faces and expression data corresponding to the fusion models of the multiple groups of faces; obtain a fusion model of a target face provided by a user, and determine the correspondence between the fusion model of the target face and the fusion models of the multiple groups of faces; based on the correspondence and the acquired expression data, generate target expression data corresponding to the fusion model of the target face; based on the fusion model of the target face and the target expression data, generate a video screen with a corresponding virtual anchor, and broadcast the video live.

[0283] It should be noted that for some contents not mentioned in the device 1800 , reference may be made to the contents of the above-mentioned device 1200 .

[0284] In addition, an embodiment of the present invention provides a computer storage medium, when the computer program is executed by one or more processors, causing the one or more processors to implement Figure 8 The steps of a method for virtual live broadcast in a method embodiment.

[0285] Based on the above similar inventive concept, another exemplary embodiment of the present application provides a method for generating a virtual customer service. The method 1900 provided in the embodiment of the present application is executed by a server, and the method 1900 includes the following steps:

[0286] 1901: Obtain fusion models of multiple groups of faces and expression data corresponding to the fusion models of multiple groups of faces.

[0287] 1902: Obtain a fusion model of a target face provided by a user, and determine a correspondence between the fusion model of the target face and fusion models of multiple groups of faces.

[0288] 1903: Generate target expression data corresponding to the fusion model of the target face based on the corresponding relationship and the acquired expression data;

[0289] 1904: Generate a corresponding virtual customer service image based on the fusion model of the target face and the target expression data for display.

[0290] Since the specific implementation of steps 1901-1903 has been described in detail above, it will not be repeated here. It is only explained that for step 1904, after determining the target expression data, the corresponding virtual customer service image, or virtual customer service image, can be generated according to the fusion model of the corresponding target face. By transforming, or selecting the expression corresponding to the target expression data, the virtual customer service changes the expression and generates different images, or a virtual customer service animation or a small video of the virtual customer service. The final virtual customer service can be generated by an intelligent terminal, such as a server sending the corresponding target expression data and the fusion model of the target face to an intelligent terminal, such as a mobile phone, and the mobile phone generates a virtual customer service image, or a virtual customer service animation or a small video of the virtual customer service based on these data. It can also be generated by the server, and the generated virtual customer service image, or a virtual customer service animation or a small video of the virtual customer service is sent to the intelligent terminal for display. And the method 1900 can also be applied to intelligent terminals, such as computers.

[0291] Specifically, determining the correspondence between the fusion model of the target face and the fusion models of the multiple groups of faces includes: determining the fitting relationship between the fusion model of the target face and the fusion models of the multiple groups of faces, so as to generate target expression data corresponding to the fusion model of the target face based on the fitting relationship and the acquired expression data.

[0292] Specifically, based on the corresponding relationship and the acquired expression data, target expression data corresponding to the fusion model of the target face is generated, including: determining the fitting coefficient of the expression data corresponding to the fusion model of each face according to the fitting coefficient corresponding to the fusion model of each face; fitting the expression data according to the fitting coefficient of the corresponding expression data to generate the target expression data.

[0293] In addition, the method 1900 further includes: based on the expression data, the fusion model of the corresponding face and the fusion model of the target face, correcting the target expression data to obtain a corrected target data expression.

[0294] In addition, the method 1900 further includes: obtaining expression data corresponding to the fusion model of the target face according to the fusion model of the target face and the corresponding expression model, as target expression data.

[0295] Specifically, expression data corresponding to the fusion model of the target face is obtained according to the fusion model of the target face and the corresponding expression model, including: obtaining the offset of the preset sampling vertex according to the fusion model of the target face and the basic face model provided by the user, and inputting it into the corresponding expression model to obtain the offset of the preset sampling vertex output by the expression model for the corresponding basic expression data; determining the position data corresponding to the preset sampling vertex according to the obtained offset and the corresponding basic expression data, the position data corresponding to the partial expression data of the fusion model of the target face; filling the complete expression data according to the position data to obtain the expression data.

[0296] Since the above content has been explained in the previous article, it will not be repeated here.

[0297] In addition, the contents not described in detail in the present method 1900 may also refer to the various steps in the above-mentioned method 200.

[0298] An exemplary embodiment of the present application provides an expression generation device. The device 2000 can be applied to a server. The device 2000 includes: an acquisition module 2001, a determination module 2002, and a generation module 2003; the functions of each module are described in detail below:

[0299] The acquisition module 2001 is used to acquire fusion models of multiple groups of faces and expression data corresponding to the fusion models of multiple groups of faces.

[0300] The determination module 2002 is used to obtain the fusion model of the target face provided by the user, and determine the correspondence between the fusion model of the target face and the fusion models of multiple groups of faces.

[0301] The generating module 2003 is used to generate target expression data corresponding to the fusion model of the target face based on the corresponding relationship and the acquired expression data.

[0302] The generation module 2003 is used to generate a corresponding virtual customer service image for display according to the fusion model of the target face and the target expression data.

[0303] Specifically, the determination module 2002 is used to determine the fitting relationship between the fusion model of the target face and the fusion models of multiple groups of faces, so as to generate target expression data corresponding to the fusion model of the target face based on the fitting relationship and the acquired expression data.

[0304] Specifically, the generation module 2003 includes: a first determination unit, which is used to determine the fitting coefficient of the expression data corresponding to the fusion model of each face according to the fitting coefficient corresponding to the fusion model of each face; a fitting unit, which is used to fit the expression data according to the fitting coefficient of the corresponding expression data to generate target expression data.

[0305] In addition, the device 2000 further includes: a correction module, which is used to correct the target expression data based on the expression data, the fusion model of the corresponding face and the fusion model of the target face to obtain the corrected target data expression.

[0306] In addition, the determination module 2002 is further used to obtain expression data corresponding to the fusion model of the target face according to the fusion model of the target face and the corresponding expression model, as target expression data.

[0307] Specifically, the determination module 2002 includes: a second determination unit, which is used to obtain the offset of the preset sampling vertex according to the fusion model of the target face and the basic face model provided by the user, and input it into the corresponding expression model to obtain the offset of the preset sampling vertex for the corresponding basic expression data output by the expression model; a third determination unit, which is used to determine the position data corresponding to the preset sampling vertex according to the obtained offset and the corresponding basic expression data, and the position data corresponds to part of the expression data of the fusion model of the target face; a filling unit, which is used to fill the complete expression data according to the position data to obtain the expression data.

[0308] It should be noted that for some contents not mentioned in the device 2000 , reference may be made to the contents of the above-mentioned device 900 .

[0309] The internal functions and structure of the device 2000 are described above. In a possible design, the structure of the device 2000 can be implemented as a computing device, such as a cloud server. The device 2100 may include: a memory 2101 and a processor 2102 .

[0310] The memory 2101 is used to store computer programs.

[0311] Processor 2102 is used to execute the computer program to: obtain fusion models of multiple groups of faces and expression data corresponding to the fusion models of the multiple groups of faces; obtain a fusion model of a target face provided by a user, and determine the correspondence between the fusion model of the target face and the fusion models of the multiple groups of faces; based on the correspondence and the acquired expression data, generate target expression data corresponding to the fusion model of the target face; and generate a corresponding image of a virtual customer service for display according to the fusion model of the target face and the target expression data.

[0312] Specifically, the processor 2102 is specifically used to: determine the fitting relationship between the fusion model of the target face and the fusion models of multiple groups of faces, so as to generate target expression data corresponding to the fusion model of the target face based on the fitting relationship and the acquired expression data.

[0313] Specifically, processor 2102 is specifically used to: determine the fitting coefficient of the expression data corresponding to the fusion model of each face according to the fitting coefficient corresponding to the fusion model of each face; fit the expression data according to the fitting coefficient of the corresponding expression data to generate target expression data.

[0314] In addition, the processor 2102 is further configured to: modify the target expression data based on the expression data, the fusion model of the corresponding face, and the fusion model of the target face to obtain a modified target data expression.

[0315] In addition, the processor 2102 is further configured to obtain, according to the fusion model of the target face and the corresponding expression model, expression data corresponding to the fusion model of the target face as target expression data.

[0316] Specifically, processor 2102 is specifically used to: obtain the offset of the preset sampling vertex according to the fusion model of the target face and the basic face model provided by the user, and input it into the corresponding expression model to obtain the offset of the preset sampling vertex for the corresponding basic expression data output by the expression model; determine the position data corresponding to the preset sampling vertex according to the obtained offset and the corresponding basic expression data, the position data corresponding to part of the expression data of the fusion model of the target face; fill the complete expression data according to the position data to obtain the expression data.

[0317] It should be noted that for some contents not mentioned in the device 2100, reference may be made to the contents of the above-mentioned device 1200.

[0318] In addition, an embodiment of the present invention provides a computer storage medium. When a computer program is executed by one or more processors, the one or more processors implement the steps of a method for generating a virtual customer service in the embodiment of method 1900.

[0319] In addition, in some of the processes described in the above embodiments and the accompanying drawings, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel, and the sequence numbers of the operations, such as 201, 202, 203, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0320] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0321] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by adding a necessary general hardware platform, and of course can also be implemented by combining hardware and software. Based on such an understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a computer product, and the present invention can be in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0322] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable multimedia data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable multimedia data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0323] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable multimedia data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0324] These computer program instructions may also be loaded onto a computer or other programmable multimedia data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0325] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0326] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0327] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0328] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating an expression, characterized in that: include: Acquire a basic face model, a fusion model of multiple groups of faces obtained by deforming the basic face model, expression data of multiple expression types corresponding to the basic face model, and expression data of the multiple expression types corresponding to the fusion model of each group of faces in the fusion model of the multiple groups of faces; Acquire a fusion model of a target face, determine a correspondence between an offset of a preset sampling vertex on the fusion model of the target face relative to the preset sampling vertex on the base face model, and an offset of the preset sampling vertex on the fusion models of multiple groups of faces relative to the preset sampling vertex on the base face model; Based on the corresponding relationship and the acquired expression data, target expression data of the multiple expression types corresponding to the fusion model of the target face are generated.

2. The method according to claim 1, characterized in that The acquiring of the fusion models of the multiple groups of faces and the expression data corresponding to the fusion models of the multiple groups of faces includes: Acquire a basic face model and a fusion model of multiple groups of faces obtained by deforming the basic face model; The expression data is obtained based on the basic expression data corresponding to the basic face model and the corresponding fusion model of the face.

3. The method according to claim 1, characterized in that The step of obtaining the fusion model of the target face includes: Based on a determination operation of the user, model data corresponding to a fusion model of a target face corresponding to the determination operation is acquired.

4. The method according to claim 1, characterized in that: The step of obtaining the fusion model of the target face includes: A deformation coefficient is obtained, and a fusion model of the target face is generated according to the deformation coefficient and a basic face model.

5. The method according to claim 1, characterized in that The determining of the correspondence between the offset of the preset sampling vertices on the fusion model of the target face relative to the preset sampling vertices on the basic face model and the offset of the preset sampling vertices on the fusion model of multiple groups of faces relative to the preset sampling vertices on the basic face model includes: Determine a fitting relationship between an offset of a preset sampling vertex on the fusion model of the target face relative to the preset sampling vertex on the base face model and an offset of the preset sampling vertex on the fusion model of multiple groups of faces relative to the preset sampling vertex on the base face model, so as to generate target expression data of the multiple expression types corresponding to the fusion model of the target face based on the fitting relationship and the acquired expression data.

6. The method according to claim 1, characterized in that The step of generating target expression data of the plurality of expression types corresponding to the fusion model of the target face based on the corresponding relationship and the acquired expression data includes: Determine the fitting coefficient of the expression data of each expression type corresponding to the fusion model of each face according to the fitting coefficient corresponding to the determined fusion model of each face; According to the fitting coefficient of the corresponding expression data of each expression type, the expression data of each expression type is fitted to generate the target expression data of each expression type.

7. The method according to claim 1, characterized in that The method further comprises: Based on the expression data, the fusion model of the corresponding face and the fusion model of the target face, the target expression data is corrected to obtain a corrected target data expression.

8. The method according to claim 1, characterized in that The method further comprises: According to the fusion model of the target face and the expression model corresponding to each expression type, expression data of each expression type corresponding to the fusion model of the target face is obtained as target expression data of each expression type.

9. The method according to claim 8, characterized in that The step of obtaining expression data of each expression type corresponding to the fusion model of the target face according to the fusion model of the target face and the expression model corresponding to each expression type comprises: According to the fusion model of the target face and the basic face model provided by the user, the offset of the preset sampling vertex is obtained, and the offset is input into the expression model corresponding to each expression type, so as to obtain the offset of the preset sampling vertex output by the expression model for the corresponding basic expression data; Determine, according to the obtained offset and the corresponding basic expression data, position data corresponding to the preset sampling vertex, wherein the position data corresponds to partial expression data of the fusion model of the target face; The complete expression data is filled according to the position data to obtain the expression data of each expression type.

10. The method according to claim 1, characterized in that The method further comprises: Generate corresponding multiple expressions according to the target expression data of the multiple expression types for display.

11. A method for generating an expression model, characterized in that: include: Acquire a plurality of fusion models of target faces and corresponding target expression data, wherein the fusion models of the plurality of target faces are obtained by deforming a basic face model, the basic face model has corresponding basic expression data of a plurality of expression types, and each fusion model of the target face has corresponding target expression data of the plurality of expression types; For the fusion model of the target face, obtaining a first offset of a preset sampling vertex on the fusion model of the target face relative to the preset sampling vertex on the base face model; For the target expression data of each expression type, obtaining a second offset of the preset sampling vertex on the target expression data of each expression type relative to the preset sampling vertex on the basic expression data of each expression type; For the basic expression data of each expression type, the first offset corresponding to each expression type is used as input training data, and the second offset corresponding to each expression type is used as output training data, and a preset initial expression model is trained to obtain an expression model for each expression type, so that the expression data of each expression type is generated based on the expression model.

12. The method according to claim 11, characterized in that The method further comprises: A facial fusion model provided by a user is obtained, and expression data of multiple expression types corresponding to the provided facial fusion model are obtained based on the provided facial fusion model and expression models of multiple corresponding expression types.

13. A method for generating an expression, characterized in that: include: Acquire a facial fusion model and a basic facial model, and obtain a first offset of a preset sampling vertex on the facial fusion model relative to the preset sampling vertex on the basic facial model, wherein the facial fusion model is obtained by deforming the basic facial model, and the basic facial model has basic expression data corresponding to multiple expression types; Inputting the first offset into an expression model corresponding to the expression type, and obtaining a second offset output by the expression model for the preset sampling vertex on the corresponding basic expression data; According to the second offset and the corresponding basic expression data, the position data corresponding to the preset sampling vertex is determined to obtain the expression data of the expression type corresponding to the facial fusion model.

14. The method according to claim 13, characterized in that The position data corresponds to part of the expression data, and the method further includes: The complete expression data is filled according to the position data to obtain the expression data.

15. A virtual live broadcast method, characterized in that: include: Acquire a basic face model and a fusion model of a face provided by a user, and obtain a first offset of a preset sampling vertex on the fusion model of the face relative to the preset sampling vertex on the basic face model, wherein the fusion model of the face is obtained by deforming the basic face model, and the basic face model has basic expression data corresponding to multiple expression types; Inputting the first offset into an expression model corresponding to the expression type, and obtaining a second offset output by the expression model for the preset sampling vertex on the corresponding basic expression data; Determine the position data corresponding to the preset sampling vertex according to the second offset and the corresponding basic expression data, so as to obtain expression data of the expression type corresponding to the facial fusion model; Based on the facial fusion model and the corresponding expression data, a video screen with a corresponding virtual anchor is generated, and the video is broadcast live.

16. A virtual live broadcast method, characterized in that: include: Acquire a basic face model, a fusion model of multiple groups of faces obtained by deforming the basic face model, expression data of multiple expression types corresponding to the basic face model, and expression data of the multiple expression types corresponding to the fusion model of each group of faces in the fusion model of the multiple groups of faces; Acquire a fusion model of a target face provided by a user, and determine a correspondence between an offset of a preset sampling vertex on the fusion model of the target face relative to the preset sampling vertex on the base face model, and an offset of the preset sampling vertex on the fusion models of multiple groups of faces relative to the preset sampling vertex on the base face model; Based on the corresponding relationship and the acquired expression data, generating target expression data of the multiple expression types corresponding to the fusion model of the target face; According to the fusion model of the target face and the target expression data, a video screen with a corresponding virtual anchor is generated, and the video is broadcast live.

17. A method for generating a virtual customer service, characterized in that: include: Acquire a basic face model, a fusion model of multiple groups of faces obtained by deforming the basic face model, expression data of multiple expression types corresponding to the basic face model, and expression data of the multiple expression types corresponding to the fusion model of each group of faces in the fusion model of the multiple groups of faces; Acquire a fusion model of a target face provided by a user, and determine a correspondence between an offset of a preset sampling vertex on the fusion model of the target face relative to the preset sampling vertex on the base face model, and an offset of the preset sampling vertex on the fusion models of multiple groups of faces relative to the preset sampling vertex on the base face model; Based on the corresponding relationship and the acquired expression data, generating target expression data of the multiple expression types corresponding to the fusion model of the target face; According to the fusion model of the target face and the target expression data, a corresponding image of the virtual customer service is generated for display.

18. The method according to claim 17, characterized in that The determining of the correspondence between the offset of the preset sampling vertices on the fusion model of the target face relative to the preset sampling vertices on the basic face model and the offset of the preset sampling vertices on the fusion model of multiple groups of faces relative to the preset sampling vertices on the basic face model includes: Determine a fitting relationship between an offset of a preset sampling vertex on the fusion model of the target face relative to the preset sampling vertex on the base face model and an offset of the preset sampling vertex on the fusion model of multiple groups of faces relative to the preset sampling vertex on the base face model, so as to generate target expression data of the multiple expression types corresponding to the fusion model of the target face based on the fitting relationship and the acquired expression data.

19. The method according to claim 17, characterized in that The generating target expression data corresponding to the fusion model of the target face based on the corresponding relationship and the acquired expression data includes: Determine the fitting coefficient of the expression data corresponding to the fusion model of each face according to the fitting coefficient corresponding to the fusion model of each face; The expression data are fitted according to the fitting coefficient of the corresponding expression data to generate the target expression data.

20. The method according to claim 17, characterized in that The method further comprises: Based on the expression data, the fusion model of the corresponding face and the fusion model of the target face, the target expression data is corrected to obtain a corrected target data expression.

21. The method according to claim 17, characterized in that The method further comprises: According to the fusion model of the target face and the expression model corresponding to each expression type, expression data of each expression type corresponding to the fusion model of the target face is obtained as the target expression data of each expression type.

22. The method according to claim 21, characterized in that The step of obtaining expression data of each expression type corresponding to the fusion model of the target face according to the fusion model of the target face and the expression model corresponding to each expression type comprises: According to the fusion model of the target face and the basic face model provided by the user, the offset of the preset sampling vertex is obtained, and the offset is input into the expression model corresponding to each expression type, so as to obtain the offset of the preset sampling vertex output by the expression model for the corresponding basic expression data; Determine, according to the obtained offset and the corresponding basic expression data, position data corresponding to the preset sampling vertex, wherein the position data corresponds to partial expression data of the fusion model of the target face; The complete expression data is filled according to the position data to obtain the expression data of each expression type.

23. A computing device comprising: Memory, processor; The memory is used to store computer programs; The processor is used to execute the computer program to obtain a basic face model, a fusion model of multiple groups of faces obtained by deforming the basic face model, expression data of multiple expression types corresponding to the basic face model, and expression data of the multiple expression types corresponding to the fusion model of the fusion model of each group of faces in the multiple groups of faces; Acquire a fusion model of a target face, determine a correspondence between an offset of a preset sampling vertex on the fusion model of the target face relative to the preset sampling vertex on the base face model, and an offset of the preset sampling vertex on the fusion models of multiple groups of faces relative to the preset sampling vertex on the base face model; Based on the corresponding relationship and the acquired expression data, target expression data of the multiple expression types corresponding to the fusion model of the target face are generated.

24. A computing device comprising: Memory, processor; The memory is used to store computer programs; The processor is configured to execute the computer program to: Acquire a plurality of fusion models of target faces and corresponding target expression data, wherein the fusion models of the plurality of target faces are obtained by deforming a basic face model, the basic face model has corresponding basic expression data of a plurality of expression types, and each fusion model of the target face has corresponding target expression data of the plurality of expression types; For the fusion model of the target face, obtaining a first offset of a preset sampling vertex on the fusion model of the target face relative to the preset sampling vertex on the base face model; For each target expression data of each expression type, obtaining a second offset of the preset sampling vertex on the target expression data of each expression type relative to the preset sampling vertex on the corresponding basic expression data, wherein the basic expression data corresponds to the basic face model; For the basic expression data of each expression type, the first offset corresponding to each expression type is used as input training data, and the second offset corresponding to each expression type is used as output training data, and a preset initial expression model is trained to obtain an expression model for each expression type, so that the expression data of each expression type is generated based on the expression model.

25. A computing device comprising: Memory, processor; The memory is used to store computer programs; The processor is configured to execute the computer program to: Acquire a facial fusion model and a basic facial model, and obtain a first offset of a preset sampling vertex on the facial fusion model relative to the preset sampling vertex on the basic facial model, wherein the facial fusion model is obtained by deforming the basic facial model, and the basic facial model has basic expression data corresponding to multiple expression types; Inputting the first offset into an expression model corresponding to the expression type, and obtaining a second offset output by the expression model for the preset sampling vertex on the corresponding basic expression data; According to the second offset and the corresponding basic expression data, the position data corresponding to the preset sampling vertex is determined to obtain the expression data of the expression type corresponding to the facial fusion model.

26. A computing device comprising: Memory, processor; The memory is used to store computer programs; The processor is configured to execute the computer program to: Acquire a basic face model, a fusion model of multiple groups of faces obtained by deforming the basic face model, expression data of multiple expression types corresponding to the basic face model, and expression data of the multiple expression types corresponding to the fusion model of each group of faces in the fusion model of the multiple groups of faces; Acquire a fusion model of a target face provided by a user, and determine a correspondence between an offset of a preset sampling vertex on the fusion model of the target face relative to the preset sampling vertex on the base face model, and an offset of the preset sampling vertex on the fusion models of multiple groups of faces relative to the preset sampling vertex on the base face model; Based on the corresponding relationship and the acquired expression data, generating target expression data of the multiple expression types corresponding to the fusion model of the target face; According to the fusion model of the target face and the target expression data, a corresponding image of the virtual customer service is generated for display.

27. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by one or more processors, the one or more processors are caused to implement the steps of the method according to any one of claims 1 to 22.

Images