Image processing system and method and image stylization platform

Abstract

The invention provides an image processing system and method and an image stylization platform, the image processing system comprises a client and an image stylization platform, the client sends an image processing request to the image stylization platform, and the image processing request comprises stylization attribute information and stylization area information; and the image stylization platform obtains a to-be-processed image based on the image processing request, obtains a region feature value of the to-be-processed image according to the stylization region information, and carries out stylization processing on the to-be-processed image according to the region feature value and the stylization attribute information to obtain a target stylization image. The image is stylized by combining the regional features of the image, so that the accuracy of the stylized processing of the image is improved.

Description

[0001] This disclosure claims priority to Chinese Patent Application No. 202411095792.5, filed on August 9, 2024, entitled “Image Processing System, Method and Image Stylization Platform”, the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of computer technology, and in particular to image processing systems and methods. This application also relates to an image stylization platform, a computing device, a computer-readable storage medium, and a computer program product. Background Technology

[0003] Driven by the global digital wave, the video game industry is undergoing unprecedented transformation and growth. With continuous technological advancements and increasingly diverse player demands, game personalization has become a key trend driving industry innovation.

[0004] Currently, personalized experiences have become a key factor in attracting and retaining players. In most games, players need to choose an avatar to represent their in-game identity. Most games on the market currently provide fixed avatars that users can choose to use, and some games also support custom avatar uploads. However, these two methods often fail to ensure that the avatar style matches the overall game style, and users cannot customize their avatars, thus limiting players' personal expression to some extent.

[0005] Therefore, how to achieve individual adjustable regional customization of stylistic features based on different facial organs has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, embodiments of this application provide an image processing system and method. This application also relates to an image stylization platform, a computing device, a computer-readable storage medium, and a computer program product, to solve the aforementioned problems existing in the prior art.

[0007] According to a first aspect of the embodiments of this application, an image processing system is provided, the image generation system comprising a client and an image stylization platform, wherein: The client sends an image processing request to the image stylization platform, wherein the image processing request includes stylization attribute information and stylization region information; The image stylization platform obtains the image to be processed based on the image processing request, obtains the regional feature values ​​of the image to be processed according to the stylization region information, and performs stylization processing on the image to be processed according to the regional feature values ​​and the stylization attribute information to obtain the target stylized image.

[0008] According to a second aspect of the embodiments of this application, an image processing method is provided, comprising: The image to be processed is obtained based on the received image processing request, wherein the image processing request includes stylized attribute information and stylized region information; The region feature values ​​of the image to be processed are obtained based on the stylized region information; The image to be processed is stylized based on the region feature values ​​and the stylization attribute information to obtain the target stylized image.

[0009] According to a third aspect of the embodiments of this application, an image stylization platform is provided, the stylization platform comprising an image processing module and an image adjustment module: The image processing module is configured to acquire an image to be processed based on the image processing request, acquire regional feature values ​​of the image to be processed according to the stylized region information, and perform stylized processing on the image to be processed according to the regional feature values ​​and the stylized attribute information to obtain a target stylized image. The image adjustment module is configured to adjust the target region of the target stylized image according to the adjustment region information in the image adjustment request.

[0010] According to a fourth aspect of the present application, a computing device is provided, including a memory, a processor, and a computer program / instructions stored in the memory and executable on the processor, wherein the processor executes the computer program / instructions to implement the steps of the above-described image processing method.

[0011] According to a fifth aspect of the present application, a computer-readable storage medium is provided that stores a computer program / instructions, which, when executed by a processor, implement the steps of the above-described image processing method.

[0012] According to a sixth aspect of the embodiments of this application, a computer program product is provided, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described image processing method.

[0013] One embodiment of this application implements the acquisition of an image to be processed based on a received image processing request, wherein the image processing request includes stylized attribute information and stylized region information; regional feature values ​​of the image to be processed are obtained according to the stylized region information; and stylized processing is performed on the image to be processed according to the regional feature values ​​and the stylized attribute information to obtain a target stylized image. By combining the regional features of the image to achieve stylized processing, the accuracy of image stylized processing is improved. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the structure of an image processing system provided in one embodiment of this application; Figure 2 This is a schematic diagram of the processing flow of an image processing system provided in an embodiment of this application; Figure 3 This is a schematic diagram of a model training process provided in one embodiment of this application; Figure 4 This is a flowchart of an image processing method provided in an embodiment of this application; Figure 5 This is a schematic diagram of an image stylization platform provided in an embodiment of this application; Figure 6 This is a structural block diagram of a computing device provided in one embodiment of this application. Detailed Implementation

[0015] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0016] The terminology used in one or more embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the scope of one or more embodiments of this application. The singular forms “a,” “the,” and “the” used in one or more embodiments of this application and in the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” used in one or more embodiments of this application refers to and includes any or all possible combinations of one or more associated listed items.

[0017] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this application, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this application, and similarly, second may also be referred to as first.

[0018] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0019] First, the terms and concepts involved in one or more embodiments of this application will be explained.

[0020] StyleGAN: StyleGAN is a powerful framework that can control the attributes of generated images. It generates new images by mapping random noise into the space of the generative model.

[0021] Gan invert: refers to the technique used when performing image translation using Generative Adversarial Networks (GANs), which inverts a given image back into the latent space of a pre-trained GAN model.

[0022] This application provides an image processing system and method, and also relates to an image stylization platform, a computing device, a computer-readable storage medium, and a computer program product, which will be described in detail in the following embodiments.

[0023] See Figure 1 , Figure 1 A schematic diagram of the structure of an image processing system provided in an embodiment of this application is shown, as follows: Figure 1 As shown, the image processing system includes a client 102 and an image stylization platform 104.

[0024] The client 102 sends an image processing request to the image stylization platform, wherein the image processing request includes stylization attribute information and stylization region information; The image stylization platform 104 obtains the image to be processed based on the image processing request, obtains the regional feature values ​​of the image to be processed according to the stylization region information, and performs stylization processing on the image to be processed according to the regional feature values ​​and the stylization attribute information to obtain the target stylized image.

[0025] Among them, the image stylization platform refers to a platform that can perform stylization processing on images, and the client refers to a terminal that can upload images to the image stylization platform or obtain and display images on the image stylization platform.

[0026] See Figure 2 , Figure 2 A schematic diagram of the processing flow of an image processing system provided in an embodiment of this application is shown, such as... Figure 2 As shown, the image system provided in this application embodiment includes a client and an image stylization platform.

[0027] Step 202: The client sends an image processing request to the image stylization platform, wherein the image processing request includes stylization attribute information and stylization region information.

[0028] Step 204: The image stylization platform obtains the image to be processed based on the image processing request, obtains the regional feature values ​​of the image to be processed according to the stylization region information, and performs stylization processing on the image to be processed according to the regional feature values ​​and the stylization attribute information to obtain the target stylized image.

[0029] After the target stylized image is generated on the image stylization platform, it can be sent to the client for display.

[0030] Among them, stylization attribute information refers to the attribute information for stylizing the image, such as adjusting the eyes of a person in a real-life image to cartoon-style eyes; stylization region information refers to the coordinates, positions, and other information of the regions in the image to be stylized; the image to be processed refers to the image that needs to be stylized; the region feature value refers to the feature value of the target region in the image to be processed; and the target stylized image refers to the image obtained after stylizing the image to be processed.

[0031] In scenarios involving stylization of game avatars (images containing faces), current facial stylization algorithms typically perform stylization transfer on the entire face. However, this doesn't always align with user expectations for the degree of stylization in different areas, as the stylization level varies across different facial features. Simply adjusting the overall stylization parameter cannot achieve a satisfactory level for each feature. For example, the eyes and nose might resemble the desired style, but the mouth might still look like its original form. In such cases, simply increasing the stylization transfer coefficient could lead to excessive stylization of the eyes and nose, resulting in distortion. Therefore, to provide users with greater customization options for different facial features, we've made substantial improvements to avatar stylization technology by offering individually adjustable, region-specific customized stylization.

[0032] Specifically, the process begins by detecting facial landmarks to determine the regions of different facial features. Then, regional stylization is customized in both training and generation stages. During training, a weighted method is used to adjust the stylization level of a region by averaging the difference in features between the discriminative network and the facial recognition network, enabling customized stylization editing for each region. In the actual generation process after training, the degree of stylization is customized by interpolating the feature values ​​of different layers of real faces and stylized faces within the corresponding region.

[0033] The discriminant network feature difference is calculated by inputting the stylized face image and the training samples of the stylized image into the discriminant network of StyleGAN, which is trained based on real face images. The L1 loss between the feature maps output from layers 1, 3, 4, and 5 of the discriminant network is then calculated as the discriminant network feature difference. This is used to improve style similarity while ensuring that the real face image and the training sample image are not completely similar. The face recognition network feature difference is calculated by extracting features from the input face and the stylized face image using the ArcFace face recognition network, calculating the cosine similarity between the features, and subtracting the cosine similarity from 1. The result is used as the face recognition network feature difference loss.

[0034] Furthermore, after obtaining the stylized image, the image stylization platform also has the function of adjusting the degree of stylization of local areas of the image according to user needs.

[0035] That is, the image stylization platform includes an image adjustment module; the image adjustment module receives the image adjustment request sent by the client, and adjusts the target area of ​​the target stylized image according to the adjustment area information in the image adjustment request.

[0036] The image adjustment module refers to the module that can adjust the stylization level of the target region of the image. The adjustment function of the image adjustment module is based on the difference of features of the discrimination network and the difference of features of the face recognition network. The image adjustment request refers to the request to adjust the stylization level of a local region in the stylized image. The adjustment region information refers to the location information of the region in the image that needs to be stylized.

[0037] In scenarios involving the generation of stylized game avatars, the image adjustment module can, in conjunction with user requests, edit the stylized image by region to enhance the richness and flexibility of image generation, thereby enabling personalized image generation.

[0038] In addition, to further enhance the richness of image generation, the image stylization platform also has the function of adjusting image attributes according to user needs.

[0039] The image stylization platform includes an object attribute editing module; the object attribute editing module receives an object attribute editing request sent by the client, and adjusts the object attribute information in the target stylized image based on the attribute adjustment in the object attribute editing request.

[0040] Among them, the object attribute editing module refers to the module that adjusts the attribute information of objects in the image; the object attribute editing request refers to the request to edit the attributes of objects in the image; the object attribute information refers to the attribute information of objects in the image, such as gender, age, face orientation, etc.

[0041] Specifically, the object attribute editing module first trains the manifold direction of the corresponding attribute in the face feature space through a face attribute classifier, and then adjusts the face feature value in the manifold direction to achieve changes in face attributes.

[0042] In practical applications, the stylization processing capabilities of image stylization platforms require model training to acquire the corresponding processing abilities. Specific training methods can include: Acquire a sample image, and perform key point detection on the objects contained in the sample image to obtain the key point coordinates; Adjust the coordinates of the key points to obtain the coordinates of the target key points, and generate a target sample image based on the coordinates of the target key points; The image processing module is trained based on the sample image and the target sample image to obtain a trained image processing module.

[0043] Specifically, stylization of faces with large deformations—that is, when the stylized result differs significantly from the real image—is ineffective. For example, StyleGAN-based algorithms are not effective at stylizing faces with large deformations. Since the StyleGAN base model is trained with a large number of real faces of normal proportions, when the stylized face proportions differ too much from normal faces, the organ feature positions of the base model trained with normal face proportions no longer correspond. In other words, the proportions of the real faces used in the original StyleGAN pre-training are too different, resulting in significant differences in the corresponding organ feature positions during pre-training, which easily leads to unsatisfactory stylization results.

[0044] To address the aforementioned problem, this application employs a progressive training method. Specifically, this involves performing keypoint detection on objects in sample images, then adjusting the detection results to obtain multiple sample images with varying degrees of stylization for progressive model training. The adjustment of the detection results can be achieved by: affine mapping the sample images to the proportions of real human faces, then training on these adjusted images until model convergence, and finally gradually adjusting the sample images back to their original proportions, thus achieving progressive training of the model.

[0045] For example, keypoint detection is performed on the face of the style template. Then, the keypoint coordinates of the style template face are interpolated with the keypoint coordinates of the average face of a real person according to different weights (e.g., 20% to 80%, 40% to 60%, 60% to 40%, and 20% to 80%) to obtain new face keypoint coordinates. By transforming the original face keypoints to the new face keypoint coordinates through affine transformation, the original style template face is transformed into multiple style template faces with different proportions.

[0046] After obtaining stylized templates with different proportions, the training process begins with a template face that more closely resembles the key points of a real human face. During training, the template face is replaced with one that is even closer to the stylized template after a certain number of iterations, until the original stylized template face is used for training. This progressive learning approach addresses the stylization of faces with large deformations, improves the unintended consequences of stylizing faces with large deformations, and enhances the accuracy of the stylization process.

[0047] Furthermore, during the training process targeting different stylizations, to enrich the sample images used for training, multiple style sample images can be generated from a single style sample image for model training, including: Obtain style sample images and encode the style sample images to obtain image codes; Adjust the image encoding to obtain at least one style image encoding, and generate a style sample image set based on each style image encoding; The image processing module is trained based on the style sample images in the style sample image set to obtain the trained image processing module.

[0048] Image coding is the encoding of the style sample image output by the encoder; adjustments to image coding can be made by applying random Gaussian fluctuations to the encoding layer in the encoder that is related to the image structure.

[0049] In a specific embodiment of this application, the model training process is described in detail in the scenario of processing human face images, including: Considering the significant server and time costs of training a model from scratch, StyleGAN, pre-trained on a large-scale dataset of real human faces, was chosen as the base model. The pre-trained model can be used for real human face generation, generating faces from random input signals. Specifically, it first transforms features from Z-space to W-space using fully connected layers to obtain the feature structure, and then uses convolution and adaptive instance regularization to embed different styles and generate images.

[0050] Considering the need to stylize faces across different styles to obtain two-dimensional stylized faces, and given the limited number of template images for each style, a single-sample face stylization technique was adopted that requires only a single template image for each style.

[0051] Specifically, such as Figure 3 As shown in the schematic diagram of a model training process provided in one embodiment of this application, the style template face image is first encoded using the Gaininvert method, i.e., a real face encoder. Each encoding layer in the face encoder corresponds to different attributes of the face, such as attributes related to structure and attributes related to texture (skin color, same color, hair color, etc.). In order to ensure the consistency of the overall structure of the face during attribute adjustment, this embodiment only performs random Gaussian fluctuations on the non-structure-related layers of encoding, such as texture-related layers, to obtain more face encodings, thereby obtaining more real images corresponding to style template faces. In the case of only one style template image, more data pairs of real images and corresponding stylized images are generated.

[0052] For example, in one specific embodiment, the face encoder includes 26 encoding layers. Layers 9, 11, 15, 16, and 17, which are texture-related layers, are selected and subjected to random Gaussian fluctuation processing to obtain multiple face codes.

[0053] Furthermore, the new data pairs are used to train and fine-tune the neural network generator, which has already been pre-trained on a real-person face dataset. The loss value between the real-person image obtained from the original real-person model and the stylized image generated by the stylization model is calculated, and then a more stable gradient descent is achieved through a weighted moving average. Specifically, the loss value is supervised by the feature difference of the pre-trained neural network discriminator and the feature difference of face recognition, and a more stable gradient descent is achieved through a weighted moving average. After fine-tuning, the face encoder can encode the facial features of the input real-person face, and then input it into the trained stylization network model to generate a stylized face image. For the problem of style template asymmetry in single-sample style transfer, data augmentation methods such as flipping, rotation, and random cropping can be used.

[0054] The above approach can solve the problem of insufficient sample size, and by training the corresponding style using only a single template image, the training efficiency of the model can be improved.

[0055] This application includes an image processing system comprising a client and an image stylization platform. During the training of the image processing module, an algorithm model capable of handling multiple styles is trained using single-sample data. This enables the model to accurately identify the features of user-uploaded photos and generate 2D stylized avatars with the user's personal style and typical features, reducing the need for stylized images used for training. Sample expansion technology overcomes the overfitting problem caused by training with few samples. Progressive learning addresses the problem of stylizing large-deformation faces by transforming the original facial keypoints to new facial keypoint coordinates through affine transformation, converting the original style template face into multiple style template faces of different proportions. This solves the uncertainty of the stylization effect for large deformations, ensuring that the generated avatar meets the expectations of stylization processing. By introducing the identification of facial recognition network feature differences in different regions, the training model is adjusted for customized facial style interpolation, achieving individually adjustable regional customized stylization technology based on different facial organ regions. This makes the stylization degree of different regions more consistent with user expectations, solving the technical problem of not being able to adjust the stylization effect of a single facial part.

[0056] Figure 4 The flowchart of an image processing method according to an embodiment of this application is shown, specifically including the following steps: Step 402: Obtain the image to be processed based on the received image processing request, wherein the image processing request includes stylized attribute information and stylized region information.

[0057] Step 404: Obtain the region feature values ​​of the image to be processed based on the stylized region information.

[0058] Step 406: Perform stylization processing on the image to be processed based on the region feature values ​​and the stylization attribute information to obtain the target stylized image.

[0059] Figure 5 The diagram shows an image stylization platform provided in one embodiment of this application, which includes an image processing module 502 and an image adjustment module 504.

[0060] The image processing module 502 is configured to acquire an image to be processed based on the image processing request, acquire regional feature values ​​of the image to be processed according to the stylized region information, and perform stylized processing on the image to be processed according to the regional feature values ​​and the stylized attribute information to obtain a target stylized image. The image adjustment module 504 is configured to adjust the target region of the target stylized image according to the adjustment region information in the image adjustment request.

[0061] Furthermore, the image stylization platform includes an image adjustment module; The image adjustment module receives an image adjustment request sent by the client and adjusts the target area of ​​the target stylized image according to the adjustment area information in the image adjustment request.

[0062] Furthermore, the image stylization platform includes an object attribute encoding module; The object attribute editing module receives an object attribute editing request sent by the client and adjusts the object attribute information in the target stylized image based on the attribute adjustment in the object attribute editing request.

[0063] Furthermore, the image processing module of the image stylization platform is trained through the following steps: Acquire a sample image, and perform key point detection on the objects contained in the sample image to obtain the key point coordinates; Adjust the coordinates of the key points to obtain the coordinates of the target key points, and generate a target sample image based on the coordinates of the target key points; The image processing module is trained based on the sample image and the target sample image to obtain a trained image processing module.

[0064] Furthermore, the image processing module of the image stylization platform is trained through the following steps: Obtain style sample images and encode the style sample images to obtain image codes; Adjust the image encoding to obtain at least one style image encoding, and generate a style sample image set based on each style image encoding; The image processing module is trained based on the style sample images in the style sample image set to obtain the trained image processing module.

[0065] The image stylization platform of this application is equipped with an image adjustment module and an object attribute editing module, which solves the limitation of the platform only having face stylization function. That is, by designing face attribute editing function on the basis of generating stylized images, it adds fine editing of attributes such as gender, expression, face orientation, hair volume, beard, age, etc., to obtain a user-friendly platform that integrates all functions into one image stylization platform.

[0066] Figure 6A structural block diagram of a computing device 600 according to an embodiment of this application is shown. The components of the computing device 600 include, but are not limited to, a memory 610 and a processor 620. The processor 620 is connected to the memory 610 via a bus 630, and a database 650 is used to store data.

[0067] The computing device 600 also includes an access device 640, which enables the computing device 600 to communicate via one or more networks 660. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 640 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, a Near Field Communication (NFC) interface, and so on.

[0068] In one embodiment of this application, the aforementioned components of the computing device 600 and Figure 6 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 6 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this application. Those skilled in the art can add or replace other components as needed.

[0069] The computing device 600 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 600 can also be a mobile or stationary server.

[0070] The processor 620 implements the steps of the image processing method when executing the computer program / instructions.

[0071] The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device and the technical solution of the image processing method described above belong to the same concept. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the image processing method described above.

[0072] An embodiment of this application also provides a computer-readable storage medium storing a computer program / instructions that, when executed by a processor, implement the steps of the image processing method as described above.

[0073] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium and the technical solution of the image processing method described above belong to the same concept. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the image processing method described above.

[0074] An embodiment of this application also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described image processing method.

[0075] The above is an illustrative scheme of a computer program product according to this embodiment. It should be noted that the technical solution of this computer program product and the technical solution of the image processing method described above belong to the same concept. For details not described in detail in the technical solution of the computer program product, please refer to the description of the technical solution of the image processing method described above.

[0076] The foregoing has described specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0077] The computer program / instructions include computer program code, which may be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0078] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in this application are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0079] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0080] The preferred embodiments disclosed above are merely illustrative of this application. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this application. These embodiments are selected and specifically described in this application to better explain the principles and practical applications of this application, thereby enabling those skilled in the art to better understand and utilize this application. This application is limited only by the claims and their full scope and equivalents.

Claims

1. An image processing system, characterized in that, The image generation system includes a client and an image stylization platform, wherein: The client sends an image processing request to the image stylization platform, wherein the image processing request includes stylization attribute information and stylization region information; The image stylization platform obtains the image to be processed based on the image processing request, obtains the regional feature values ​​of the image to be processed according to the stylization region information, and performs stylization processing on the image to be processed according to the regional feature values ​​and the stylization attribute information to obtain the target stylized image.

2. The image processing system as described in claim 1, characterized in that, The image stylization platform includes an image adjustment module; The image adjustment module receives an image adjustment request sent by the client and adjusts the target area of ​​the target stylized image according to the adjustment area information in the image adjustment request.

3. The image processing system as described in claim 1, characterized in that, The image stylization platform includes an object attribute encoding module; The object attribute editing module receives an object attribute editing request sent by the client and adjusts the object attribute information in the target stylized image based on the attribute adjustment in the object attribute editing request.

4. The image processing system as described in claim 1, characterized in that, The image processing module of the image stylization platform is trained through the following steps: Acquire a sample image and perform key point detection on the objects contained in the sample image to obtain the key point coordinates; Adjust the coordinates of the key points to obtain the coordinates of the target key points, and generate a target sample image based on the coordinates of the target key points; The image processing module is trained based on the sample image and the target sample image to obtain the trained image processing module.

5. The image processing system as described in claim 1, characterized in that, The image processing module of the image stylization platform is trained through the following steps: Obtain style sample images and encode the style sample images to obtain image codes; Adjust the image encoding to obtain at least one style image encoding, and generate a style sample image set based on each style image encoding; The image processing module is trained based on the style sample images in the style sample image set to obtain the trained image processing module.

6. An image processing method, characterized in that, include: The image to be processed is obtained based on the received image processing request, wherein the image processing request includes stylized attribute information and stylized region information; The region feature values ​​of the image to be processed are obtained based on the stylized region information; The image to be processed is stylized based on the region feature values ​​and the stylization attribute information to obtain the target stylized image.

7. An image stylization platform, characterized in that, The stylization platform includes an image processing module and an image adjustment module: The image processing module is configured to acquire an image to be processed based on the image processing request, acquire regional feature values ​​of the image to be processed according to the stylized region information, and perform stylized processing on the image to be processed according to the regional feature values ​​and the stylized attribute information to obtain a target stylized image. The image adjustment module is configured to adjust the target region of the target stylized image according to the adjustment region information in the image adjustment request.

8. A computing device, comprising a memory, a processor, and a computer program / instructions stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program / instructions, it implements the steps of the method of claim 6.

9. A computer-readable storage medium storing a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method of claim 6.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method of claim 6.

Images