CN117972673B - Semantic verification code generation method, device, equipment and medium based on style transfer - Google Patents

Abstract

The embodiment of the present invention discloses a method, device, equipment and medium for generating a semantic verification code based on style transfer, and relates to the field of verification code technology. The method comprises: constructing a style picture library based on style pictures; constructing a content picture library based on content pictures; constructing a style transfer picture library based on style transfer pictures; obtaining a target style picture and an interference style picture from the style transfer picture library, obtaining a target style transfer picture corresponding to the target style picture from the style transfer picture library, obtaining a target content picture corresponding to the target style transfer picture from the content picture library, and generating a semantic verification code based on the target style picture, the interference style picture, the target style transfer picture and the target content picture. The method proposed by the present invention is highly efficient in generating a semantic verification code of style transfer, and is easy for humans to accurately identify.

Description

Semantic verification code generation method, device, equipment and medium based on style transfer

Technical Field

The present invention relates to the field of verification code technology, and in particular to a semantic verification code generation method, device, equipment and medium based on style transfer.

Background technique

In recent years, neural style transfer has been mainly used to interfere with text content recognition in text-clicking graphic verification codes. However, stylized images still cannot hide the position of text or object targets in the image. Since modern OCR technology has a powerful ability to extract and recognize content from complex pattern images, the above method is easy to crack and has low security.

Existing literature proposes a style matching verification code based on neural network style transfer. After observing the content image and the style-transferred image, the user needs to perceive the semantic association between the two, and then select the correct style from the optional style images to match the perceived semantic association. This method cleverly turns the content recognition problem into a style recognition problem. Although the neural style transfer algorithm can easily stylize the content image according to the specified style image, it is a very difficult task for the machine to extract the image style from the stylized image. On the contrary, humans have a strong ability to distinguish different image styles. The style matching verification code has a good balance between usability and security. However, this method relies on fine-tuning the vgg19 neural network model, and a large number of images need to be prepared for training to realize the style transfer algorithm. The algorithm itself has the problem of long calculation time. The image data used in this document are all collected from public data sets, and the number of images is limited, making it difficult to put into production. On the other hand, when the interference style and the target style are similar in texture features, color, etc., it will increase the difficulty of human recognition. Simply increasing the number of image styles can only reduce the probability of problems.

Summary of the invention

The embodiments of the present invention provide a method, apparatus, device and medium for generating a semantic verification code based on style transfer, aiming to solve the problem that the existing semantic verification code generation based on style transfer is inefficient and difficult to identify.

In a first aspect, an embodiment of the present invention provides a method for generating a semantic verification code based on style transfer, which includes:

Acquire style text description information, generate style pictures matching the style text description information based on a preset style picture model, and build a style picture library based on the style pictures;

Acquire content text description information, generate content pictures matching the content text description information based on a preset text image macro model, and build a content picture library based on the content pictures;

The content image is obtained from the content image library and input into a preset image-generated image model; the style image is obtained from the style image library and input into a preset adapter; the adapter uses the style image as a guide for style control to control the image-generated image model to perform stylized migration on the content image to obtain a style-migrated image, and a style-migrated image library is constructed based on the style-migrated image;

A target style picture and an interference style picture are obtained from the style picture library, a target style migration picture corresponding to the target style picture is obtained from the style migration picture library, a target content picture corresponding to the target style migration picture is obtained from the content picture library, and a semantic verification code is generated based on the target style picture, the interference style picture, the target style migration picture and the target content picture.

A further technical solution is that the step of obtaining the style text description information includes:

Get objects from the preset object library, colors from the preset color library, and art styles from the preset art style library;

The object, the color and the artistic style are combined into the style text description information.

A further technical solution is that the method of generating a style picture matching the style text description information based on a preset style picture model includes:

Inputting the object, the color, and the artistic style included in the style text description information into a preset first prompt word template to obtain a first prompt word sentence, wherein the first prompt word template defines that the output image is a seamless texture image;

The first prompt word sentence is input into the style picture model to obtain a style picture matching the style text description information.

A further technical solution is that the step of obtaining the content text description information includes:

Get the content object from the preset content library and randomly select white or black as the background color;

The content object and the background color are combined into the content text description information.

A further technical solution is that the content image matching the content text description information is generated based on the preset text image macro model, including:

Inputting the content object and the background color included in the content text description information into a preset second prompt word template to obtain a second prompt word sentence;

The second prompt word sentence is input into the text image macro model to obtain a content image matching the content text description information.

A further technical solution is that obtaining the target style picture and the interference style picture from the style picture library includes:

Determining target style text description information and interference style text description information based on the object library, the color library, and the art style library, wherein the target style text description information and the interference style text description information both include objects, colors, and art styles;

The target style picture is obtained from the style picture library based on the target style text description information, and the interference style picture is obtained from the style picture library based on the interference style text description information.

A further technical solution is that the semantic verification code is generated based on the target style picture, the interference style picture, the target style migration picture and the target content picture, including:

Constructing a topic based on the target content picture and the target style transfer picture, and constructing options based on the target style picture and the interference style picture to generate the semantic verification code;

The method for generating a semantic verification code based on style transfer also includes:

Determine whether the option selected by the user is the target style picture;

If the option selected by the user is the target style picture, the verification is determined to be successful;

If the option selected by the user is not the target style image, the verification is determined to have failed.

In a second aspect, an embodiment of the present invention further provides a semantic verification code generation device based on style transfer, which includes a unit for executing the above method.

In a third aspect, an embodiment of the present invention further provides a computer device, which includes a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the above method when executing the computer program.

In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores a computer program, and the computer program can implement the above method when executed by a processor.

The technical solution of the embodiment of the present invention utilizes an adapter (such as ip-adapter) to transfer the style of an image. Compared with the traditional convolutional neural network such as VGG16, the adapter has a strong generalization ability and can transfer almost any image style without repeatedly training the model. It is also compatible with large image generation models (such as stable diffusion XL), which greatly improves the generation efficiency of style transfer images.

On the other hand, using a large model of Vincent graphs (for example, dalle3) to generate style images can accurately control the color, object and artistic style of the style, and generating interference style images can accurately avoid situations where the style images are similar or repeated to the target style images. This greatly improves the stability of generating style transfer verification code images that are easy for humans to pass. In terms of the ability to prevent black and gray industries from cracking, existing OCR technology has no effect on this style transfer verification code, and extracting style from images is still a very difficult task in the field of scientific research. Using a large model of Vincent graphs to generate a large number of style image samples can further increase the cost of the black and gray industries to collect images for manual labeling and training models. The present invention takes into account both ease of use and security.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

FIG1 is a schematic diagram of a flow chart of a method for generating a semantic verification code based on style transfer provided by an embodiment of the present invention;

FIG2 is another schematic flow chart of a method for generating a semantic verification code based on style transfer provided by an embodiment of the present invention;

FIG3 is a style picture provided by an embodiment of the present invention;

FIG4 is a content picture provided by an embodiment of the present invention;

FIG5 is a schematic diagram of a reference topic provided by an embodiment of the present invention;

FIG6 is a schematic diagram of a semantic verification code provided by an embodiment of the present invention;

FIG7 is a schematic block diagram of a semantic verification code generation device based on style transfer provided by an embodiment of the present invention;

FIG8 is a schematic block diagram of a computer device provided in an embodiment of the present invention.

Detailed ways

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

It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and/or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and/or combinations thereof.

It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

It should be further understood that the term "and/or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

As used in this specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

Referring to FIG. 1 and FIG. 2 , an embodiment of the present invention provides a semantic verification code generation method based on style transfer. As shown in FIG. 1 , the method includes the following steps:

S1. Acquire style text description information, generate style pictures matching the style text description information based on a preset style picture model, and build a style picture library based on the style pictures.

In specific implementation, style picture materials should select pictures of a single object, with a distinctive style, and simple and easy to identify. Seamless texture pictures are very suitable as style picture materials.

The style text description information includes information such as object, color, and artistic style, etc. Based on different objects, colors, and artistic styles, multiple style text description information can be combined.

The large model of texture images can be specifically dalle3. Dalle3 is a large model of texture images suitable for automatically generating texture images. Dalle3 has currently opened its API and Python library, and images can be generated for free through Python. Dalle3 has a strong natural language understanding ability and can simply construct text prompt words to generate images that are highly matched with text input. Using this feature, it is possible to automatically generate style images and label them with feature tags corresponding to the style.

By inputting the style text description information into the Wenshengtu model, the Wenshengtu model can generate a style picture matching the style text description information. Multiple style text description information can obtain multiple style pictures. Multiple style pictures are stored in a database to construct a style picture library.

In one embodiment, the above step of “obtaining style text description information” specifically includes:

S11, obtaining an object from a preset object library, obtaining a color from a preset color library, and obtaining an art style from a preset art style library.

In the specific implementation, an object library, a color library, and an art style library are pre-built. The object library includes multiple objects, such as walls, bubbles, etc. The color library includes multiple colors, such as white, black, and silver, etc. The art style library includes multiple art styles, such as science fiction cyberpunk style, etc.

Specifically, objects are acquired from a preset object library, colors are acquired from a preset color library, and art styles are acquired from a preset art style library.

S12, combining the object, the color and the artistic style into the style text description information.

In a specific implementation, an object is selected from the object library, a color is selected from the color library, and an art style is selected from the art style library to combine and obtain a style text description information. Furthermore, by exhaustively enumerating all combinations, multiple style text description information can be obtained, and then multiple style images can be obtained accordingly.

In one embodiment, the above step of “generating a style picture matching the style text description information based on a preset style picture model” specifically includes:

S101, inputting the object, the color and the artistic style included in the style text description information into a preset first prompt word template to obtain a first prompt word sentence, wherein the first prompt word template defines that the output image is a seamless texture image.

In a specific implementation, the first prompt word template reserves the filling positions of the object, the color and the artistic style. At the same time, in the first prompt word template, it is limited that the output image is a seamless texture image. Specifically, the first prompt word template uses "seamless texture" as the activation word for the seamless texture image to ensure that the output image is a texture image, which is convenient for users to distinguish. For example, in one embodiment, the prompt word template is: (color), (object) seamless texture image, (artistic style).

In one embodiment, the first prompt word sentence is: highly saturated blue + silver metal wall seamless texture map, science fiction cyberpunk style. The object is a metal wall, the color is highly saturated blue + silver, and the art style is science fiction cyberpunk style.

It is understandable that, in order to facilitate dale3 recognition, the first prompt word sentence can be translated into English. The translation process can be implemented by calling the interface of the translation software.

S102: Input the first prompt word sentence into the style picture model to obtain a style picture matching the style text description information.

In a specific implementation, the first prompt word sentence is input into the Wenshengtu model, and the Wenshengtu model generates a style picture matching the style text description information based on the first prompt word sentence. Specifically, the first prompt word sentence is input into the Wenshengtu model through the python library calling the dalle3 interface to generate a style picture. It can be set to output 4 pictures at a time to improve the diversity of pictures of a single style. Then, the dalle3 interface is called by python to download the style picture, and the style picture is imported into the style picture library. At the same time, the local storage path of the style picture, the generation time, and the corresponding object, color, artistic style and other style-related feature fields are generated.

For example, the style image generated based on the first prompt word sentence is shown in FIG3 .

S2. Obtain content text description information, generate content pictures matching the content text description information based on a preset text-image macro model, and build a content picture library based on the content pictures.

In the specific implementation, the content image is the basic image used to synthesize with the style image to generate the stylized image, which can be understood as the pad image in the image generation model. The content image does not require overly complex content composition, nor does it require a specific artistic style. The content image should try to choose a clear subject, a simple background, and a picture close to real life.

In the embodiment of the present invention, the content text description information includes content objects and background colors, and the background colors are black or white. Based on different content objects and background colors, multiple content text description information can be combined. The large model of the text image can be specifically dalle3.

By inputting the content text description information into the text image model, the text image model can generate a content image matching the content text description information. Multiple content text description information can obtain multiple content images. Multiple content images are stored in a database to construct a content image library.

In one embodiment, the above step of “obtaining content text description information” specifically includes:

S21, obtaining a content object from a preset content library, and randomly selecting white or black as the background color.

In a specific implementation, a content library is pre-built, and the content library includes multiple content objects, such as robots, houses, etc. Content objects are randomly obtained from the preset content library, and white or black is randomly selected as the background color.

S22: Combining the content object and the background color into the content text description information.

In a specific implementation, an object and a background color can be combined to obtain a content text description information. Exhaustive combinations of all objects and background colors can obtain multiple content text description information, and then multiple content images can be obtained.

In one embodiment, the above step of “generating a content image matching the content text description information based on a preset text image macro model” specifically includes:

S201: Input the content object and the background color included in the content text description information into a preset second prompt word template to obtain a second prompt word sentence.

In a specific implementation, the second prompt word template reserves positions for content objects and background colors. For example, in one embodiment, the prompt word template is: a (content object) in a (background color) background, photorealism.

For example, a content object is randomly selected from the content library and input into the second prompt word template. The object can have an adjective attribute, such as "a steel robot in the shape of a monster". The robot is the object, and the steel in the shape of a monster is the attribute adjective. Attribute adjectives can increase the diversity of content images. Randomly select white or black as the background color, and the second prompt word sentence obtained is: a white steel robot in the shape of a monster on a black background, photorealism.

It is understandable that, in order to facilitate dalle3 recognition, the second prompt word sentence can be translated into English. The translation process can be implemented by calling the interface of the translation software.

S202: Input the second prompt word sentence into the text-generated image macro model to obtain a content image that matches the content text description information.

In a specific implementation, the second prompt word sentence is input into the text-generated image large model, and the text-generated image large model generates a content image matching the content text description information based on the second prompt word sentence.

Specifically, the second prompt word sentence is input into the text image model through the python library calling the dalle3 interface to generate content images. It can be set to output 4 images at a time to improve the image diversity of a single content. Then, the dalle3 interface is called by python to download the content image, import the content image into the content image library, and generate the local storage path of the content image, the generation time, and the corresponding content object, background color and other content-related feature fields.

For example, a content picture generated based on the second prompt word sentence is shown in FIG4 .

S3. Obtain the content image from the content image library and input it into a preset image-generated large model. Obtain the style image from the style image library and input it into a preset adapter. The adapter uses the style image as a guide for style control to control the image-generated large model to perform stylized migration on the content image to obtain a style-migrated image, and construct a style-migrated image library based on the style-migrated image.

In the specific implementation, the image generation model can be specifically a stable diffusion model. The adapter is Ip-adapter. Ip-adapter is a lightweight adapter that uses image input as a prompt word to guide the stable diffusion model to generate images. By using the content image as the input of the stable diffusion image generation mode and the style image input to the ip-adapter as a guide for style control, the image stylization migration is achieved, and the purpose of automatic style migration can be achieved. The advantage of ip-adapter is its flexibility and strong compatibility. Only one style image is needed to guide the image generation model, and there is no need to prepare an image fine-tuning model. The function of ip-adapter is not to simply guide style generation, but to convert the content style and other features of the reference image into text prompts combined with the pad image to guide the image generation. As long as the redrawing amplitude of the image generation is controlled between 0.5-0.55, the content of the image generation will be more consistent with the content of the pad image, and will not be too affected by the reference image, while the style will be greatly affected by the reference image. Therefore, ip-adapter can better achieve the effect of style migration of images. The specific operation steps are as follows:

Get a content image from the content image library, input ip-adapter, configure the parameters, get a style image from the style image library, input stablediffusion XL 1.0base image generation mode, set the image size to 1024x1024, select 0.55 for the redrawing amplitude, set the sampling method to DPM+2M SDE Karras, set the sampling step to 32, and set the batch size to 4 (one content image and one style image generate four style-transferred images). Import the generated stylized image into the stylized image library, and generate the storage path, generation time, content image ID, style image ID and other associated fields of the stylized image.

S4. Obtain a target style picture and an interference style picture from the style picture library, obtain a target style migration picture corresponding to the target style picture from the style migration picture library, obtain a target content picture corresponding to the target style migration picture from the content picture library, and generate a semantic verification code based on the target style picture, the interference style picture, the target style migration picture and the target content picture.

In a specific implementation, a target style picture and an interference style picture are obtained from the style picture library. The target style picture is one, and the interference style picture may be multiple, for example, 8 pictures.

A target style-transferred picture corresponding to the target style picture is obtained from the style-transferred picture library, that is, a style-transferred picture obtained after style transfer based on the target style picture.

A target content picture corresponding to the target style-transferred picture is obtained from the content picture library, that is, a content picture of the target style-transferred picture before style migration.

Furthermore, a topic is constructed based on the target content picture and the target style transfer picture, and an option is constructed based on the target style picture and the interference style picture to generate the semantic verification code.

In order to help users understand the meaning of the questions, reference questions can be provided, as follows: randomly select a style-transferred picture from the style-transferred picture library as the answer to the reference question, and then select a content picture and a style picture corresponding to the style-transferred picture from the content picture library and the style picture library respectively to form the reference question.

Specifically, the question may be: target content picture +? = target style transfer picture. The options are composed of the target style picture and the interference style picture.

Based on the above semantic verification code, the specific verification process includes: determining whether the option selected by the user is the target style picture; if the option selected by the user is the target style picture, determining that the verification is passed; if the option selected by the user is not the target style picture, determining that the verification is failed.

In one embodiment, the above step of “obtaining the target style picture and the interference style picture from the style picture library” specifically includes:

S41 : determining target style text description information and interference style text description information based on the object library, the color library, and the artistic style library, wherein the target style text description information and the interference style text description information both include objects, colors, and artistic styles.

In the specific implementation, three objects are randomly obtained from the object library, which are different from the objects in the style image in the reference topic. One of the three objects is randomly selected as the target object, and the remaining two are used as interference objects.

Randomly select colors and artistic styles from the color library and the artistic style library to form three color-artistic style combinations, requiring that the colors and artistic styles of the three color-artistic style combinations are different from each other, randomly select one color-artistic style combination as the target color-artistic style combination, and the other two color-artistic style combinations as interference color-artistic style combinations. The target object and the target color-artistic style combination constitute the target style text description information, and the target object and the two interference color-artistic style combinations respectively constitute two interference style text description information.

Furthermore, for the other two interference objects, six color-artistic style combinations are selected from the color library and the art style library. The six color-artistic style combinations are required to be different from the target color-artistic style combination in one of the colors or art styles. The six color-artistic style combinations are not required to be different from each other, and even if there are the same six color-artistic style combinations, it is OK, although the probability of this happening is extremely low (when the color library and the art style library have sufficient data). After that, one interference object and three of the above six color-artistic style combinations respectively form three interference style text description information, and the two interference objects can obtain six interference style text description information.

Thus, a total of 1 target style text description information and 8 interference style text description information can be obtained.

S42: acquiring the target style picture from the style picture library based on the target style text description information, and acquiring the interference style picture from the style picture library based on the interference style text description information.

In a specific implementation, after the above step S42, 1 target style text description information and 8 interference style text description information are obtained. Based on the target style text description information, the target style picture can be obtained from the style picture library, and based on the 8 interference style text description information, 8 interference style pictures can be obtained from the style picture library.

The verification code generation example is as follows:

The reference topic is shown in Figure 5.

The object of the style picture in the reference question is poisonous bubbles. Then randomly select three objects: snowflake-shaped crystal, stone wall and carpet. Then randomly select carpet as the target object. Next, randomly select three combinations of blue-green photography style, red-blue national trend style, gold and emerald green decorative art style, and randomly select blue-green photography style as the target style picture color and art style combination, and the other two are interference combinations. For the two objects of snowflake-shaped crystal and stone wall, select 3 combinations different from blue-green photography style from the style picture library and combine them with 2 objects to form another 6 interference combinations. For example, brass steampunk stone wall, silver-blue science fiction cyberpunk stone wall, gray-white photography style stone wall. Dark gold crystal combined with ink sketch and watercolor. Blue-green abstract expressionism crystal, cartoon style purple crystal. The semantic verification code obtained at the end is shown in Figure 6.

The technical solution of the embodiment of the present invention utilizes an adapter (such as ip-adapter) to transfer the style of an image. Compared with the traditional convolutional neural network such as VGG16, the adapter has a strong generalization ability and can transfer almost any image style without repeatedly training the model. It is also compatible with large image generation models (such as stable diffusion XL), which greatly improves the generation efficiency of style transfer images.

On the other hand, using a large model of Vincent graphs (for example, dalle3) to generate style images can accurately control the color, object and artistic style of the style, and generating interference style images can accurately avoid situations where the style images are similar or repeated to the target style images. This greatly improves the stability of generating style transfer verification code images that are easy for humans to pass. In terms of the ability to prevent black and gray industries from cracking, existing OCR technology has no effect on this style transfer verification code, and extracting style from images is still a very difficult task in the field of scientific research. Using a large model of Vincent graphs to generate a large number of style image samples can further increase the cost of the black and gray industries to collect images for manual labeling and training models. The present invention takes into account both ease of use and security.

See Figure 7, which is a schematic block diagram of a semantic verification code generation device 20 based on style transfer provided by an embodiment of the present invention. Corresponding to the above semantic verification code generation method based on style transfer, the present invention also provides a semantic verification code generation device 20 based on style transfer. The semantic verification code generation device 20 based on style transfer includes a unit for executing the above semantic verification code generation method based on style transfer, and the semantic verification code generation device 20 based on style transfer can be configured in a desktop computer, a tablet computer, a laptop, and other terminals. Specifically, the semantic verification code generation device 20 based on style transfer includes:

The first construction unit 21 is used to obtain style text description information, generate a style picture matching the style text description information based on a preset style picture model, and construct a style picture library based on the style picture;

A second construction unit 22 is used to obtain content text description information, generate content pictures matching the content text description information based on a preset text image macro model, and construct a content picture library based on the content pictures;

The third construction unit 23 is used to obtain the content image from the content image library and input it into a preset image-generated image model, obtain the style image from the style image library and input it into a preset adapter, the adapter uses the style image as a guide for style control, controls the image-generated image model to perform stylized migration on the content image, obtains a style-migrated image, and constructs a style-migrated image library based on the style-migrated image;

A generating unit 24 is used to obtain a target style picture and an interference style picture from the style picture library, obtain a target style migration picture corresponding to the target style picture from the style migration picture library, obtain a target content picture corresponding to the target style migration picture from the content picture library, and generate a semantic verification code based on the target style picture, the interference style picture, the target style migration picture and the target content picture.

In one embodiment, the obtaining of the style text description information includes:

Get objects from the preset object library, colors from the preset color library, and art styles from the preset art style library;

The object, the color and the artistic style are combined into the style text description information.

In one embodiment, the generating of a style picture matching the style text description information based on a preset style picture model includes:

Inputting the object, the color, and the artistic style included in the style text description information into a preset first prompt word template to obtain a first prompt word sentence, wherein the first prompt word template defines that the output image is a seamless texture image;

The first prompt word sentence is input into the style picture model to obtain a style picture matching the style text description information.

In one embodiment, obtaining content text description information includes:

Get the content object from the preset content library and randomly select white or black as the background color;

The content object and the background color are combined into the content text description information.

In one embodiment, the generating of a content image matching the content text description information based on a preset text image macro model includes:

Inputting the content object and the background color included in the content text description information into a preset second prompt word template to obtain a second prompt word sentence;

The second prompt word sentence is input into the text image macro model to obtain a content image matching the content text description information.

In one embodiment, obtaining the target style picture and the interference style picture from the style picture library includes:

Determining target style text description information and interference style text description information based on the object library, the color library, and the art style library, wherein the target style text description information and the interference style text description information both include objects, colors, and art styles;

The target style picture is obtained from the style picture library based on the target style text description information, and the interference style picture is obtained from the style picture library based on the interference style text description information.

In one embodiment, the generating a semantic verification code based on the target style picture, the interference style picture, the target style migration picture and the target content picture includes:

Constructing a topic based on the target content picture and the target style transfer picture, and constructing options based on the target style picture and the interference style picture to generate the semantic verification code;

The semantic verification code generation device 20 based on style transfer also includes:

A judging unit, used to judge whether the option selected by the user is the target style picture;

A first determination unit, configured to determine that the verification is successful if the option selected by the user is the target style picture;

The second determination unit is configured to determine that the verification fails if the option selected by the user is not the target style picture.

It should be noted that those skilled in the art can clearly understand that the specific implementation process of the above-mentioned semantic verification code generation device 20 based on style transfer and each unit can refer to the corresponding description in the aforementioned method embodiment, and for the convenience and brevity of description, it will not be repeated here.

The above-mentioned semantic verification code generation device based on style transfer can be implemented in the form of a computer program, and the computer program can be run on the computer device shown in FIG. 8 .

Please refer to FIG8 , which is a schematic block diagram of a computer device provided in an embodiment of the present application. The computer device 500 may be a terminal or a server, wherein the terminal may be an electronic device with communication functions such as a smart phone, a tablet computer, a laptop computer, a desktop computer, a personal digital assistant, and a wearable device. The server may be an independent server or a server cluster composed of multiple servers.

The computer device 500 includes a processor 502 , a memory, and a network interface 505 connected via a system bus 501 , wherein the memory may include a non-volatile storage medium 503 and an internal memory 504 .

The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. When the computer program 5032 is executed, the processor 502 may execute a semantic verification code generation method based on style transfer.

The processor 502 is used to provide computing and control capabilities to support the operation of the entire computer device 500 .

The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a semantic verification code generation method based on style transfer.

The network interface 505 is used to communicate with other devices over the network.

The above structure is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device 500 to which the solution of the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

The processor 502 is used to run a computer program 5032 stored in the memory to implement the steps of a semantic verification code generation method based on style transfer provided in any of the above method embodiments:

It should be understood that in the embodiment of the present application, the processor 502 may be a central processing unit (CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

It can be understood by those skilled in the art that all or part of the processes in the method for implementing the above embodiment can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a storage medium, which is a computer-readable storage medium. The computer program is executed by at least one processor in the computer system to implement the process steps of the embodiment of the above method.

Therefore, the present invention also provides a storage medium. The storage medium may be a computer-readable storage medium. The storage medium stores a computer program. When the computer program is executed by a processor, the processor executes the steps of a semantic verification code generation method based on style transfer provided in any of the above method embodiments.

The storage medium is a physical, non-transient storage medium, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk, or an optical disk, etc., which can store program codes. The computer-readable storage medium can be non-volatile or volatile.

Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of each unit is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

The steps in the method of the embodiment of the present invention can be adjusted in order, combined and deleted according to actual needs. The units in the device of the embodiment of the present invention can be combined, divided and deleted according to actual needs. In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, terminal, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention.

In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or substitutions within the technical scope disclosed by the present invention, and these modifications or substitutions should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims (10)

1. A semantic verification code generation method based on style migration is characterized by comprising the following steps:

acquiring style text description information, generating a style picture matched with the style text description information based on a preset text generation graph large model, and constructing a style picture library based on the style picture;

acquiring content text description information, generating a content picture matched with the content text description information based on a preset text-generated graph large model, and constructing a content picture library based on the content picture;

Acquiring the content picture from the content picture library, inputting the content picture into a preset photo-generated large model, acquiring the style picture from the style picture library, inputting the style picture into a preset adapter, and controlling the photo-generated large model to perform stylized migration on the content picture by using the style picture as a guide of style control by the adapter to obtain a style migration picture, and constructing a style migration picture library based on the style migration picture;

obtaining a target style picture and an interference style picture from the style picture library, obtaining a target style migration picture corresponding to the target style picture from the style migration picture library, obtaining a target content picture corresponding to the target style migration picture from the content picture library, and generating a semantic verification code based on the target style picture, the interference style picture, the target style migration picture and the target content picture.

2. The style migration-based semantic verification code generation method according to claim 1, wherein the obtaining style text description information comprises:

Acquiring an object from a preset object library, acquiring a color from a preset color library and acquiring an artistic style from a preset artistic style library;

And composing the object, the color and the artistic style into the style text description information.

3. The semantic verification code generation method based on style migration according to claim 2, wherein the generating a style picture matched with the style text description information based on the preset text-generated graph large model comprises:

inputting the object, the color and the artistic style contained in the style text description information into a preset first prompting word template to obtain a first prompting word sentence, wherein the first prompting word template defines that the output picture is a seamless texture picture;

and inputting the first prompt word sentence into the draft map large model to obtain a style picture matched with the style text description information.

4. The style migration-based semantic verification code generation method according to claim 1, wherein the acquiring content text description information comprises:

obtaining a content object from a preset content library, and randomly selecting white or black as background color;

And forming the content text description information by the content object and the background color.

5. The style migration-based semantic verification code generation method according to claim 4, wherein the generating a content picture matching the content text description information based on the preset text-to-text graph large model comprises:

Inputting the content object and the background color contained in the content text description information into a preset second prompting word template to obtain a second prompting word sentence;

And inputting the second prompt word sentence into the draft graph large model to obtain a content picture matched with the content text description information.

6. The method for generating semantic verification codes based on style migration according to claim 3, wherein the step of obtaining the target style picture and the interference style picture from the style picture library comprises:

Determining target style text description information and disturbance style text description information based on the object library, the color library and the artistic style library, wherein the target style text description information and the disturbance style text description information comprise objects, colors and artistic styles;

And acquiring the target style picture from the style picture library based on the target style text description information, and acquiring the interference style picture from the style picture library based on the interference style text description information.

7. The style migration-based semantic verification code generation method according to claim 1, wherein the generating the semantic verification code based on the target style picture, the interference style picture, the target style migration picture, and the target content picture comprises:

constructing a theme based on the target content picture and the target style migration picture, and generating the semantic verification code based on the target style picture and the interference style picture construction option;

the semantic verification code generation method based on style migration further comprises the following steps:

judging whether the option selected by the user is the target style picture or not;

If the option selected by the user is the target style picture, judging that the verification is passed;

and if the option selected by the user is not the target style picture, judging that the verification is not passed.

8. A semantic verification code generating apparatus based on style migration, comprising means for performing the method of any one of claims 1-7.

9. A computer device, characterized in that it comprises a memory on which a computer program is stored and a processor which, when executing the computer program, implements the method according to any of claims 1-7.

10. A computer readable storage medium, characterized in that the storage medium stores a computer program which, when executed by a processor, implements the method according to any of claims 1-7.