Graphic information processing method, device, equipment, storage medium and program product

Abstract

The present invention provides a method, device, equipment, storage medium and program product for processing graphic information, which are applied to the field of computer vision technology. The method includes: obtaining first graphic information, the first graphic information includes a first text and a first image; adding an indication token to the first image to obtain second graphic information, the indication token includes an image start token and an image end token, the image start token is located before the first image, and the image end token is located after the first image; determining the text features corresponding to the first text, determining the image features corresponding to the first image, determining the token features corresponding to the indication token, and determining an input sequence based on the text features, the image features and the token features; inputting the input sequence into a large language model to obtain corresponding graphic understanding positioning information.

Description

Technical Field

[0001] The present invention relates to the field of computer vision technology, and in particular to a method, device, equipment, storage medium and program product for processing graphic information. Background Art

[0002] In the development of artificial intelligence, multimodal understanding and localization tasks are an important research direction. Existing multimodal models often face the problem of unclear reference between text and image and difficulty in understanding the relationship between images when dealing with complex multi-image and text scenes. These problems make it difficult for the model to effectively handle multi-image and text understanding and localization tasks in practical applications.

[0003] In the prior art, a combination of a visual encoder and a large language model can be used to improve the model's ability to understand the relationship between text and images and between multiple images. However, when it is applied to the task of understanding and localizing multiple images, it does not achieve good results. Summary of the invention

[0004] The present invention provides a method, device, equipment, storage medium and program product for processing graphic information, which are used to solve the problem that the prior art cannot improve the model's ability to understand the relationship between text and image and the relationship between multiple images in multi-graphic understanding and positioning tasks.

[0005] The present invention provides a method for processing graphic information, comprising: obtaining first graphic information, wherein the first graphic information comprises a first text and a first image; adding an indication token to the first image to obtain second graphic information, wherein the indication token comprises an image start token and an image end token, wherein the image start token is located before the first image, and the image end token is located after the first image; determining text features corresponding to the first text, determining image features corresponding to the first image, determining token features corresponding to the indication token, and determining an input sequence according to the text features, the image features and the token features; and inputting the input sequence into a large language model to obtain corresponding graphic understanding positioning information.

[0006] According to a method for processing graphic information provided by the present invention, the first graphic information also includes a second image, and the second image includes at least one image; the method also includes: adding indication tokens to the at least one image respectively to obtain the second graphic information; wherein the indication token of the first image corresponds to a first identifier, and the indication token of the second image corresponds to a second identifier.

[0007] According to a method for processing graphic information provided by the present invention, the first image includes at least two objects, and the at least two objects include a target object with graffiti information added; determining the image features corresponding to the first image includes: segmenting the first mask features corresponding to the target object in the first image through a segmentation model, and inputting the first mask features into the graffiti channel of the visual encoder to determine the image features.

[0008] According to a method for processing graphic information provided by the present invention, the value of the graffiti channel includes a first value and a second value, the first value represents the image foreground, and the second value represents the image background.

[0009] The present invention also provides a graphic information processing device, comprising the following modules: an acquisition module and a processing module; the acquisition module is used to acquire first graphic information, the first graphic information includes a first text and a first image; the processing module is used to add an indication token to the first image to obtain second graphic information, the indication token includes an image start token and an image end token, the image start token is located before the first image, and the image end token is located after the first image; determine the text features corresponding to the first text, determine the image features corresponding to the first image, determine the token features corresponding to the indication token, and determine an input sequence based on the text features, the image features and the token features; input the input sequence into a large language model to obtain corresponding graphic understanding positioning information.

[0010] According to a graphic information processing device provided by the present invention, the first graphic information also includes a second image, and the second image includes at least one image; the processing module is also used to add indication tokens to the at least one image respectively to obtain the second graphic information; wherein the indication token of the first image corresponds to a first identifier, and the indication token of the second image corresponds to a second identifier.

[0011] According to a graphic information processing device provided by the present invention, the first image includes at least two objects, and the at least two objects include a target object with graffiti information added; the processing module is used to segment the first mask feature corresponding to the target object in the first image through a segmentation model, and input the first mask feature into the graffiti channel of the visual encoder to determine the image feature.

[0012] According to a graphic information processing device provided by the present invention, the value of the graffiti channel includes a first value and a second value, the first value represents the image foreground, and the second value represents the image background.

[0013] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, any of the above-mentioned graphic information processing methods is implemented.

[0014] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for processing graphic information as described above is implemented.

[0015] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned graphic information processing methods.

[0016] The graphic information processing method, device, equipment, storage medium and program product provided by the present invention can obtain first graphic information, wherein the first graphic information includes a first text and a first image; add an indication token to the first image to obtain second graphic information, wherein the indication token includes an image start token and an image end token, wherein the image start token is located before the first image, and the image end token is located after the first image; determine the text features corresponding to the first text, determine the image features corresponding to the first image, determine the token features corresponding to the indication token, and determine an input sequence based on the text features, the image features and the token features; input the input sequence into a large language model to obtain corresponding graphic understanding positioning information. Through this scheme, since an indication token can be added to the first image in the first graphic information and an input sequence containing token features can be obtained, the ability of the large language model to distinguish images can be improved, thereby enabling the model to understand the complex referential relationship between text and image, thereby improving the ability of the large language model to understand the relationship between text and image. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 This is one of the flow charts of the method for processing graphic information provided by the present invention;

[0019] Figure 2 This is the second flow chart of the method for processing graphic information provided by the present invention;

[0020] Figure 3 It is a processing flow diagram of the visual encoder provided by the present invention;

[0021] Figure 4 It is a structural schematic diagram of the graphic information processing device provided by the present invention;

[0022] Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0024] It should be noted that, in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0025] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0026] In order to facilitate a clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, words such as "first" and "second" are used to distinguish between identical or similar items with substantially the same functions and effects. Those skilled in the art can understand that words such as "first" and "second" do not limit the quantity and order of execution.

[0027] The embodiments of the present application describe some exemplary embodiments for the purpose of explanation. It should be understood that the present application can be implemented in other ways that are not specifically shown in the drawings.

[0028] like Figure 1 As shown, the embodiment of the present application provides a method for processing graphic information, which can be applied to a graphic information processing device. The graphic information processing method may include S101-S104:

[0029] S101: A graphic information processing device obtains first graphic information.

[0030] The first graphic information includes a first text and a first image.

[0031] Optionally, the first graphic information may be a query instruction related to the first image. For example, the first graphic information may be “please find the main subject in the first image”, “please output the position coordinates of the main subject in the first image”, etc.

[0032] Optionally, the first image may include at least two objects, and the at least two objects include a target object to which graffiti information is added.

[0033] Specifically, when referring to an image, the image often contains only one object, and the model needs to return the mask of the object that belongs to the same category as the query object in the image to be retrieved. The reference form of the image with graffiti information is because the image contains multiple objects at the same time, and the user needs to use graffiti such as dots, circles, and lines to highlight the object to be referred to, and then let the model return the mask position of the object of the same category as the graffiti object from the image to be retrieved.

[0034] That is, when the first image includes multiple objects, the target object that the user wants to refer to can be determined through the graffiti information. For example, the first graphic information can be "Please find the photographed object referred to by the graffiti information in the first image".

[0035] S102: The graphic information processing device adds an indication token to the first image to obtain second graphic information.

[0036] The above-mentioned indication token includes an image start token and an image end token, wherein the image start token is located before the first image, and the image end token is located after the first image.

[0037] Typically, users may use text descriptions in query instructions to refer to specific images. This reference can provide the multimodal large language model with information about the visual content mentioned in the text, enabling it to learn the alignment relationship between the two modalities. In order to more accurately link text and images, this application can generate an image declaration template for each image in the multi-image and text mixed input.

[0038] Specifically, for the first graphic information, the graphic information processing device can use a relationship-enhanced unified expression paradigm of multiple graphic instructions to adjust it. That is, before and after the image features, exclusive image start tokens and image end tokens are assigned to refer to the visual features of the first image, and at the same time, natural language prompts are used to establish the reference between text and image, and an explicit text reference "first image" is added to the image declaration and placed before the image start token. In this way, the model's ability to associate text with corresponding images can be greatly improved, increasing the flexibility of users' questions. After that, no matter where the user uses the text "first image", the model can understand the reference content of the "first image".

[0039] Optionally, the first graphic information also includes a second image, and the second image includes at least one image; the graphic information processing device can also add indication tokens to the at least one image respectively to obtain the second graphic information; wherein, the indication token of the first image corresponds to the first identifier, and the indication token of the second image corresponds to the second identifier.

[0040] That is, when the first graphic information includes multiple images, one indication token may correspond to one token identifier, and different exclusive identifiers may be provided for different images, so that the model can distinguish different images, thereby facilitating the model to understand the complex relationship between multiple images.

[0041] For example, Figure 2 As shown, take the first image as image i and the second image as image j as an example. If the first graphic information is "Please find the target object referred to by the graffiti information in image j "image j" in image i "image i", then the second graphic information after adding the indication token can be "Please find the target object referred to by the graffiti information in image i "Image i" Find image j "ImageJ" The target object referred to by the graffiti information.

[0042] S103. The graphic information processing device determines text features corresponding to the first text, determines image features corresponding to the first image, determines token features corresponding to the indication token, and determines an input sequence according to the text features, the image features, and the token features.

[0043] Specifically, the graphic information processing device can determine the text features corresponding to the first text through a word segmenter, determine the image features corresponding to the first image and the token features corresponding to the indicator token through a visual encoder, and reorder the text features, the image features and the token features to obtain an input sequence.

[0044] Optionally, the visual encoder may include a graffiti channel. When the first image includes graffiti information, the graphic information processing device may segment the first mask feature corresponding to the target object in the first image through a segmentation model, and input the first mask feature into the graffiti channel of the visual encoder to determine the image feature.

[0045] Specifically, in order to unify the two reference modes of image and image with graffiti, this application designs the use of an additional graffiti channel to highlight the object referred to by the user in the image. For the input of the entire image, it is only necessary to expand the graffiti range to the entire image, that is, the entire image is the object to be referred to, and the unification of the reference paradigm can be completed.

[0046] Optionally, the value of the graffiti channel includes a first value and a second value, the first value represents the image foreground, and the second value represents the image background.

[0047] Specifically, Figure 3 As shown, the visual encoder with graffiti channel proposed in this application is based on the original CLIP image encoder with subtle structural modifications, which can inherit the pre-training parameters of CLIP and retain prior knowledge. In the ViT structure of the CLIP image encoder, this application applies RGB convolution to the image in the first layer. In addition, this application introduces an additional graffiti convolution layer, which is parallel to the RGB convolution layer, so that the CLIP image encoder can accept additional graffiti channels as input. The value range of the graffiti channel is set between [0, 1], where 1 represents the foreground and 0 represents the background. In order to allow the user's graffiti in any form to be accurately understood by the model, this application first uses the segmentation model to predict the masks of all objects in the image, and then uses the matching algorithm to match the mask with the user's graffiti, and selects the mask of the object that the user wants to refer to. Finally, the mask of the object is the input of the graffiti channel. In this way, the various forms of graffiti of the user can be converted into pixel-level object masks, which unifies various forms of graffiti and increases the accuracy of the model in identifying prominent areas.

[0048] S104: The graphic and text information processing device inputs the input sequence into a large language model to obtain corresponding graphic and text understanding positioning information.

[0049] Specifically, the mixed feature input sequence of images and text will be input into the efficient backbone network that is dual-adaptive for image and text modalities. Finally, the backbone network will answer various questions from users in a unified generative way. If the question is a comprehension task, the model will output the answer in the form of pure text; if the question includes a positioning task, the model will first output the structured features that refer to the object, and then output the coordinate position of the object's circumscribed polygon.

[0050] Optionally, this application also designs a two-stage training paradigm to ensure the convergence of the model:

[0051] The first stage is the pre-training stage, which is designed to help the model align image and text features. During this stage, both the pre-trained visual encoder and the large language model remain frozen. The scribble convolution layer and the projection layer that maps visual features to text space will be trained to learn visual embeddings that can be understood by the large language model.

[0052] The second stage is the instruction adjustment for multi-image and text scenarios. At this stage, this application aims to solve the limitation that the current existing methods can only cope with single-image and text scenarios, and expand the model capabilities to multi-image and text understanding and positioning scenarios by fine-tuning multi-image and text instructions, so that the model can understand the complex reference relationship between text and images and the complex relationship between multiple images, and can locate the pixel-level position of objects according to the user's multimodal and multi-form reference, and finally obtain a multimodal large language model with a higher intelligence level. At this stage, the image encoder, graffiti convolution layer, projection layer and backbone model will all be trained jointly.

[0053] In an embodiment of the present application, since an indication token can be added to the first image in the first graphic information and an input sequence containing token features can be obtained, the ability of the large language model to distinguish images can be improved, thereby enabling the model to understand the complex referential relationship between text and images, thereby improving the large language model's ability to understand the relationship between text and images.

[0054] The above mainly introduces the solution provided by the embodiment of the present application from the perspective of the method. In order to realize the above functions, it includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiment disclosed herein, the embodiment of the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware 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 exceed the scope of the present application.

[0055] The graphic information processing method provided in the embodiment of the present application can be executed by a graphic information processing device or a control module for graphic information processing in the graphic information processing device. The graphic information processing device provided in the embodiment of the present application is described by taking the graphic information processing device executing the graphic information processing method as an example.

[0056] It should be noted that the embodiment of the present application can divide the functional modules of the graphic information processing device according to the above method example. For example, each functional module can be divided according to each function, or two or more functions can be integrated into one processing module. The above integrated module can be implemented in the form of hardware or in the form of software functional modules. Optionally, the division of modules in the embodiment of the present application is schematic and is only a logical functional division. There may be other division methods in actual implementation.

[0057] like Figure 4 As shown, an embodiment of the present application provides a graphic information processing device 400. The graphic information processing device 400 includes: an acquisition module 401 and a processing module 402. The acquisition module 401 can be used to acquire first graphic information, wherein the first graphic information includes a first text and a first image; the processing module 402 is used to add an indication token to the first image to obtain second graphic information, wherein the indication token includes an image start token and an image end token, wherein the image start token is located before the first image, and the image end token is located after the first image; determine the text features corresponding to the first text, determine the image features corresponding to the first image, determine the token features corresponding to the indication token, and determine an input sequence based on the text features, the image features and the token features; input the input sequence into a large language model to obtain corresponding graphic understanding positioning information.

[0058] Optionally, the first graphic information also includes a second image, and the second image includes at least one image; the processing module 402 is also used to add indication tokens to the at least one image respectively to obtain the second graphic information; wherein, the indication token of the first image corresponds to the first identifier, and the indication token of the second image corresponds to the second identifier.

[0059] Optionally, the first image includes at least two objects, and the at least two objects include a target object with graffiti information added; the processing module 402 is used to segment the first mask feature corresponding to the target object in the first image through a segmentation model, and input the first mask feature into the graffiti channel of the visual encoder to determine the image feature.

[0060] Optionally, the value of the graffiti channel includes a first value and a second value, the first value represents the image foreground, and the second value represents the image background.

[0061] In an embodiment of the present application, since an indication token can be added to the first image in the first graphic information and an input sequence containing token features can be obtained, the ability of the large language model to distinguish images can be improved, thereby enabling the model to understand the complex referential relationship between text and images, thereby improving the large language model's ability to understand the relationship between text and images.

[0062] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530 and a communication bus 540, wherein the processor 510, the communication interface 520 and the memory 530 communicate with each other through the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute the image-text information processing method, which includes: obtaining first image-text information, wherein the first image-text information includes a first text and a first image; adding an indication token to the first image to obtain second image-text information, wherein the indication token includes an image start token and an image end token, wherein the image start token is located before the first image and the image end token is located after the first image; determining the text feature corresponding to the first text, determining the image feature corresponding to the first image, determining the token feature corresponding to the indication token, and determining an input sequence according to the text feature, the image feature and the token feature; inputting the input sequence into a large language model to obtain corresponding image-text understanding positioning information.

[0063] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0064] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the graphic information processing method provided by the above-mentioned methods, which includes: obtaining first graphic information, the first graphic information including a first text and a first image; adding an indication token to the first image to obtain second graphic information, the indication token including an image start token and an image end token, the image start token is located before the first image, and the image end token is located after the first image; determining text features corresponding to the first text, determining image features corresponding to the first image, determining token features corresponding to the indication token, and determining an input sequence based on the text features, the image features and the token features; inputting the input sequence into a large language model to obtain corresponding graphic understanding positioning information.

[0065] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the graphic information processing method provided by the above-mentioned methods, the method comprising: obtaining first graphic information, the first graphic information comprising a first text and a first image; adding an indication token to the first image to obtain second graphic information, the indication token comprising an image start token and an image end token, the image start token being located before the first image, and the image end token being located after the first image; determining text features corresponding to the first text, determining image features corresponding to the first image, determining token features corresponding to the indication token, and determining an input sequence based on the text features, the image features and the token features; inputting the input sequence into a large language model to obtain corresponding graphic understanding positioning information.

[0066] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0067] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for processing graphic information, characterized in that: include: Acquire first graphic information, where the first graphic information includes a first text and a first image; Adding an indication token to the first image to obtain second image-text information, wherein the indication token includes an image start token and an image end token, wherein the image start token is located before the first image, and the image end token is located after the first image; Determine text features corresponding to the first text in the second graphic information, determine image features corresponding to the first image in the second graphic information, determine token features corresponding to the indication token in the second graphic information, and determine an input sequence according to the text features, the image features, and the token features; Inputting the input sequence into a large language model to obtain corresponding image-text understanding positioning information; The first graphic information further includes a second image, and the second image includes at least one image; the method further includes: adding an indication token to each of the at least one image to obtain the second graphic information; The indication token of the first image corresponds to a first identifier, and the indication token of the second image corresponds to a second identifier; In the second graphic information, the text reference of the first image is included before the image start token; The step of determining the input sequence according to the text feature, the image feature and the token feature comprises: The text features, the image features, and the token features are reordered according to the text reference of the first image to determine the input sequence.

2. The method for processing graphic information according to claim 1, characterized in that: The first image includes at least two objects, and the at least two objects include a target object to which graffiti information is added; and determining the image feature corresponding to the first image includes: The first mask feature corresponding to the target object in the first image is segmented by a segmentation model, and the first mask feature is input into a scribble channel of a visual encoder to determine the image feature.

3. The method for processing graphic information according to claim 2, characterized in that: The value of the graffiti channel includes a first value and a second value, the first value represents the image foreground, and the second value represents the image background.

4. A graphic information processing device, characterized in that: include: Acquisition module and processing module; The acquisition module is used to acquire first graphic information, where the first graphic information includes a first text and a first image; The processing module is used to add an indication token to the first image to obtain second image-text information, wherein the indication token includes an image start token and an image end token, wherein the image start token is located before the first image, and the image end token is located after the first image; Determine text features corresponding to the first text in the second graphic information, determine image features corresponding to the first image in the second graphic information, determine token features corresponding to the indicator token in the second graphic information, and determine an input sequence based on the text features, the image features, and the token features; input the input sequence into a large language model to obtain corresponding graphic understanding positioning information; The first graphic information also includes a second image, and the second image includes at least one image; the processing module is further used to add an indication token to each of the at least one image to obtain the second graphic information; wherein the indication token of the first image corresponds to the first identifier, and the indication token of the second image corresponds to the second identifier; In the second graphic information, the text reference of the first image is included before the image start token; The processing module is used to reorder the text features, the image features and the token features according to the text reference of the first image to determine the input sequence.

5. The graphic information processing device according to claim 4, characterized in that: The first image includes at least two objects, and the at least two objects include a target object with graffiti information added; the processing module is used to segment the first mask feature corresponding to the target object in the first image through a segmentation model, and input the first mask feature into the graffiti channel of the visual encoder to determine the image feature.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the graphic information processing method according to any one of claims 1 to 3 is implemented.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for processing graphic information as claimed in any one of claims 1 to 3 is implemented.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for processing graphic information as claimed in any one of claims 1 to 3 is implemented.

Images