Input method skin generation method and device, equipment, storage medium and program product
Through the image descriptive text and stylistic word text input by the user, the input method skin is generated using the literary and graphic model, which solves the problems of high cost, low efficiency and low user experience in the prior art, and realizes efficient and personalized input method skin generation.
Patent Information
- Application Number
- CN202411901874.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-23
- Publication Date
- 2025-04-18
AI Technical Summary
In the prior art, the input method has high skin design cost and low design efficiency, and cannot meet the personalized needs of all users, resulting in a reduced user experience.
By obtaining the image descriptive text and stylistic word text input by the user, the literary and biographical image model is used to generate the input method background image and function key image that meets the user's needs, thereby automatically generating the input method skin.
It realizes efficient and personalized skin generation of input method, reduces design costs and time, and improves user experience.
Smart Images
Figure CN119850765A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an input method skin generation method, device, equipment, storage medium and program product. Background Art
[0002] With the rapid development of technology, there are more and more user devices; among them, input method is one of the indispensable tools in user devices. Input method is responsible for converting keyboard input or touch screen input of user devices into characters and text. As people's requirements for personalized experience of input method continue to rise, users are no longer satisfied with the unified input method skin, and hope to show their personality and improve the user experience through customized skin.
[0003] At present, many input method skins of various styles are designed in advance for users to choose their favorite input method skins. However, the existing technology requires input method skin designers to manually design a variety of input method skins, which has high design costs and low design efficiency; and the input method skins designed in advance are limited and cannot meet the needs of all users, resulting in a reduced user experience. Summary of the invention
[0004] The present invention provides an input method skin generation method, device, equipment, storage medium and program product, which are used to solve the defects of high input method skin design cost and low design efficiency, as well as low user experience in the prior art, and realize efficient and high-experience input method skin generation.
[0005] The present invention provides an input method skin generation method, comprising: Acquire input text; the input text includes image descriptive text input by the user, and the image descriptive text includes image content information of the input method background image; Inputting the input text into a first text-generated image model to obtain an input method background image output by the first text-generated image model; the input method background image is used to generate an input method skin; The first text-generated graph model is used to generate a corresponding input method background graph based on content information of the input text.
[0006] According to an input method skin generation method provided by the present invention, the input text further includes a style word text, and the style word text is used to indicate the style of the input method background image.
[0007] According to a method for generating an input method skin provided by the present invention, the step of inputting the input text into a first text-generated image model to obtain an input method background image output by the first text-generated image model comprises: Inputting the image descriptive text into a text polishing model to obtain a polished text output by the text polishing model; Inputting the polished text into a first text-generated image model to obtain an input method background image output by the first text-generated image model; The text polishing model is a pre-trained large language model.
[0008] According to a method for generating an input method skin provided by the present invention, after obtaining the input text, the method further includes: Inputting the input text into an icon descriptive text generation model to obtain an icon descriptive text output by the icon descriptive text generation model; the icon descriptive text includes image content information of the function key icon; Inputting the icon descriptive text into a second text-generated image model to obtain a function key image output by the second text-generated image model; the function key image includes a function key icon, and the input method background image and the function key icon are used together to generate an input method skin; Among them, the icon descriptive text generation model is a pre-trained large language model; the second text generation model is used to generate corresponding function key images based on the content information of the icon descriptive text.
[0009] According to a method for generating an input method skin provided by the present invention, after inputting the icon descriptive text into a second text-generated image model to obtain a function key image output by the second text-generated image model, the method further includes: Inputting the function key image into a foreground segmentation model to obtain a function key icon output by the foreground segmentation model; The foreground segmentation model is used to segment and obtain foreground information of the function key image.
[0010] According to an input method skin generation method provided by the present invention, the icon descriptive text includes a plurality of icon descriptive sub-texts, any of the icon descriptive sub-texts includes image content information of a function key icon; the number of the function key images is multiple; The step of inputting the function key image into a foreground segmentation model to obtain a function key icon output by the foreground segmentation model includes: For any of the function key images, input the function key image into a foreground segmentation model to obtain a foreground icon output by the foreground segmentation model; Inputting the foreground icon into an image quality assessment model to obtain an image quality assessment result output by the image quality assessment model; Based on the image quality evaluation result, determining whether to use the foreground icon as a function key icon; Obtaining a number of said function key icons; Wherein, the image quality assessment model is obtained by training based on sample foreground icons.
[0011] According to an input method skin generation method provided by the present invention, the input text further includes a style word text, and the style word text is used to indicate the style of the input method background image; The step of inputting the icon descriptive text into a second text-generated image model to obtain a function key image output by the second text-generated image model includes: The icon descriptive text and the style word text are input into a second text-generated image model to obtain a function key image output by the second text-generated image model.
[0012] According to a method for generating an input method skin provided by the present invention, after inputting the input text into a first text-generated image model to obtain an input method background image output by the first text-generated image model, the method further comprises: Obtaining input method skin configuration information corresponding to a target device; the target device is a device for which an input method skin needs to be generated; Based on the input method skin configuration information, a generation method of the input method skin is determined.
[0013] The present invention also provides an input method skin generating device, comprising: An acquisition module, used to acquire input text; the input text includes image descriptive text input by a user, and the image descriptive text includes image content information of an input method background image; An input module, used for inputting the input text into a first text-generated image model to obtain an input method background image output by the first text-generated image model; the input method background image is used to generate an input method skin; The first text-generated graph model is used to generate a corresponding input method background graph based on content information of the input text.
[0014] 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 one of the input method skin generation methods described above is implemented.
[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the input method skin generation method as described in any one of the above is implemented.
[0016] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the input method skin generation method as described above is implemented.
[0017] The input method skin generation method, device, equipment, storage medium and program product provided by the present invention obtain input text, and the input text includes image descriptive text input by the user, and the image descriptive text includes image content information of the input method background image, so that the input text is input into the first text-generated image model to obtain the input method background image output by the first text-generated image model, and the input method background image is used to generate the input method skin, so that the user can input the image descriptive text according to his own needs, and generate an input method skin that meets the user's personalized needs, and can meet the personalized needs of all users, thereby improving the user experience; and the first text-generated image model is used to generate the corresponding input method background image based on the content information of the input text, thereby eliminating the need to manually design a variety of input method skins in advance, reducing the design cost of the input method skin, and through the above method, only text needs to be input to quickly generate the input method skin, thereby improving the generation efficiency of the input method skin and avoiding the problem of low design efficiency of the input method skin. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] 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.
[0019] Figure 1 This is one of the flow charts of the input method skin generation method provided by the present invention.
[0020] Figure 2 This is the second flow chart of the input method skin generation method provided by the present invention.
[0021] Figure 3 It is a schematic diagram of image segmentation of the foreground segmentation model provided by the present invention.
[0022] Figure 4 It is a display schematic diagram of the input method skin provided by the present invention.
[0023] Figure 5 It is a structural schematic diagram of the input method skin generating device provided by the present invention.
[0024] Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0025] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings 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.
[0026] With the rapid development of science and technology, there are more and more user devices; among them, input method is one of the indispensable tools in user devices. Input method is responsible for converting keyboard input or touch screen input of user devices into characters and text. With the popularization of smart devices and the increasing demand for input methods, input method is not only a simple input tool, but also gradually becomes an important part of user interface aesthetics and human-computer interaction experience. Based on this, it is necessary to generate input method skin. Among them, input method skin belongs to the intersection of human-computer interaction and user interface design.
[0027] As people's requirements for personalized experience of input methods continue to rise, users are no longer satisfied with the unified input method skins, and hope to express their personality and improve the user experience through customized skins. Therefore, it is necessary to design and develop input method skins to meet the diverse needs of users.
[0028] At present, many input method skins of various styles are designed in advance for users to choose their favorite input method skins. However, the existing technology requires input method skin designers to manually design a variety of input method skins, which has high design costs and low design efficiency; and the input method skins designed in advance are limited and cannot meet the needs of all users, resulting in a reduced user experience.
[0029] Specifically, the existing input method skin generation mainly relies on manual design. The design process revolves around the creation and layout of visual elements, and its principle is to improve user experience through visual design. First, designers use tools such as Adobe Photoshop and Illustrator to create key visual elements such as function keys, backgrounds, fonts, etc. These elements must take into account both aesthetics and functionality. For example, the design of function keys must have both artistic beauty and ensure the accuracy and comfort of user clicks. Based on this, although the existing input method skin design scheme has made great progress in aesthetics and personalization, it still has some shortcomings and disadvantages; first, the design and development costs are high. Creating high-quality input method skins requires professional designers to use complex graphic tools for fine design, which not only increases the development cost, but also may extend the development cycle, that is, the manual design cost is high and the design efficiency is low; second, designers need to design a variety of styles of input method skins in advance so that users can choose their favorite input method skins from them, but the number of such preset input method skins is limited, and it is difficult to fully meet the needs of all users, resulting in a reduced user experience.
[0030] In view of the above problems, the present invention proposes the following embodiments. Figure 1-Figure 4 The input method skin generation method of the present invention is described.
[0031] Figure 1 is one of the flow charts of the input method skin generation method provided by the present invention, such as Figure 1 As shown, the input method skin generation method includes the following steps 110 and 120.
[0032] Step 110, obtaining input text.
[0033] In the embodiment of the present invention, the execution subject of the input method skin generation method can be the cloud (server) or the user device for which the input method skin needs to be generated. If the execution subject is the cloud, the user device sends the input text to the cloud, and finally the cloud sends the generated input method background image to the user device, so that the user device can generate the corresponding input method skin.
[0034] The input text includes image descriptive text input by a user, and the image descriptive text includes image content information of an input method background image.
[0035] Here, the image descriptive text is a descriptive language in the form of text input by the user, which is used to describe the input method background image to be generated. The image content information is used to characterize the image content of the input method background image, so that the corresponding input method background image can be generated based on the image content information. For example, the image descriptive text is "a young girl and her pet dog, walking on the grassland".
[0036] If the execution subject of the input method skin generation method is the cloud, the image descriptive text is obtained by the user inputting on the user device; if the execution subject of the input method skin generation method is the user device, the image descriptive text is obtained by the user inputting on the user device.
[0037] Furthermore, before the user inputs the image descriptive text, a text prompt may be provided, for example, the prompt text is “Please describe the content of the input method background image you want to generate”.
[0038] Step 120: input the input text into a first text-generated image model to obtain an input method background image output by the first text-generated image model.
[0039] The first text-generated graph model is used to generate a corresponding input method background graph based on content information of the input text.
[0040] In one embodiment, the first text-generated graph model is trained based on sample input text. Further, the first text-generated graph model is trained based on sample input text and its corresponding input method background graph label.
[0041] In another embodiment, the first text-generated image model is a pre-trained large model. Further, the large model can be fine-tuned based on the sample input text to obtain a first text-generated image model that is more suitable for the input method skin generation scenario, thereby obtaining a more accurate input method background image; that is, the fine-tuning technology makes the pre-trained large model suitable for input method skin generation, greatly increasing the controllability of the input method skin generation result.
[0042] The input method background image is used to generate the input method skin. The input method background image is the main background of the input method skin.
[0043] If the executor of the input method skin generation method is the cloud, the input method background image is sent to the user device so that the user device can generate the input method skin based on the input method background image; if the executor of the input method skin generation method is the user device, the user device generates the input method skin based on the input method background image.
[0044] The input method skin generation method provided in the embodiment of the present invention obtains input text, and the input text includes image descriptive text input by the user, and the image descriptive text includes image content information of the input method background image, so that the input text is input into the first text-generated image model to obtain the input method background image output by the first text-generated image model, and the input method background image is used to generate the input method skin, so that the user can input the image descriptive text according to his own needs, and generate an input method skin that meets the user's personalized needs, and can meet the personalized needs of all users, thereby improving the user experience; and the first text-generated image model is used to generate the corresponding input method background image based on the content information of the input text, thereby eliminating the need to manually design a variety of input method skins in advance, reducing the design cost of the input method skin, and through the above method, only text needs to be input to quickly generate the input method skin, thereby improving the generation efficiency of the input method skin, and avoiding the problem of low design efficiency of the input method skin.
[0045] Based on any of the above embodiments, in the method, the input text further includes a style word text, and the style word text is used to indicate the style of the input method background image.
[0046] Here, the style word text can be a descriptive language in the form of text input by the user; the style word text can also be a style word selected by the user, that is, multiple style words are provided in advance for the user to choose. The style word text is used to describe the style (style) of the input method background image to be generated. For example, the style word text is "oil painting style", "anime style", "cute cartoon style" or "vivid realistic style".
[0047] If the execution body of the input method skin generation method is the cloud, the style word text is obtained by the user inputting or selecting on the user device; if the execution body of the input method skin generation method is the user device, the style word text is obtained by the user inputting or selecting on the user device.
[0048] Accordingly, the above step 120 includes: inputting the image descriptive text and the style word text into the first text-generated image model to obtain the input method background image output by the first text-generated image model.
[0049] In one embodiment, the image descriptive text and the style word text may be merged into a merged text, and then the merged text may be input into the first text-generated image model to obtain an input method background image output by the first text-generated image model. For example, if the image descriptive text is "a young girl and her pet dog, strolling on the grassland", and the style word text is "cute cartoon style", then the merged text is "a young girl and her pet dog, strolling on the grassland, cute cartoon style".
[0050] In another embodiment, the image descriptive text and the style word text are used as two inputs of the first context graph model.
[0051] The input method skin generation method provided by the embodiment of the present invention, the input text also includes style word text, and the style word text is used to indicate the style of the input method background image, so that the user can input the style word text according to his own needs to generate an input method skin that better meets the user's personalized needs, and can meet the personalized needs of all users, thereby improving the user experience.
[0052] Based on any of the above embodiments, in the method, the above step 120 includes: Inputting the image descriptive text into a text polishing model to obtain a polished text output by the text polishing model; The polished text is input into a first text-generated image model to obtain an input method background image output by the first text-generated image model.
[0053] In some embodiments, the text polishing model can be used to rewrite, expand, abbreviate, continue, and imitate the image descriptive text. In one embodiment, the text polishing model can determine which text polishing process to perform. In another embodiment, the user can determine which text polishing process to perform.
[0054] Considering that the image descriptive text is input by the user, but the user's input text is not necessarily standardized, for example, there are typos, missing words, incoherence, less content, more content, overly colloquial language and other problems. Based on this, the image descriptive text is input into the text polishing model to obtain a polished text that better meets the input requirements of the first text-to-image model.
[0055] It should be noted that, considering that the image descriptive text is more likely to be non-standard, while the style word text is usually selected by the user with pre-set style words, based on this, the style word text is less likely to be non-standard, so the style word text does not need to be polished. In addition, since the content to be described in the image descriptive text is relatively unfixed and complex, while the content to be described in the style word text is relatively fixed and basically has some style types, therefore, only the image descriptive text needs to be polished.
[0056] Of course, the style word text can also be polished according to actual needs to obtain a polished style word text. For a specific implementation, reference can be made to the text polishing method of the image descriptive text.
[0057] The text polishing model is a pre-trained large language model. Furthermore, the large language model can be fine-tuned based on the sample image descriptive text to obtain a text polishing model that is more suitable for the input method skin generation scenario, thereby obtaining a polished text that better meets the input requirements of the first text-generated image model; that is, the fine-tuning technology makes the pre-trained large language model suitable for input method skin generation, greatly increasing the controllability of the input method skin generation results.
[0058] The input method skin generation method provided by the embodiment of the present invention takes into account that the image descriptive text is input by the user, and the user's input text is not necessarily standardized. The image descriptive text is input into the text polishing model to obtain the polished text that better meets the input requirements of the first text image model, and then the polished text is input into the first text image model to obtain a more accurate input method background image output by the first text image model, thereby ensuring that the generated input method skin is more in line with the user's personalized needs, thereby improving the user experience.
[0059] Based on any of the above embodiments, Figure 2 FIG. 2 is a flow chart of the input method skin generation method provided by the present invention. Figure 2 As shown, after the above step 110, the method further includes: step 130 and step 140.
[0060] Step 130: input the input text into an icon descriptive text generation model to obtain an icon descriptive text output by the icon descriptive text generation model.
[0061] Among them, the icon descriptive text generation model is a pre-trained large language model. Furthermore, the large language model can be fine-tuned based on the sample input text to obtain an icon descriptive text generation model that is more suitable for the input method skin generation scenario, thereby obtaining more accurate icon descriptive text, thereby improving the generation accuracy of the input method skin, and further improving the user experience; that is, the fine-tuning technology makes the pre-trained large language model suitable for input method skin generation, greatly increasing the controllability of the input method skin generation results.
[0062] The icon descriptive text includes image content information of the function key icon. It is used to describe the function key icon to be generated. The image content information is used to characterize the image content information of the function key icon, so that the corresponding function key image can be generated based on the image content information. For example, the icon descriptive text is "a flower", "a clover", "a rabbit", "a butterfly", "a stone", "a bird", "a cloud", "a kite", "a bone", "a tent", "a guitar", "a hat", "a pair of boots", "a bee", "a stream" or "a hot air balloon".
[0063] In one embodiment, if the input text also includes a style word text, the image descriptive text and the style word text can be merged into a merged text, and then the merged text is input into the icon descriptive text generation model to obtain the icon descriptive text output by the icon descriptive text generation model. For example, if the image descriptive text is "a young girl and her pet dog, walking on the grassland", and the style word text is "cute cartoon style", then the merged text is "a young girl and her pet dog, walking on the grassland, cute cartoon style".
[0064] In another embodiment, if the input text also includes style word text, the image descriptive text and the style word text are used as two inputs of the icon descriptive text generation model.
[0065] In some embodiments, the icon descriptive text includes multiple icon descriptive subtexts (prompt words), and any icon descriptive subtext includes image content information of a function key icon. Accordingly, the number of function key images is multiple. Further, after obtaining multiple function key icons, the multiple function key icons can be randomly mapped to each function key of the input method, or mapped according to preset rules, so as to facilitate the subsequent generation of input method skins.
[0066] Exemplarily, the image descriptive text is “a young girl and her pet puppy, walking on the grassland”, the style word text is “cute cartoon style”, and the icon descriptive text includes the following icon descriptive sub-texts: “a flower”, “a clover”, “a rabbit”, “a butterfly”, “a stone”, “a bird”, “a cloud”, “a kite”, “a bone”, “a tent”, “a guitar”, “a hat”, “a pair of boots”, “a bee”, “a stream” or “a hot air balloon”.
[0067] It should be understood that in order to maintain the consistency of the overall style of the input method skin, the function key icon must be associated with the input method background image. Based on this, the input text for generating the input method background image is also input into the icon descriptive text generation model, thereby outputting icon descriptive text that is highly related to the input method background image.
[0068] Step 140: input the icon descriptive text into a second text-generated image model to obtain a function key image output by the second text-generated image model.
[0069] The second text-generated image model is used to generate a corresponding function key image based on the content information of the icon descriptive text.
[0070] In one embodiment, the second text-generated image model is trained based on the sample icon descriptive text. Further, the second text-generated image model is trained based on the sample icon descriptive text and its corresponding function key image label.
[0071] In another embodiment, the second text-generated image model is a pre-trained large model. Further, the large model can be fine-tuned based on the sample icon descriptive text to obtain a second text-generated image model that is more suitable for the input method skin generation scenario, thereby obtaining a more accurate function key image; that is, the fine-tuning technology makes the pre-trained large model suitable for input method skin generation, greatly increasing the controllability of the input method skin generation result.
[0072] In another embodiment, the second text-generated graph model and the first text-generated graph model may be the same model. If training is required, model training is performed based on a mixture of training samples of the two.
[0073] The function key image includes a function key icon, and the input method background image and the function key icon are used together to generate an input method skin.
[0074] If the execution subject of the input method skin generation method is the cloud, the input method background image and the function key icons are sent to the user device, so that the user device generates the input method skin based on the input method background image and the function key icons; if the execution subject of the input method skin generation method is the user device, the user device generates the input method skin based on the input method background image and the function key icons. Specifically, based on the input method background image that has been generated, the function key icons are superimposed on the input method background image according to the arrangement of the input method function keys, so as to obtain a complete input method skin.
[0075] In a specific embodiment, the icon descriptive text includes a plurality of icon descriptive subtexts, and the plurality of icon descriptive subtexts are respectively input into the second text-generated image model to obtain a plurality of function key images output by the second text-generated image model, that is, one icon descriptive subtext is used to generate one function key image.
[0076] The input method skin generation method provided by the embodiment of the present invention inputs input text into an icon descriptive text generation model to obtain an icon descriptive text output by the icon descriptive text generation model, inputs the icon descriptive text into a second text-generated image model to obtain a function key image output by the second text-generated image model, and the input text includes an image descriptive text input by a user, the image descriptive text includes image content information of an input method background image, and the function key image includes a function key icon, the input method background image and the function key icon are used together to generate an input method skin, so that the user can input the image descriptive text according to his or her own needs, and generate an input method skin that meets the user's personalized needs, and can meet the personalized needs of all users, thereby improving the user experience; and the second text-generated image model It is used to generate corresponding function key images based on the content information of the icon descriptive text, thereby eliminating the need to manually design multiple input method skins in advance, reducing the design cost of the input method skin, and through the above method, only text needs to be entered to quickly generate the input method skin, thereby improving the generation efficiency of the input method skin and avoiding the problem of low design efficiency of the input method skin; at the same time, the input text for generating the input method background image is also input into the icon descriptive text generation model, thereby outputting icon descriptive text that is highly related to the input method background image, thereby ensuring that the function key icon and the input method background image are associated, that is, ensuring the consistency of the overall style of the input method skin, thereby improving the rationality of the generated input method skin, ensuring that the input method skin is beautiful and in line with design, and ultimately improving the user experience.
[0077] Based on any of the above embodiments, considering that the function key image is the original image output by the second text image model, that is, the function key image includes not only the required foreground elements (function key icons) but also some screen backgrounds; based on this, after the above step 140, the method further includes: The function key image is input into a foreground segmentation model to obtain a function key icon output by the foreground segmentation model.
[0078] The foreground segmentation model is used to segment and obtain the foreground information of the function key image, that is, the foreground segmentation model is used to segment and obtain the foreground icon of the function key image.
[0079] In one embodiment, the foreground segmentation model is trained based on sample function key images. Further, the foreground segmentation model is trained based on sample function key images and their corresponding function key icon labels. Based on this, a foreground segmentation model that is more suitable for the input method skin generation scenario can be obtained, thereby obtaining a more accurate function key icon.
[0080] For example, assuming that the icon descriptive text is "a flower basket", such as Figure 3 As shown, Figure 3The left side is the function key image (original image) output by the second text image model. The function key image is input into the foreground segmentation model to obtain the function key icon output by the foreground segmentation model ( Figure 3 image on the right).
[0081] The input method skin generation method provided in an embodiment of the present invention takes into account that the function key image is the original image output by the second text image model, that is, the function key image includes not only the required foreground elements but also some screen backgrounds. Therefore, the function key image is input into the foreground segmentation model to obtain the function key icon output by the foreground segmentation model, thereby ensuring that the input method skin is generated based on the input method background image and the function key icon, thereby improving the rationality of the generated input method skin, ensuring that the input method skin is beautiful and conforms to design, and ultimately improving the user experience.
[0082] Based on any of the above embodiments, in the method, the icon descriptive text includes a plurality of icon descriptive sub-texts, any of the icon descriptive sub-texts includes image content information of a function key icon; and the number of the function key images is multiple.
[0083] Exemplarily, the image descriptive text is “a young girl and her pet puppy, walking on the grassland”, the style word text is “cute cartoon style”, and the icon descriptive text includes the following icon descriptive sub-texts: “a flower”, “a clover”, “a rabbit”, “a butterfly”, “a stone”, “a bird”, “a cloud”, “a kite”, “a bone”, “a tent”, “a guitar”, “a hat”, “a pair of boots”, “a bee”, “a stream” and “a hot air balloon”.
[0084] Considering that the foreground segmentation model is not stable enough, that is, the function key icon output by the foreground segmentation model may have low quality, based on this, in the method, the function key image is input to the foreground segmentation model to obtain the function key icon output by the foreground segmentation model, including: For any of the function key images, input the function key image into a foreground segmentation model to obtain a foreground icon output by the foreground segmentation model; Inputting the foreground icon into an image quality assessment model to obtain an image quality assessment result output by the image quality assessment model; Based on the image quality evaluation result, determining whether to use the foreground icon as a function key icon; Get several function key icons.
[0085] Here, the foreground icon is also a function key icon, but here only qualified foreground images are used as function key icons.
[0086] The image quality assessment model is obtained by training based on sample foreground icons. Further, the image quality assessment model is obtained by training based on sample foreground icons and their corresponding image quality assessment result labels.
[0087] In some embodiments, the image quality evaluation model is a binary classification model, so that it can be directly determined whether the foreground icon can be used as a function key icon based on the image quality evaluation result. That is, the image quality evaluation model is used to score the foreground icon, so that the foreground icon with a qualified score can be used as the function key icon.
[0088] In other embodiments, the image quality evaluation model is a regression model, so that the image quality evaluation result can be a specific score. In one embodiment, foreground icons with scores greater than a preset score threshold are used as function key icons, otherwise they are not used as function key icons. In another embodiment, based on the image quality evaluation results of all foreground icons, several foreground icons are selected as function key icons in descending order of scores.
[0089] It should be noted that the number of multiple icon descriptive subtexts is greater than the number of function key icons to be generated, that is, the number of foreground icons is greater than the number of function key icons to be generated. Based on this, function key icons with better quality can be screened out from more foreground icons.
[0090] Exemplarily, the icon descriptive text includes the following icon descriptive sub-texts: "a flower", "a clover", "a rabbit", "a butterfly", "a stone", "a bird", "a cloud", "a kite", "a bone", "a tent", "a guitar", "a hat", "a pair of boots", "a bee", "a stream" and "a hot air balloon", that is, the number of the multiple icon descriptive sub-texts is 16, and the number of function key icons to be generated is 8. Based on this, 8 can be selected from the 16 foreground icons as function key icons.
[0091] Furthermore, after obtaining a plurality of function key icons, the plurality of function key icons can be randomly mapped to the function keys of the input method, or mapped according to a preset rule, so as to facilitate the subsequent generation of an input method skin.
[0092] The input method skin generation method provided by the embodiment of the present invention takes into account the insufficient stability of the foreground segmentation model, that is, the function key icon output by the foreground segmentation model may have low quality. Therefore, for any function key image, the function key image is input into the foreground segmentation model to obtain the foreground icon output by the foreground segmentation model, and the foreground icon is input into the image quality evaluation model to obtain the image quality evaluation result output by the image quality evaluation model. Based on the image quality evaluation result, it is determined whether to use the foreground icon as a function key icon, thereby obtaining a number of function key icons with higher quality, that is, screening out foreground icons with complete structure and more aesthetics as function key icons, thereby improving the quality of the generated input method skin, ensuring that the input method skin is beautiful and conforms to design, and ultimately improving the user experience.
[0093] Based on any of the above embodiments, in the method, the input text further includes a style word text, and the style word text is used to indicate the style of the input method background image. Accordingly, the above step 140 includes: The icon descriptive text and the style word text are input into a second text-generated image model to obtain a function key image output by the second text-generated image model.
[0094] In one embodiment, the image descriptive text and the style word text may be merged into a merged text, and then the merged text may be input into the second text-generated image model to obtain a function key image output by the second text-generated image model. For example, if the image descriptive text is "a young girl and her pet dog, strolling on the grassland", and the style word text is "cute cartoon style", then the merged text is "a young girl and her pet dog, strolling on the grassland, cute cartoon style".
[0095] In another embodiment, the image descriptive text and the style word text are used as two inputs of the second context graph model.
[0096] In some embodiments, the icon descriptive text includes a plurality of icon descriptive subtexts, and for any icon descriptive subtext, the icon descriptive subtext and the style word text are input into the second text-generated image model to obtain a function key image output by the second text-generated image model. That is, an icon descriptive subtext and a style word text are used to generate a function key image.
[0097] The input method skin generation method provided by the embodiment of the present invention inputs the icon descriptive text and the style word text into the second text-generated image model to obtain the function key image output by the second text-generated image model, that is, the style word text used to indicate the input method background image is also input into the second text-generated image model, ensuring that the output function key image is highly related to the input method background image, that is, further ensuring the consistency of the overall style of the input method skin, thereby further improving the rationality of the generated input method skin, ensuring that the input method skin is beautiful and conforms to design, and ultimately further enhancing the user experience.
[0098] Based on any of the above embodiments, after the above step 140, the method further includes: The input method background image is adjusted based on the function key icons and the position information of the function keys on the input method background image.
[0099] Here, the position information of the function keys on the input method background image is known. Considering that in the process of generating the input method skin, the function key icons are superimposed on the input method background image according to the arrangement of the input method function keys based on the already generated input method background image to obtain a complete input method skin, based on this, it is necessary to adjust the input method background image to adjust the pixels of the position of the function keys on the input method background image, so that the function key icons can be better rendered at the position of the function keys on the input method background image. That is, the input method background image still needs to be post-processed before it can be used as the background image of the input method skin. Of course, if there are multiple function key icons, the processing method for each function key icon is the same.
[0100] The input method skin generation method provided in an embodiment of the present invention adjusts the input method background image based on the function key icons and the position information of the function keys on the input method background image, thereby obtaining a higher quality input method background image, which facilitates better rendering of the function key icons at the positions of the function keys on the input method background image, thereby improving the quality of the generated input method skin, ensuring that the input method skin is beautiful and conforms to design, and ultimately improving the user experience.
[0101] Based on any of the above embodiments, considering that in the prior art, after designing the input method skin, responsive design and multi-resolution adaptation are also required, that is, to ensure that the input method skin performs consistently on different devices, designers usually use vector graphics or high-resolution bitmaps to maintain the clarity and details of the graphics on various screen sizes. However, when it is necessary to adapt to user devices with multiple resolutions, the workload will increase significantly, that is, the design cost is high and the design efficiency is low; based on this, after the above step 120, the method further includes: Obtaining input method skin configuration information corresponding to a target device; the target device is a device for which an input method skin needs to be generated; Based on the input method skin configuration information, a generation method of the input method skin is determined.
[0102] Here, the input method skin configuration information may include but is not limited to at least one of the following: resolution, layout information of each element, size information of each element, position information of each element, transparency of each element, and setting parameters of interaction effects, etc. That is, the layout, size, position, transparency and other parameters of each element are defined through the input method skin configuration information. That is, the input method skin configuration information guides the rendering effect of the input method skin on the target device to ensure the consistency and aesthetics of the interface on the target device. Among them, the setting of the interaction effect is such as the dynamic change when the function key is pressed, so as to enhance the user's interaction experience.
[0103] If the executor of the input method skin generation method is the cloud, the input method skin configuration information can be sent to the cloud by the target device, or the cloud can automatically obtain the corresponding input method skin configuration information based on the identification information of the target device; if the executor of the input method skin generation method is the user device (target device), the input method skin configuration information can be stored in the target device in advance, or it can be obtained from the cloud when needed.
[0104] Here, the input method skin generation method is used to guide how to generate an input method skin suitable for a target device based on an input method background image. Further, the input method skin generation method is used to guide how to generate an input method skin suitable for a target device based on an input method background image and function key icons.
[0105] The input method skin generation method provided in the embodiment of the present invention obtains input method skin configuration information corresponding to the target device, and the target device is a device for which an input method skin needs to be generated. Based on the input method skin configuration information, the generation method of the input method skin is determined, thereby eliminating the need to manually design input method skins for multiple devices in advance, reducing the design cost of the input method skin, and through the above method, only text needs to be input to quickly generate the input method skin, thereby improving the generation efficiency of the input method skin and avoiding the problem of low design efficiency of the input method skin.
[0106] To facilitate understanding of the above embodiments, Figure 4 As shown in FIG. 1 , for example, the image descriptive text is “a young girl and her pet dog walking on the grassland”, and the style word text is “cute cartoon style”. Based on this, the image descriptive text and the style word text are input into the first text image model, and the following is obtained: Figure 4The large background (input method background image) shown in FIG. 1 is inputted into the icon descriptive text generation model to obtain the icon descriptive text outputted by the icon descriptive text generation model. The icon descriptive text includes a plurality of icon descriptive sub-texts. Any icon descriptive sub-text includes the image content information of a function key icon. Thus, the plurality of icon descriptive sub-texts are respectively inputted into the second text generation model to obtain a plurality of function key images outputted by the second text generation model, thereby generating the following image: Figure 4 The function key icons of the function keys shown. Figure 4 The 26-key input method is shown, and Figure 4 There are 8 function keys.
[0107] Based on the above embodiments, the present invention proposes a customized input method skin generation solution based on text control. Users can directly generate the required input method skin according to their own needs, thereby omitting the tedious steps of manual design and development, thereby greatly saving the generation cost of the input method skin. Specifically, the user only needs to input text, and with the help of a large language model and a text-generated graph model, the relevant input method skin can be quickly generated. This solution can not only meet the personalized needs of users, but also quickly generate input method skins.
[0108] The input method skin generating device provided by the present invention is described below. The input method skin generating device described below and the input method skin generating method described above can be referenced to each other.
[0109] Figure 5 is a schematic diagram of the structure of the input method skin generating device provided by the present invention, such as Figure 5 As shown, the input method skin generating device includes an acquisition module 510 and an input module 520 .
[0110] The acquisition module 510 is used to acquire input text; the input text includes image descriptive text input by the user, and the image descriptive text includes image content information of the input method background image.
[0111] The input module 520 is used to input the input text into the first text-generated image model to obtain an input method background image output by the first text-generated image model; the input method background image is used to generate an input method skin.
[0112] The first text-generated graph model is used to generate a corresponding input method background graph based on content information of the input text.
[0113] The input method skin generation device provided by the embodiment of the present invention obtains input text, and the input text includes image descriptive text input by the user, and the image descriptive text includes image content information of the input method background image, so that the input text is input into the first text-generated image model to obtain the input method background image output by the first text-generated image model, and the input method background image is used to generate the input method skin, so that the user can input the image descriptive text according to his own needs, and generate an input method skin that meets the user's personalized needs, and can meet the personalized needs of all users, thereby improving the user experience; and the first text-generated image model is used to generate the corresponding input method background image based on the content information of the input text, thereby eliminating the need to manually design a variety of input method skins in advance, reducing the design cost of the input method skin, and through the above method, only text needs to be input to quickly generate the input method skin, thereby improving the generation efficiency of the input method skin, and avoiding the problem of low design efficiency of the input method skin.
[0114] Based on any of the above embodiments, the input text further includes style word text, and the style word text is used to indicate the style of the input method background image.
[0115] Based on any of the above embodiments, the input module 520 is specifically used for: Inputting the image descriptive text into a text polishing model to obtain a polished text output by the text polishing model; Inputting the polished text into a first text-generated image model to obtain an input method background image output by the first text-generated image model; The text polishing model is a pre-trained large language model.
[0116] Based on any of the above embodiments, the device further includes: A text output module, used for inputting the input text into the icon descriptive text generation model to obtain the icon descriptive text output by the icon descriptive text generation model; the icon descriptive text includes image content information of the function key icon; An image output module, used for inputting the icon descriptive text into a second text-generated image model to obtain a function key image output by the second text-generated image model; the function key image includes a function key icon, and the input method background image and the function key icon are used together to generate an input method skin; Among them, the icon descriptive text generation model is a pre-trained large language model; the second text generation model is used to generate corresponding function key images based on the content information of the icon descriptive text.
[0117] Based on any of the above embodiments, the device further includes: An icon output module, used for inputting the function key image into a foreground segmentation model to obtain a function key icon output by the foreground segmentation model; The foreground segmentation model is used to segment and obtain foreground information of the function key image.
[0118] Based on any of the above embodiments, the icon descriptive text includes multiple icon descriptive sub-texts, any of the icon descriptive sub-texts includes image content information of a function key icon; the number of the function key images is multiple; the icon output module is specifically used for: For any of the function key images, input the function key image into a foreground segmentation model to obtain a foreground icon output by the foreground segmentation model; Inputting the foreground icon into an image quality assessment model to obtain an image quality assessment result output by the image quality assessment model; Based on the image quality evaluation result, determining whether to use the foreground icon as a function key icon; Obtaining a number of said function key icons; Wherein, the image quality assessment model is obtained by training based on sample foreground icons.
[0119] Based on any of the above embodiments, the input text further includes a style word text, and the style word text is used to indicate the style of the input method background image; the image output module is specifically used to: The icon descriptive text and the style word text are input into a second text-generated image model to obtain a function key image output by the second text-generated image model.
[0120] Based on any of the above embodiments, the device further includes: An information acquisition module, used to acquire input method skin configuration information corresponding to a target device; the target device is a device for which an input method skin needs to be generated; The method determination module is used to determine the generation method of the input method skin based on the input method skin configuration information.
[0121] Figure 6 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 6As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the input method skin generation method, the method comprising: obtaining input text; the input text comprises an image descriptive text input by a user, the image descriptive text comprises image content information of an input method background image; inputting the input text into a first text image model to obtain an input method background image output by the first text image model; the input method background image is used to generate an input method skin; wherein the first text image model is used to generate a corresponding input method background image based on the content information of the input text.
[0122] In addition, the logic instructions in the above-mentioned memory 630 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.
[0123] On the other hand, the present invention also provides a computer program product, which includes a computer program, and 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 input method skin generation method provided by the above-mentioned methods, and the method includes: obtaining input text; the input text includes an image descriptive text input by a user, and the image descriptive text includes image content information of an input method background image; inputting the input text into a first text-generated image model to obtain an input method background image output by the first text-generated image model; the input method background image is used to generate an input method skin; wherein, the first text-generated image model is used to generate a corresponding input method background image based on the content information of the input text.
[0124] 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 input method skin generation method provided by the above-mentioned methods, the method comprising: obtaining input text; the input text comprises an image descriptive text input by a user, and the image descriptive text comprises image content information of an input method background image; the input text is input into a first text-generated image model to obtain an input method background image output by the first text-generated image model; the input method background image is used to generate an input method skin; wherein the first text-generated image model is used to generate a corresponding input method background image based on the content information of the input text.
[0125] 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.
[0126] 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.
[0127] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for generating an input method skin, characterized in that: include: Acquire input text; the input text includes image descriptive text input by the user, and the image descriptive text includes image content information of the input method background image; Inputting the input text into a first text-generated image model to obtain an input method background image output by the first text-generated image model; the input method background image is used to generate an input method skin; The first text-generated graph model is used to generate a corresponding input method background graph based on content information of the input text.
2. The input method skin generation method according to claim 1, characterized in that: The input text also includes style word text, and the style word text is used to indicate the style of the input method background image.
3. The input method skin generation method according to claim 1, characterized in that: The step of inputting the input text into a first text-generated graph model to obtain an input method background image output by the first text-generated graph model includes: Inputting the image descriptive text into a text polishing model to obtain a polished text output by the text polishing model; Inputting the polished text into a first text-generated image model to obtain an input method background image output by the first text-generated image model; The text polishing model is a pre-trained large language model.
4. The input method skin generation method according to any one of claims 1 to 3, characterized in that: After obtaining the input text, the method further includes: Inputting the input text into an icon descriptive text generation model to obtain an icon descriptive text output by the icon descriptive text generation model; the icon descriptive text includes image content information of the function key icon; Inputting the icon descriptive text into a second text-generated image model to obtain a function key image output by the second text-generated image model; the function key image includes a function key icon, and the input method background image and the function key icon are used together to generate an input method skin; Among them, the icon descriptive text generation model is a pre-trained large language model; the second text generation model is used to generate corresponding function key images based on the content information of the icon descriptive text.
5. The input method skin generation method according to claim 4, characterized in that: After inputting the icon descriptive text into the second text-generated image model to obtain the function key image output by the second text-generated image model, the method further includes: Inputting the function key image into a foreground segmentation model to obtain a function key icon output by the foreground segmentation model; The foreground segmentation model is used to segment and obtain foreground information of the function key image.
6. The input method skin generation method according to claim 5, characterized in that: The icon descriptive text includes a plurality of icon descriptive sub-texts, any of which includes image content information of a function key icon; the number of the function key images is multiple; The step of inputting the function key image into a foreground segmentation model to obtain a function key icon output by the foreground segmentation model includes: For any of the function key images, input the function key image into a foreground segmentation model to obtain a foreground icon output by the foreground segmentation model; Inputting the foreground icon into an image quality assessment model to obtain an image quality assessment result output by the image quality assessment model; Based on the image quality evaluation result, determining whether to use the foreground icon as a function key icon; Obtaining a number of said function key icons; Wherein, the image quality assessment model is obtained by training based on sample foreground icons.
7. The input method skin generation method according to claim 4, characterized in that: The input text also includes a style word text, and the style word text is used to indicate the style of the input method background image; The step of inputting the icon descriptive text into a second text-generated image model to obtain a function key image output by the second text-generated image model includes: The icon descriptive text and the style word text are input into a second text-generated image model to obtain a function key image output by the second text-generated image model.
8. The input method skin generation method according to any one of claims 1 to 3, characterized in that: After inputting the input text into the first text-generated graph model to obtain the input method background image output by the first text-generated graph model, the method further includes: Obtaining input method skin configuration information corresponding to a target device; the target device is a device for which an input method skin needs to be generated; Based on the input method skin configuration information, a generation method of the input method skin is determined.
9. An input method skin generating device, characterized in that: include: An acquisition module, used to acquire input text; the input text includes image descriptive text input by a user, and the image descriptive text includes image content information of an input method background image; An input module, used for inputting the input text into a first text-generated image model to obtain an input method background image output by the first text-generated image model; the input method background image is used to generate an input method skin; The first text-generated graph model is used to generate a corresponding input method background graph based on content information of the input text.
10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the input method skin generation method according to any one of claims 1 to 8 is implemented.
11. 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 input method skin generation method according to any one of claims 1 to 8 is implemented.
12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the input method skin generation method according to any one of claims 1 to 8 is implemented.