Page difference detection method and device, readable storage medium and electronic equipment

Abstract

The embodiment of the application discloses a page difference detection method and device, a readable storage medium and an electronic equipment. The specific implementation scheme is: obtaining a page image of a to-be-detected page and a corresponding design reference image; performing difference detection on the page image and the design reference image to determine initial difference information; inputting the page image, the design reference image and the initial difference information into a large model to enable the large model to perform checking processing on the initial difference information based on the page image and the design reference image; and determining target difference information based on the output of the large model. In the scheme, the initial difference information is determined by performing preliminary difference detection on the page image and the design reference image, and then the large model performs checking processing on the initial difference information, thereby ensuring the accuracy of the final target difference information and effectively realizing accurate detection of the difference between the page image and the design reference image.

Description

Technical Field

[0001] This application relates to the fields of image processing and large language model technology, and in particular to a page difference detection method, apparatus, readable storage medium and electronic device. Background Technology

[0002] In the field of user interface (UI) design and development, in order to ensure that the final page effect is consistent with the original design reference image, it is usually necessary to perform difference detection between the design reference image and the actual page image to locate areas where there are visual differences.

[0003] How to effectively and accurately detect the differences between the design baseline image and the page image has become an important technical problem. Summary of the Invention

[0004] This application provides a page difference detection method, apparatus, readable storage medium, and electronic device.

[0005] This application provides the following solution: According to the first aspect, a page difference detection method is provided, the method comprising: Obtain the page image of the page to be inspected and the corresponding design baseline image; Perform difference detection between the page image and the design baseline image to determine initial difference information; The page image, design baseline image, and initial difference information are input into the large model so that the large model can verify the initial difference information based on the page image and design baseline image. The target difference information is determined based on the output of the large model.

[0006] According to a second aspect, a page difference detection device is provided, the device comprising: The image acquisition module is used to acquire the page image of the page to be inspected and the corresponding design baseline image; The initial difference detection module is used to detect the differences between the page image and the design reference image in order to determine the initial difference information; The initial difference verification module is used to input the page image, design baseline image and initial difference information into the large model, so that the large model can verify the initial difference information based on the page image and design baseline image; The target difference determination module is used to determine target difference information based on the output of the large model.

[0007] According to a third aspect, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the page difference detection method described above.

[0008] According to the fourth aspect, an electronic device is provided, comprising: One or more processors; and A memory associated with one or more processors, the memory being used to store program instructions, which, when read and executed by one or more processors, perform the steps of the page difference detection method described above.

[0009] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described page difference detection method.

[0010] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application acquires the page image of the page to be detected and the corresponding design reference image, performs difference detection between the page image and the design reference image to determine initial difference information, and inputs the page image, design reference image, and initial difference information into a large model. The large model then verifies the initial difference information based on the page image and design reference image, and determines the target difference information based on the output of the large model. In this scheme, by performing preliminary difference detection on the page image and design reference image to determine initial difference information, and then verifying the initial difference information through a large model, the accuracy of the final target difference information is ensured, effectively achieving precise detection of the difference between the page image and the design reference image.

[0011] Of course, any product implementing this application does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a system architecture diagram applicable to the embodiments of this application.

[0014] Figure 2 This is a schematic flowchart of the page difference detection method provided in the embodiments of this application.

[0015] Figure 3 This is a schematic diagram of a difference comparison page provided in an embodiment of this application.

[0016] Figure 4 This is a schematic diagram of the page difference detection device provided in an embodiment of this application.

[0017] Figure 5 A schematic block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0019] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. The singular forms “a,” “the,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0020] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0021] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0022] In the field of user interface (UI) design and development, to detect differences between design reference images and actual page images, relevant technologies typically employ manual comparison methods or pixel-based difference comparison methods. Manual comparison requires developers to meticulously compare the visual details of the design reference image and the page image item by item, which is labor-intensive and prone to overlooking differences. Pixel-based difference comparison methods locate discrepancy areas by calculating the pixel differences between corresponding pixels in the design reference image and the page image. While this method can comprehensively compare pixel-level differences, it lacks an understanding of semantic information such as the structure of page elements. This results in an inability to perform semantic attribution judgments on detected differences, such as failing to determine differences in the color or positional offset of page elements, leading to poor accuracy in difference detection.

[0023] Therefore, there is an urgent need for a page difference detection method that can accurately detect the differences between the design reference image and the page image.

[0024] In view of this, this application provides a new approach. To facilitate understanding of this application, the system architecture on which this application is based is first described. For example... Figure 1 The diagram shows an exemplary system architecture that can be applied to embodiments of this application, such as... Figure 1 As shown, the system architecture may include: a server-side component and a user-side component running on a user terminal.

[0025] The server-side and the client-side are the two main components of an application service. The server-side uses a server as its primary hardware infrastructure and can include one or more software service modules. The client-side can be a client application, a mini-program, or a web application running through a browser.

[0026] The user terminal can include, but is not limited to, smart mobile terminals, wearable devices, and PCs (Personal Computers). Smart mobile devices can include smartphones, tablets, PDAs (Personal Digital Assistants), and connected car terminals. Wearable devices can include smartwatches, smart glasses, smart bracelets, VR (Virtual Reality) devices, AR (Augmented Reality) devices, and mixed reality devices (devices that support both virtual and augmented reality), etc.

[0027] A server can be a standalone server, a server cluster, or a cloud server. A cloud server, also known as a cloud computing server or cloud host, is a hosting product within the cloud computing service system, designed to address the shortcomings of traditional physical hosts and Virtual Private Servers (VPS) services, such as high management difficulty and weak service scalability.

[0028] As an alternative approach, the user can obtain the page image and design reference image of the page to be detected through the client, and send the page image and design reference image of the page to be detected to the server so that the server can execute the page difference detection method provided in this application and finally obtain the target difference information.

[0029] It should be understood that, Figure 1 The number of servers, clients, and user terminals shown is merely illustrative. Depending on implementation needs, there can be any number of servers, clients, and user terminals.

[0030] Figure 2 This is a flowchart illustrating the page difference detection method provided in an embodiment of this application. Figure 2 As shown, the method may include the following steps: Step S210: Obtain the page image of the page to be inspected and the corresponding design baseline image; Step S220: Perform difference detection between the page image and the design reference image to determine the initial difference information; Step S230: Input the page image, design baseline image, and initial difference information into the large model so that the large model can perform verification processing on the initial difference information based on the page image and design baseline image; Step S240: Determine the target difference information based on the output of the large model.

[0031] As can be seen from the above process, this application acquires the page image of the page to be detected and the corresponding design reference image, performs difference detection on the page image and the design reference image to determine the initial difference information, and inputs the page image, design reference image, and initial difference information into a large model. The large model then verifies the initial difference information based on the page image and design reference image, and determines the target difference information based on the output of the large model. In this scheme, by performing preliminary difference detection on the page image and design reference image to determine the initial difference information, and then verifying the initial difference information through a large model, the accuracy of the final target difference information is ensured, effectively achieving precise detection of the difference between the page image and the design reference image.

[0032] The following describes in detail each step of the above process and the effects that can be further produced, with reference to the embodiments. It should be noted that the terms "first" and "second" involved in this application do not have limitations in terms of size, order, or quantity, but are only used to distinguish them in name. For example, "first region" and "second region" are used to distinguish two different image regions.

[0033] First, the above step S210, namely "obtaining the page image of the page to be detected and the corresponding design baseline image", will be described in detail with reference to the embodiments.

[0034] The design benchmark image refers to a static image that serves as a visual reference standard for the page to be inspected. For example, the design benchmark image can be designed and exported using UI design tools.

[0035] A page image refers to the image displayed after the page to be tested is actually shown. For example, a page image can be obtained through methods such as screenshots.

[0036] As one possible implementation, the page difference detection method provided in this application embodiment can be applied to UI design tools (such as Figma) as a plugin. The UI design tool can call this plugin, directly upload the design reference image and the page image, and then initiate difference detection between the design reference image and the page image.

[0037] As one possible approach, if the design reference image is designed in a UI design tool, the design reference image can be selected directly in the UI design tool, while simultaneously uploading the page image, and then a difference detection between the design reference image and the page image can be initiated.

[0038] The following describes in detail step S220, namely "performing difference detection between the page image and the design reference image to determine initial difference information," with reference to an embodiment.

[0039] The initial difference information can be understood as the detection result obtained by performing preliminary difference detection between the page image and the design reference image.

[0040] As an optional approach, baseline specification parameters of each page element (such as text, icons, etc.) in the design baseline image can be obtained, including attributes such as the page element's representation, type, position, size, color, rounded corners, and font. During difference detection, based on the rule parameters of each page element, corresponding regions can be found in the page image through feature matching or position matching. Then, the actual rule parameters of the page elements in that region are identified and compared with the baseline specification parameters.

[0041] For example, the basic specifications of page elements in the design reference image can be obtained from UI design tools or by identifying page elements in the design reference image using image recognition algorithms.

[0042] As an optional method, difference detection is performed between the page image and the design baseline image, including: performing difference detection between the page image and the design baseline image based on difference checking rules corresponding to the types of each page element in the page to be detected.

[0043] Different types of page elements require different dimensions and indicators for difference detection. Therefore, corresponding difference detection rules can be configured for different types of page elements to serve as refined difference detection standards, thereby improving the accuracy of the generated initial difference detection information.

[0044] For example, for icon elements, the dimensions for difference detection can include color, graphic outline, stroke thickness, size, and position. For search box elements, the dimensions for difference detection can include border color and background color, by performing difference detection on each separately. For card elements, the dimensions for difference detection can include attributes such as shadow, rounded corners, and padding. By configuring the above difference detection rules, the generated initial difference information can serve as a problem description pointing to specific repair actions, such as "incorrect icon color," "inconsistent search box border color," and "inconsistent card rounded corners."

[0045] For example, the difference detection rules can also include a tolerance strategy, that is, setting threshold ranges for different specification parameters, and only judging the existence of a difference when the value exceeds the threshold range. For example, when the positional offset of the same element in the design reference image and the page image is no more than 1 pixel, the positional difference can be ignored; when the positional offset is greater than 1 pixel but no more than 2 pixels, a slight positional difference can be judged, with a low severity level; when the positional offset is greater than 3 pixels, a significant positional difference can be judged, with a high severity level. As another example, for color differences, the internationally accepted color difference formula (such as CIEDE2000) can be used to calculate the ΔE value. When the ΔE value is less than 3, the color difference is too small to be easily noticed and can be ignored; when the ΔE value is greater than or equal to 3, a color difference can be judged, with a low severity level; when the ΔE value is greater than or equal to 5, a significant color difference can be judged, with a high severity level. As yet another example, when the font size difference of the same text element in the design reference image and the page image is less than 1 pixel, the font size difference can be ignored. By setting a tolerance strategy, subtle differences introduced by issues such as rendering anti-aliasing and image compression can be effectively filtered out, reducing the misjudgment rate of initial difference information.

[0046] For example, the initial difference information can be specifically in the form of a list of candidate difference questions. The question information for each candidate difference question may specifically include, for example, the identifier of the page element, the coordinates of the corresponding difference area of ​​the page element, the difference type, the difference quantification value (such as color difference, font size difference, etc.), and a confidence score determined based on a preset algorithm. In this example, the initial difference information can serve as evidence with preliminary classification, but may contain a set of candidate questions with misclassification and redundancy, providing a foundation for subsequent targeted analysis of the initial difference information based on a large model.

[0047] The following describes in detail, with reference to the embodiments, step S230, namely, "inputting the page image, the design reference image, and the initial difference information into the large model so that the large model can perform verification processing on the initial difference information based on the page image and the design reference image" and step S240, namely, "determining the target difference information based on the output of the large model".

[0048] In this embodiment, the large model is a multimodal large model with visual understanding capabilities, capable of image analysis and comparison of page images and design baseline images, thereby verifying initial difference information. Verification processing can be understood as the large model reviewing and correcting initial difference information based on semantic understanding and reasoning of the design baseline image and page image.

[0049] For example, the output of the large model can be the result of the verification process, which can represent the review and correction information for each initial difference information. Based on the result of the verification process and the initial difference information, the final target difference information can be generated.

[0050] In summary, based on steps S210 to S240, this application determines the initial difference information by performing preliminary difference detection between the page image and the design reference image, and then verifies the initial difference information using a large model, thereby ensuring the accuracy of the final target difference information and achieving accurate detection of the difference between the page image and the design reference image.

[0051] In this solution, by verifying and processing the initial difference information through a large model, semantic descriptions of page element differences can be achieved, improving the readability of the difference information and facilitating a clearer and more accurate understanding of the root causes and specific deviations that lead to the differences. This helps guide subsequent repair work and improves development efficiency.

[0052] In one optional embodiment of this application, the page image, the design baseline image, and initial difference information are input into a large model, so that the large model performs verification processing on the initial difference information based on the page image and the design baseline image, including: Based on the page image and the design reference image, a alignment image is generated. The alignment image is used to characterize the positional relationship between the page elements of the page to be detected in the page image and the design reference image. Based on the page image, design baseline image, alignment image, and initial difference information, prompt instructions are generated. These prompt instructions are used to prompt the large model to perform verification processing on the initial difference information based on the page image, design baseline image, and alignment image. Input the prompt command into the large model so that the large model can undergo verification processing.

[0053] Alignment images are auxiliary images synthesized to represent the positional relationships between elements in the page image and the design reference image. Alignment images enable large models to accurately understand the positional relationships between elements.

[0054] For example, a page image can be overlaid with a design reference image, and a specific color blending mode can be applied to the overlaid image. For instance, the design reference image can be set as the background overlaid at the bottom, and the page image can be overlaid on top of it with a certain degree of transparency (e.g., 50%) to create a matching image. In this matching image, the positional differences between the page image and the design reference image will be clearly displayed. In this example, before generating the matching image, the image sizes of the page image and the design reference image can be adjusted to match, such as maintaining the same width, to ensure the effect of the generated matching image.

[0055] A prompt instruction is a structured piece of natural language text that serves as the core command and contextual description guiding a large model to perform a specific task. For example, page images, design reference images, and alignment images can be embedded as attachments to the prompt instructions, or links to access the page images, design reference images, and alignment images can be embedded within the prompt instructions.

[0056] In this approach, by adding alignment images representing the positional relationships of page elements to the prompt instructions, the large model's ability to understand the positional relationships of page elements can be improved, which helps to enhance the detection capability of positional differences.

[0057] In one optional embodiment of this application, the output of the large model includes correction information for the initial difference information, and the prompting instruction includes a first sub-prompting instruction, which is used to prompt the large model to perform the following processing: For each page element in the page to be detected, determine the degree of similarity between the first region of the page element in the page image and the second region of the page element in the design reference image; In response to the similarity level meeting the preset similarity conditions, the first removal prompt message or the second removal prompt message is determined as the correction information corresponding to the page element. The first removal prompt message is used to remove the initial difference information corresponding to the page element, and the second removal prompt message is used to remove the initial difference information corresponding to the page element whose difference level does not meet the preset difference conditions. In response to the similarity level not meeting the preset similarity conditions, the style differences and positional differences of the page element in the page image and the design baseline image are determined, and the correction information corresponding to the page element is determined based on the style differences and positional differences.

[0058] Among them, the correction information is a set of specific modifications to the initial difference information output by the large model after performing the verification process.

[0059] For any page element, the first region is the image area where the page element is located within the page image, and the second region is the image area where the page element is located within the design reference image. For example, the first and second regions can be determined based on the position coordinates of the page element. These position coordinates can be obtained from UI design tools or based on the identification and positioning of the page element.

[0060] Similarity is used to quantify the similarity between the first and second regions. When the similarity meets the preset similarity conditions, it indicates that the first and second regions are basically identical. In this case, the initial difference information of the page element is likely to be entirely misjudged. In this situation, the first sub-hint instruction can prompt the large model to use the first removal hint as correction information. That is, the correction method suggested in the first removal hint is to remove all the initial difference information corresponding to the page element. If the degree of difference described by the initial difference information corresponding to the page element is too large when the first and second regions are basically identical, it does not meet the premise that the first and second regions are basically identical. These initial difference information may be misjudged. The first sub-hint instruction can prompt the large model to use the second removal hint as correction information. That is, the correction method suggested by the first sub-hint instruction is to remove the initial difference information corresponding to the page element whose degree of difference meets the preset difference conditions. For example, if a page element has a color difference ΔE=7 between the page image and the design reference image, this color difference is relatively obvious, and its degree of difference does not meet the preset difference conditions. In this case, the correction information can include the second removal hint to remove this initial difference information describing a significant color difference.

[0061] Conversely, if the similarity does not meet the preset similarity conditions, it indicates that the first region and the second region are inconsistent. In this case, the first sub-prompt instruction can prompt the large model to conduct in-depth analysis to specifically determine the style differences (such as fill color, stroke color, font, rounded corners, etc.) and position differences (such as horizontal or vertical offset, center line misalignment, etc.) between the element in the page image and the design baseline image. Then, based on the analyzed style differences or position differences, the initial difference information is verified to generate the corresponding correction information for the element. For example, if an error is found in a certain initial difference information, correction information is generated to correct the error; if the analyzed style difference or position difference does not exist in the initial difference information, correction information is generated to supplement the initial difference information with the style difference or position difference.

[0062] In this solution, by judging the similarity between the corresponding regions in the page image and the design reference image, and taking different methods to generate correction information for different similarity levels, it can avoid the large model consuming resources on subtle differences in image regions with high similarity, while also enabling in-depth difference analysis for regions with low similarity, thus ensuring the accuracy of the generated correction information.

[0063] In one alternative embodiment of this application, the prompt instruction includes a second sub-prompt instruction, which prompts the large model to identify style and positional differences between page elements in the page image and the design baseline image based on difference checking rules corresponding to the type of page elements.

[0064] The page element type refers to the classification of page elements based on their function and other factors, such as icons, search boxes, buttons, text, and card containers. Each page element type can be associated with a set of preset difference checking rules. These rules define the dimensions to be checked and the corresponding difference thresholds when inspecting for differences in that page element.

[0065] In this solution, the difference checking rules corresponding to the type of page element can be embedded into the second sub-hint instruction, so that the large model can perform difference analysis on the corresponding type of page element based on the difference checking rules, thereby improving the accuracy of difference detection.

[0066] In one optional embodiment of this application, determining the similarity between a first region of the page element in the page image and a second region of the page element in the design reference image includes: The degree of similarity between the first region and the second region is determined based on at least one of the first similarity, the second similarity, and the similarity difference. Wherein, the first similarity represents the similarity between the first region and the second region, the second similarity represents the similarity between the first region and the second region after aligning their positions, and the similarity difference represents the difference between the first similarity and the second similarity.

[0067] The first similarity, also known as the direct similarity, is calculated by directly comparing the first and second regions. The first similarity reflects the visual matching between the first and second regions without compensating for any potential positional shifts.

[0068] For example, for a button element, if the first similarity between its first and second regions is calculated to be 0.68, then the first similarity score is 0.68. This low first similarity score may be due to positional offset, shape change, or internal structural differences.

[0069] The second similarity, also known as the optimal similarity, represents the best similarity that the first and second regions can achieve after small-scale positional alignment. For example, either the first or second region can be allowed to move within a limited distance (e.g., 10 pixels) to find a new position that maximizes their matching degree, and the image similarity at that position is the second similarity.

[0070] Continuing with the example of the button element above, after a position search, it was found that when the button area in the page image is shifted 2 pixels to the left and 1 pixel upward, its second similarity with the button area in the design baseline image is 0.95.

[0071] The similarity difference is the arithmetic difference between the second similarity and the first similarity, which can measure the contribution of positional offset to the current visual difference.

[0072] Continuing with the button example above, the similarity difference is 0.95 - 0.68 = 0.27. This large increase in similarity difference indicates that the current difference is mainly due to misalignment of the elements, while their structures may be intact or highly similar. Conversely, a small increase in similarity difference (e.g., less than 0.05) indicates that even with positional correction, the two elements remain dissimilar, and the difference primarily stems from style differences in the page elements.

[0073] As an optional approach, when calculating similarity, the aforementioned single similarity index or a combination of multiple indices can be flexibly used for judgment. For example, only the first similarity can be used, and it can be set that when the value of the first similarity is greater than a first set threshold (e.g., 0.75), it indicates that the first region and the second region are similar. Alternatively, the first similarity, the second similarity, and the similarity difference can be used in combination. For instance, it can be set that when the value of the first similarity is greater than the second set threshold (e.g., 0.75), the value of the second similarity is greater than the third set threshold (e.g., 0.7), and the similarity difference is greater than the fourth set threshold (0.2), it indicates that the first region and the second region are similar.

[0074] In this scheme, by introducing multi-dimensional and quantifiable similarity indicators, the similarity between the first region and the second region can be accurately determined, providing a key basis for generating accurate correction information in the future.

[0075] In one optional embodiment of this application, generating a alignment image based on a page image and a design reference image includes: The page image is overlaid with the design baseline image, and the image on top is color-doped to generate a alignment image. The prompt instruction includes a third sub-prompt instruction, which is used to prompt the large model to remove color differences when the page element is detected to have positional and color differences between the page image and the design reference image based on the alignment image, and the color differences do not meet the preset color difference conditions.

[0076] Alignment images are generated by overlaying a page image with a design baseline image and then applying a color-dodge effect to the overlaid image. While this method clearly reflects the positional relationships of page elements, it weakens the presentation of color differences, making it prone to misjudgments of color differences detected based on alignment images.

[0077] To ensure the large model can correctly understand and utilize the characteristics of the alignment images, the prompt instructions include a third sub-prompt instruction. This third sub-instruction clarifies a correction rule to the large model: when a positional and color difference is detected in a page element based on the alignment image, if the degree of color difference does not meet the preset color difference condition, the color difference should be removed, and only the more obvious positional difference should be retained.

[0078] The color difference condition can be a quantifiable threshold used to define whether the degree of color difference is negligible or significant. The color difference condition can follow the tolerance strategy in the previous example, for example, setting the color difference condition to "ΔE<4" or "ΔE<5".

[0079] For example, in the difference detection of a button component, the large model detected a color difference of ΔE=3.5 in the button's background area, but also found that the overall outline of the button had a horizontal offset of 3 pixels between the design reference image and the page image. The large model can determine that there is a positional difference of "button alignment offset" rather than a color difference of "button color error", thus avoiding misjudgment of color difference.

[0080] For example, the third sub-hint could be "Please note: In a positioning image, if a page element has a positional offset and the color difference of that page element area is not significant (e.g., ΔE < 4), you should prioritize classifying it as a positional difference and ignore the color difference."

[0081] In this scheme, since the color difference is weakened in the alignment image generated after color dodge processing, in order to avoid misjudgment of color difference, when color difference and position difference coexist and the color difference is not significant, the color difference can be removed, thereby continuously and reliably correcting the misjudgment of color difference and improving the accuracy of difference detection.

[0082] To address the issue of misjudgment of differences caused by dynamically changing content on the page to be detected, in one optional embodiment of this application, the output of the large model includes correction information for the initial difference information, and the prompting instruction includes a fourth sub-prompting instruction, which is used to prompt the large model to perform the following processing: Identify regions of variable page elements in the page image and the design baseline image. Variable page elements are elements in the page to be detected whose content can be switched. The third removal prompt is identified as the correction information corresponding to the variable page element. The third removal prompt is used to remove the initial difference information describing the differences in the element content of the variable page element.

[0083] Among them, variable page elements refer to page elements in the page to be tested whose content is allowed to change dynamically, such as carousel product display images, news header images, etc., as well as text that changes dynamically with the carousel images, such as product description text, etc.

[0084] Differences in the content of variable page elements are not the intended focus of inspection; rather, the style of the container holding the variable page elements (such as size, position, rounded corners, borders, and shadows) is the key area to examine. To avoid misjudgments caused by detecting differences in the content of variable page elements, the prompt instruction also includes a fourth sub-prompt instruction. This fourth sub-instruction is specifically designed to guide the large model to identify and filter out interference caused by differences in the content of the variable page elements themselves, focusing instead on the style differences of the variable page elements.

[0085] For example, the identification of variable page elements can be based on a variety of clues: such as identifying it as a product photo based on image content, or detecting that it is located in a specific position on the page to be detected, such as in the product list.

[0086] For example, if the design baseline image shows an image of product A, while the page image shows an image of product B, and the initial difference information contains information describing the differences between the images of product A and product B, this part of the information describing the differences between the images of product A and product B can be removed from the initial difference information.

[0087] In this solution, by identifying variable page elements and removing the initial difference information describing the differences in the content of dynamically switching elements within the variable page elements, the misjudgment of differences in the content of variable page elements is reduced, thereby improving the accuracy of difference detection.

[0088] In one alternative approach of this application, determining the target difference information based on the output of a large model includes: The initial difference information is removed, retained, or supplemented based on the correction information to obtain the target difference information.

[0089] Among them, after the large model outputs correction information, the initial difference information can be corrected based on the correction information.

[0090] Correction information can specifically instruct on whether to remove, retain, or supplement initial difference information. Specifically, correction information may explicitly indicate initial difference information that needs to be removed, as this could lead to false positives; such initial difference information can be removed. Correction information may explicitly indicate initial difference information that needs to be retained, as this initial difference information has passed verification and its accuracy has been confirmed; such initial difference information can be retained in the final target difference information. Correction information may explicitly indicate difference information that needs to be supplemented, as this difference information was not detected in the initial difference detection but was detected by the larger model; such difference information can be added to the final target difference information.

[0091] In this scheme, the initial difference information is removed, retained, or supplemented by the correction information output by the large model, which can automatically generate target difference information and help improve the efficiency of difference detection.

[0092] In one optional embodiment of this application, the initial difference information is removed, retained, or supplemented based on the correction information to obtain the target difference information, including: Based on the correction information, the initial difference information is removed, retained, or supplemented to obtain candidate difference information; Candidate difference information that meets the first preset condition for semantic similarity is deduplicated, and / or candidate difference information that meets the second preset condition for semantic conflict is merged to obtain target difference information.

[0093] Candidate difference information is obtained by removing, retaining, or supplementing the initial difference information through correction. Candidate difference information may contain semantic repetitions, leading to redundancy in the target difference information. It may also contain semantic conflicts, causing contradictions in the target difference information.

[0094] In this embodiment, candidate difference information whose semantic similarity meets a first preset condition can be deduplicated. The first preset condition can be set as follows: if two or more candidate difference pieces correspond to the same page element identifier and have the same difference category, then their semantic similarity is determined to meet the condition, pointing to the same type of problem within the same design element. For example, if two candidate difference pieces both indicate a color error in the "search box," although their description details or confidence levels may differ, it can be determined that they have semantic overlap, thus triggering deduplication.

[0095] Candidate discrepancies that meet the second preset condition regarding semantic conflict can be merged. The degree of semantic conflict measures the extent to which multiple candidate pieces of information disagree on the same fact. When the degree of semantic conflict meets the second preset condition, it indicates a severe semantic conflict, necessitating the merging of the multiple candidate pieces of information involved.

[0096] In this scheme, through the above deduplication and merging processes, the candidate difference information set is further purified and optimized, eliminating internal contradictions and redundancy, which helps to improve the quality of the final target difference information.

[0097] In one optional embodiment of this application, merging candidate difference information whose semantic conflict level satisfies a second preset condition includes at least one of the following: Remove candidate difference information whose confidence level does not meet the third preset condition; Remove candidate difference information whose represented degree of difference does not meet the fourth preset condition; Remove candidate difference information whose completeness of difference details does not meet the fifth preset condition; Candidate difference information that meets the second preset condition for semantic conflict is merged.

[0098] The confidence level represents the reliability of the candidate difference information. When multiple candidate difference information are semantically conflicting (e.g., multiple candidate difference information indicating different color differences for the same page element), the confidence levels of these candidate difference information can be compared, and candidate difference information with a confidence level below the threshold (e.g., 0.6) can be automatically removed.

[0099] For example, the confidence level of each initial difference information can be determined when initially performing difference detection to generate initial difference information.

[0100] In this scheme, when multiple candidate difference information have semantic conflicts, the candidate difference information with higher confidence is retained and the candidate difference information with lower confidence is removed, thereby effectively ensuring the reliability of the final generated target difference information.

[0101] The degree of difference is a quantitative indicator that characterizes the intensity of the difference, such as the ΔE value of color difference, the offset pixel value of positional difference, and the font size difference of font size issues.

[0102] The fourth preset condition is used to measure the severity of the difference. When the difference meets the fourth preset condition, it indicates a serious difference; when the difference does not meet the fourth preset condition, it indicates a minor difference. For example, the fourth preset condition is that the color difference ΔE ≥ 5, meaning that only candidate difference information with a color difference ΔE ≥ 5 is retained.

[0103] In this approach, the accuracy of the final target difference information can be improved by removing minor differences with low degree of variation.

[0104] The completeness of the difference details can assess whether a candidate difference contains the structured information necessary for effective understanding and repair. For example, a candidate difference should include the design baseline value, actual value, and deviation value of a certain specification parameter of a page element. If at least one of these three values ​​is missing, the corresponding candidate difference is incomplete and can be removed.

[0105] In this solution, by removing candidate difference information that lacks information integrity, the information integrity of the final generated target difference information can be guaranteed, which facilitates targeted repair in the future.

[0106] When there are multiple candidate difference messages with semantic conflict levels that meet the second preset condition, and all of these candidate difference messages are valuable, they can be merged. For example, two candidate difference messages with semantic conflict both confirm that button A has a style problem, but one focuses on describing color deviation and the other focuses on describing shadow missing. These two candidate difference messages can be merged into one candidate difference message that includes both of these differences, so as to retain a richer amount of information and achieve a complete description of the difference.

[0107] In one optional embodiment of this application, after obtaining the page image of the page to be inspected and the corresponding design reference image, the method further includes: Adjust the image size of the page image to match the design reference image.

[0108] The design reference image and the page image may differ in size due to their different sources; for example, the design reference image may be 750 pixels wide, while the page image may be 1080 pixels wide. Performing direct difference detection when there is a size difference between the design reference image and the page image is likely to result in detection failure.

[0109] In this solution, the page image and the design reference image are resized to ensure that their dimensions match, thus ensuring the normal operation of the difference detection.

[0110] For example, the specific method for size adjustment can be width normalization, that is, adjusting the page image and the design reference image to the same reference width. The reference width can be directly taken from the width of the design reference image, or a standard width set by the system or manually.

[0111] In one alternative embodiment of this application, the page image, the design baseline image, and initial difference information are input into the large model, including: Perform semantic enhancement processing on the initial difference information; Input the page image, the design baseline image, and the semantically enhanced initial difference information into the large model; Semantic enhancement processing includes at least one of the following: Correct the initial discrepancy information that may lead to misjudgment; The initial difference information describing multiple difference information is split; Convert the initial difference information into a form that includes preset difference parameters; Remove the part describing the difference in element content from the initial difference information corresponding to the variable page element. The variable page element is the element whose content can be switched among the page elements of the page to be detected.

[0112] To improve the data quality and structural clarity of the initial difference information, semantic enhancement processing can be performed on the initial difference information. Subsequently, the page image, the design baseline image, and the semantically enhanced initial difference information are input into the large model together, thereby providing the large model with a clearer, more structured, and easier-to-understand input, which helps to improve the accuracy of the output content of the large model.

[0113] Specifically, semantic enhancement processing can include correcting initial difference information that has been misjudged. This operation can process misjudgments in the initial difference information and improve its accuracy. For example, according to an empirical rule: when the color difference ΔE value of a region is less than 4 (i.e., the color difference is weak), but its edge position is significantly offset, the difference is more likely to be a positional difference than a color difference. When there is initial difference information with weak color difference and significant edge offset, its problem category can be automatically changed from "color misjudgment" to "spacing problem" or "alignment problem," and the description can be adjusted accordingly.

[0114] The initial difference information may contain some general, aggregated difference descriptions, which actually contain multiple differences. Semantic enhancement processing can include splitting the initial difference information describing multiple differences to ensure clear attribution of the initial difference information and to provide more granular input for the large model. For example, for an icon, the initial difference information may only contain the description "icon style error". During semantic enhancement, a specific detection rule can be applied to the icon to split the general description "icon style error" into multiple independent and more precise descriptions, such as "icon color error (ΔE=6.2)", "icon graphic outline inconsistency (similarity 0.7)", and "icon position offset (X direction +2px)".

[0115] To improve the readability and operability of initial difference information, semantic enhancement processing can convert vague descriptions in the initial difference information into parameterized descriptions. In this example, three key parameters can be preset: "design value," "actual value," and "deviation value." By including these three parameters in the semantically enhanced initial difference information, a more standardized data structure can be formed, improving the clarity and readability of the initial difference information. For example, "text color incorrect" can be transformed into "title text color error: design color value #333333, actual color value #666666, color difference ΔE=5.8"; "margin too large" can be transformed into "element right side spacing error: design value 20px, actual value 24px, deviation +4px."

[0116] Semantic enhancement processing can also include removing the portion describing content differences from the initial difference information for variable page elements. Since the content of elements within variable page elements may undergo expected changes (such as periodic switching of carousel images), the initial difference information can ignore these content differences and focus only on the style differences (such as rounded corners and borders) of the variable page elements.

[0117] When performing semantic enhancement, the initial difference information for variable page elements (such as images in a carousel) can be identified. Then, it can be determined whether there is a part in the initial difference information that describes the differences in the content of the variable page elements. If so, the part describing the differences in the content of the variable page elements can be removed or marked as "filtered", thereby filtering out invalid initial difference information in advance and reducing noise in the input large model.

[0118] As an optional approach, to ensure the stability of calling large models, embodiments of this application can provide the following mechanism: Pre-check mechanism: Before initiating a call request to a large model, basic checks are performed on the service provider, model name, interface key format, and service endpoint to ensure that they meet the basic format requirements and match the preset default values, preventing invalid calls due to configuration errors.

[0119] Timeout and retry mechanism: Set timeout for calling large models, and use strategies such as exponential backoff to automatically retry a limited number of times for retryable errors such as connection failure and response timeout, in order to cope with temporary network fluctuations or server load.

[0120] Error diagnosis mechanism: Parse the error information returned when calling the large model, such as common HTTP status codes (such as request error, authentication failure, call frequency exceeding the limit and other abnormal status codes) and network layer exceptions, and convert them into readable prompts to be fed back to the user, so as to facilitate quick location and repair of problems.

[0121] Failure rollback mechanism: When the call to the large model fails due to network or service unavailability, the system can automatically skip the large model verification process and directly output the initial difference information after semantic enhancement and other processing as the final target difference information. This ensures that the page difference detection process can still be completed and produce highly accurate target difference information even when the large model service is unavailable.

[0122] To present the target difference information to users more intuitively, in one optional embodiment of this application, after determining the target difference information, the method further includes: The comparison page displays the differences, including page images, design baseline images, and target difference information. Each target difference information is displayed in the corresponding position of the described page element.

[0123] The difference comparison page is an interactive page used to carry target difference information. The difference comparison page will display the page image and the design baseline image together, and at the same time display the target difference information in the corresponding position of the described page element, so that the target difference information can be displayed in the form of annotation information of the corresponding page element.

[0124] For example, Figure 3 This is a schematic diagram of a difference comparison page provided in an embodiment of this application.

[0125] like Figure 3 As shown, the difference comparison page displays the page image and the design baseline image side by side, and displays the target difference information through annotation boxes at positions horizontally aligned with the page elements where differences exist.

[0126] Specifically, for the page element labeled 1, namely the search box in the image, the background color of the search box is inconsistent. The specific difference is that the background fill color of the search box does not match the design. The design value is #FFFFFF, while the actual value is #000000, resulting in a color difference ΔE of 100, which is a color error that needs to be corrected.

[0127] For the page element labeled 2, i.e. component b in the diagram, component b is not aligned with the other components (component a, component c, component d). The specific difference information is: component b is not horizontally aligned with other icons in the same row. The design baseline is y=820px, but it is actually moved 5px to y=815px, which is an alignment problem.

[0128] For page element number 3, i.e., image material e in the image, which is a carousel image within an image container, the differences in image content can be ignored; only the style differences of the image container are considered. Specifically, the target difference is that the rounded corner radius of the carousel container is not restored. The design value is 0px, the actual value is 8px, a deviation of 8px. The system has ignored the differences in the image content itself and only reports the image container style issue.

[0129] In this solution, an integrated difference comparison page is used to display target difference information and to clearly compare the target difference information with the actual image, so that users can quickly and intuitively understand the target difference information.

[0130] To optimize the presentation of the difference comparison page when the displayed target difference information is densely presented, in one optional embodiment of this application, the above method further includes: In response to overlapping display areas of target difference information on the difference comparison page, the display position of the target difference information is adjusted and / or the amount of information in the target difference information is reduced so that there is no overlap in the display areas of the target difference information.

[0131] Among them, overlapping display areas refer to the visual annotation elements (such as numbered boxes, highlighted areas, or bubble labels) corresponding to the differences between two or more targets overlapping in the display area, which may cause the information to be obscured and unable to be clearly identified.

[0132] In this embodiment of the application, the overlap can be eliminated by adjusting the display position, that is, by moving the position of the annotation element without changing the correspondence between the annotation element and the underlying page element.

[0133] In this embodiment of the application, the core points of the target difference information can be retained by reducing the amount of information, and the text content directly displayed on the annotation element can be simplified, thereby reducing the space occupied by the annotation element and eliminating overlap.

[0134] This solution can avoid overlapping display areas when densely labeled target difference information is displayed. By adjusting the layout or reducing the amount of target difference information, it ensures clear and complete display of target difference information and optimizes the presentation effect of the difference comparison page.

[0135] As an optional approach, after displaying the difference comparison page, this solution also supports user editing of the target difference information. Users can click on any visual labeling element of the target difference information (such as a numbered box) to enter edit mode. Editing operations can include: modifying the issue description text, adjusting the issue category (such as changing from "color" to "style"), changing the severity priority, etc. After the user completes the editing and clicks "Apply Changes," the system can automatically update the stored target difference information and the target difference information displayed on the difference comparison page. The edited target difference information can be persistently saved by the system. The original target difference information and editing history will be recorded to form a traceable review log, which is convenient for subsequent manual review and verification.

[0136] The methods provided in this application embodiment can be applied to a variety of application scenarios, including but not limited to: as a plugin in UI design tools (such as Figma), or as an independent difference detection service, to realize the difference detection between the design reference image and the page image.

[0137] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0138] According to another embodiment, a page difference detection device is provided. Figure 4 A schematic block diagram of the page difference detection device according to one embodiment is shown. Figure 4 As shown, the page difference detection device 400 includes: Image acquisition module 410 is used to acquire the page image of the page to be detected and the corresponding design reference image; The initial difference detection module 420 is used to perform difference detection between the page image and the design reference image to determine the initial difference information; The initial difference verification module 430 is used to input the page image, the design reference image and the initial difference information into the large model, so that the large model can verify the initial difference information based on the page image and the design reference image; The target difference determination module 440 is used to determine target difference information based on the output of the large model.

[0139] As an optional approach, the initial difference verification module 430 is specifically used for: Based on the page image and the design reference image, a alignment image is generated. The alignment image is used to characterize the positional relationship between the page elements of the page to be detected in the page image and the design reference image. Based on the page image, design baseline image, alignment image, and initial difference information, prompt instructions are generated. These prompt instructions are used to prompt the large model to perform verification processing on the initial difference information based on the page image, design baseline image, and alignment image. Input the prompt command into the large model so that the large model can undergo verification processing.

[0140] As an optional approach, the output of the large model includes correction information for the initial difference information, and the prompting instructions include a first sub-prompting instruction, which prompts the large model to perform the following processing: For each page element in the page to be detected, determine the degree of similarity between the first region of the page element in the page image and the second region of the page element in the design reference image; In response to the similarity level meeting the preset similarity conditions, the first removal prompt message or the second removal prompt message is determined as the correction information corresponding to the page element. The first removal prompt message is used to remove the initial difference information corresponding to the page element, and the second removal prompt message is used to remove the initial difference information corresponding to the page element whose difference level does not meet the preset difference conditions. In response to the similarity level not meeting the preset similarity conditions, the style differences and positional differences of the page element in the page image and the design baseline image are determined, and the correction information corresponding to the page element is determined based on the style differences and positional differences.

[0141] As an optional approach, the prompt instruction includes a second sub-prompt instruction, which prompts the large model to identify style and positional differences between page elements in the page image and the design baseline image based on difference checking rules corresponding to the type of page elements.

[0142] As an alternative approach, determining the similarity between a first region of the page element in the page image and a second region of the page element in the design reference image includes: The degree of similarity between the first region and the second region is determined based on at least one of the first similarity, the second similarity, and the similarity difference. Wherein, the first similarity represents the similarity between the first region and the second region, the second similarity represents the similarity between the first region and the second region after aligning their positions, and the similarity difference represents the difference between the first similarity and the second similarity.

[0143] As an optional method, a alignment image is generated based on the page image and the design baseline image, including: The page image is overlaid with the design baseline image, and the image on top is color-doped to generate a alignment image. The prompt instruction includes a third sub-prompt instruction, which is used to prompt the large model to remove color differences when the page element is detected to have positional and color differences between the page image and the design reference image based on the alignment image, and the color differences do not meet the preset color difference conditions.

[0144] As an optional approach, the output of the large model includes correction information for the initial difference information, and the prompting instructions include a fourth sub-prompting instruction, which prompts the large model to perform the following processing: Identify regions of variable page elements in the page image and the design baseline image. Variable page elements are elements in the page to be detected whose content can be switched. The third removal prompt is identified as the correction information corresponding to the variable page element. The third removal prompt is used to remove the initial difference information describing the differences in the element content of the variable page element.

[0145] As an optional approach, the output of the large model includes correction information for the initial difference information. The target difference determination module 440 is specifically used for: The initial difference information is removed, retained, or supplemented based on the correction information to obtain the target difference information.

[0146] As an optional approach, when the target difference determination module 440 removes, retains, or supplements the initial difference information based on the correction information to obtain the target difference information, it is specifically used for: Based on the correction information, the initial difference information is removed, retained, or supplemented to obtain candidate difference information; Candidate difference information that meets the first preset condition for semantic similarity is deduplicated, and / or candidate difference information that meets the second preset condition for semantic conflict is merged to obtain target difference information.

[0147] As an optional approach, when the target difference determination module 440 merges candidate difference information whose semantic conflict level meets the second preset condition, it is specifically used for at least one of the following: Remove candidate difference information whose confidence level does not meet the third preset condition; Remove candidate difference information whose represented degree of difference does not meet the fourth preset condition; Remove candidate difference information whose completeness of difference details does not meet the fifth preset condition; Candidate difference information that meets the second preset condition for semantic conflict is merged.

[0148] As an alternative, the above-mentioned device further includes: The image size adjustment module (not shown in the figure) is used to adjust the image size of the page image and the design reference image after acquiring the page image of the page to be detected and the corresponding design reference image, so that the image size of the page image matches that of the design reference image.

[0149] As an optional approach, the initial difference detection module 420 is specifically used for: Based on the difference detection rules corresponding to the types of page elements in the page to be detected, the page image is used to detect differences between the page image and the design baseline image.

[0150] As an optional method, the initial difference verification module 430 is specifically used for: inputting the page image, design baseline image, and initial difference information into the large model. Perform semantic enhancement processing on the initial difference information; Input the page image, the design baseline image, and the semantically enhanced initial difference information into the large model; Semantic enhancement processing includes at least one of the following: Correct the initial discrepancy information that may lead to misjudgment; The initial difference information describing multiple difference information is split; Convert the initial difference information into a form that includes preset difference parameters; Remove the part describing the difference in element content from the initial difference information corresponding to the variable page element. The variable page element is the element whose content can be switched among the page elements of the page to be detected.

[0151] As an alternative, the above-mentioned device further includes: The difference comparison display module (not shown in the figure) is used to display the difference comparison page after the target difference information is determined. The difference comparison page includes page images, design baseline images, and target difference information. Each target difference information is displayed in the position corresponding to the described page element.

[0152] As an optional approach, the difference comparison display module (not shown in the figure) is also used for: In response to overlapping display areas of target difference information on the difference comparison page, the display position of the target difference information is adjusted and / or the amount of information in the target difference information is reduced so that there is no overlap in the display areas of the target difference information.

[0153] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and relevant parts can be referred to the descriptions in the method embodiments. The systems and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0154] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0155] In addition, embodiments of this application also provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method described in any of the foregoing method embodiments.

[0156] And an electronic device, comprising: One or more processors; and A memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform the steps of the method described in any of the foregoing method embodiments.

[0157] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in any of the foregoing method embodiments.

[0158] in, Figure 5An exemplary architecture of an electronic device is shown, which may include a processor 510, a video display adapter 511, a disk drive 512, an input / output interface 513, a network interface 514, and a memory 520. The processor 510, video display adapter 511, disk drive 512, input / output interface 513, network interface 514, and memory 520 can communicate with each other via a communication bus 530.

[0159] The processor 510 can be implemented using a general-purpose CPU, microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits to execute relevant programs and implement the technical solution provided in this application.

[0160] The memory 520 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 520 can store the operating system 521 for controlling the operation of the electronic device 500, and the basic input / output system (BIOS) 522 for controlling the low-level operations of the electronic device 500. Additionally, it can store a web browser 523, a data storage management system 524, and a page difference detection device 525, etc. The aforementioned page difference detection device 525 can be the application program that specifically implements the aforementioned steps in this embodiment. In summary, when implementing the technical solution provided in this application through software or firmware, the relevant program code is stored in the memory 520 and is called and executed by the processor 510.

[0161] Input / output interface 513 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0162] Network interface 514 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0163] Bus 530 includes a pathway for transmitting information between various components of the device, such as processor 510, video display adapter 511, disk drive 512, input / output interface 513, network interface 514, and memory 520.

[0164] It should be noted that although the above-described device only shows the processor 510, video display adapter 511, disk drive 512, input / output interface 513, network interface 514, memory 520, bus 530, etc., in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the solution of this application, and does not necessarily include all the components shown in the figures.

[0165] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer program product. This computer program product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0166] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for detecting page differences, characterized in that, include: Obtain the page image of the page to be inspected and the corresponding design baseline image; Perform difference detection between the page image and the design baseline image to determine initial difference information; The page image, the design baseline image, and the initial difference information are input into the large model so that the large model performs verification processing on the initial difference information based on the page image and the design baseline image. The target difference information is determined based on the output of the large model.

2. The method according to claim 1, characterized in that, The step of inputting the page image, the design baseline image, and the initial difference information into the large model, so that the large model performs verification processing on the initial difference information based on the page image and the design baseline image, includes: Based on the page image and the design reference image, a alignment image is generated. The alignment image is used to characterize the positional relationship of the page elements of the page to be detected in the page image and the design reference image. Based on the page image, the design baseline image, the alignment image, and the initial difference information, a prompt instruction is generated. The prompt instruction is used to prompt the large model to perform verification processing on the initial difference information based on the page image, the design baseline image, and the alignment image. The prompt instruction is input into the large model so that the large model performs the verification process.

3. The method according to claim 2, characterized in that, The output of the large model includes correction information for the initial difference information, and the prompting instruction includes a first sub-prompting instruction, which prompts the large model to perform the following processing: For each page element in the page to be detected, determine the similarity between the first region of the page element in the page image and the second region of the page element in the design reference image. In response to the similarity degree satisfying a preset similarity condition, a first removal prompt message or a second removal prompt message is determined as the correction information corresponding to the page element. The first removal prompt message is used to remove the initial difference information corresponding to the page element, and the second removal prompt message is used to remove the initial difference information corresponding to the page element whose difference degree does not satisfy the preset difference condition. In response to the similarity level not meeting the preset similarity condition, the style difference and position difference of the page element in the page image and the design reference image are determined, and the correction information corresponding to the page element is determined based on the style difference and position difference.

4. The method according to claim 3, characterized in that, The prompt instruction includes a second sub-prompt instruction, which prompts the large model to identify style and positional differences of the page elements in the page image and the design baseline image based on the difference check rules corresponding to the type of the page elements.

5. The method according to claim 3, characterized in that, Determining the similarity between the first region of the page element in the page image and the second region of the page element in the design reference image includes: The degree of similarity between the first region and the second region is determined based on at least one of the first similarity, the second similarity, and the similarity difference. Wherein, the first similarity represents the similarity between the first region and the second region, the second similarity represents the similarity between the first region and the second region after aligning their positions, and the similarity difference represents the difference between the first similarity and the second similarity.

6. The method according to claim 3, characterized in that, The step of generating a alignment image based on the page image and the design baseline image includes: The page image is overlaid with the design reference image, and the image on top is color-doped to generate the alignment image. The prompt instruction includes a third sub-prompt instruction, which is used to prompt the large model to remove the color difference when the page element is detected to have positional and color differences in the page image and the design reference image based on the alignment image, and the color difference does not meet the preset color difference conditions.

7. The method according to claim 2, characterized in that, The output of the large model includes correction information for the initial difference information, and the prompting instruction includes a fourth sub-prompting instruction, which prompts the large model to perform the following processing: Identify regions of variable page elements in the page image and the design baseline image, wherein the variable page elements are elements whose content can be switched among the page elements of the page to be detected; The third removal prompt is determined to be the correction information corresponding to the variable page element, and the third removal prompt is used to remove the initial difference information describing the element content difference of the variable page element.

8. The method according to claim 2, characterized in that, The output of the large model includes correction information for the initial difference information, and the determination of the target difference information based on the output of the large model includes: Based on the correction information, the initial difference information is removed, retained, or supplemented to obtain the target difference information.

9. The method according to claim 8, characterized in that, The step of removing, retaining, or supplementing the initial difference information based on the correction information to obtain the target difference information includes: Based on the correction information, the initial difference information is removed, retained, or supplemented to obtain candidate difference information; The candidate difference information that meets the first preset condition for semantic similarity is deduplicated, and / or the candidate difference information that meets the second preset condition for semantic conflict is merged to obtain the target difference information.

10. The method according to claim 9, characterized in that, Merging candidate difference information whose semantic conflict level meets a second preset condition includes at least one of the following: Remove candidate difference information whose confidence level does not meet the third preset condition; Remove candidate difference information whose represented degree of difference does not meet the fourth preset condition; Remove candidate difference information whose completeness of the difference details does not meet the fifth preset condition; The candidate difference information that meets the second preset condition for semantic conflict is merged.

11. The method according to any one of claims 1-10, characterized in that, After acquiring the page image of the page to be detected and the corresponding design baseline image, the method further includes: The page image and the design reference image are resized to match their dimensions.

12. The method according to any one of claims 1-10, characterized in that, The difference detection between the page image and the design reference image includes: Based on the difference detection rules corresponding to the types of each page element in the page to be detected, the page image and the design baseline image are subjected to difference detection.

13. The method according to any one of claims 1-10, characterized in that, The step of inputting the page image, the design baseline image, and the initial difference information into the large model includes: The initial difference information is then subjected to semantic enhancement processing; Input the page image, the design baseline image, and the semantically enhanced initial difference information into the large model; The semantic enhancement process includes at least one of the following: Correct the initial difference information that may lead to misjudgment; The initial difference information describing multiple difference information is split; The initial difference information is converted into a form containing preset difference parameters; Remove the portion describing the content difference in the initial difference information corresponding to the variable page element. The variable page element is an element whose content can be switched among the page elements of the page to be detected.

14. The method according to any one of claims 1-10, characterized in that, After determining the target difference information, the method further includes: A difference comparison page is displayed, which includes the page image, the design baseline image, and the target difference information, with each target difference information displayed at the position corresponding to the described page element.

15. The method according to claim 14, characterized in that, Also includes: In response to the overlapping display areas of the target difference information in the difference comparison page, the display position of the target difference information is adjusted and / or the amount of information of the target difference information is reduced so that the display areas of the target difference information do not overlap.

16. A page difference detection device, characterized in that, include: The image acquisition module is used to acquire the page image of the page to be inspected and the corresponding design baseline image; An initial difference detection module is used to perform difference detection between the page image and the design reference image to determine initial difference information; The initial difference verification module is used to input the page image, the design baseline image, and the initial difference information into the large model, so that the large model can perform verification processing on the initial difference information based on the page image and the design baseline image; The target difference determination module is used to determine target difference information based on the output of the large model.

17. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method according to any one of claims 1-15.

18. An electronic device, characterized in that, include: One or more processors; as well as A memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform the steps of the method according to any one of claims 1-15.

19. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the method described in any one of claims 1-15.

Images