AI-based customized clothing display method, system, storage medium and program product

Abstract

The present application provides an AI-based custom clothing display method, system, storage medium, and program product, relating to the field of electronic digital data processing. The method comprises: obtaining multiple clothing design elements contained in the custom clothing; matching the display action with the smallest occlusion ratio for each clothing design element to obtain multiple display action sets; selecting a target display action from each display action set that has the highest repetition rate with the display actions in other display action sets, and dividing the target display actions into corresponding life scenes based on the matching degree to obtain a first display action subset corresponding to each life scene; if the difference between the standard quantity value and the corresponding quantity value of the target display action is greater than a preset quantity threshold, then updating the first display action subset according to the priority order of the clothing design elements, thereby generating a dynamic display video of the custom clothing. Implementation of this method can enable the characteristics of the custom clothing to be fully and efficiently displayed within a limited time.

Description

Technical Field

[0001] The present application relates to the field of electronic digital data processing, and in particular to an AI-based customized clothing display method, system, storage medium and program product. Background Art

[0002] With the development of society and the improvement of people's living standards, clothing has become more than just a basic function of warmth and coverage; it has become a vital medium for expressing individuality, reflecting fashion, and embodying cultural connotations. In particular, in the field of fashion design, designers use diverse design elements such as fabric materials, color combinations, tailoring techniques, and textures and patterns to imbue clothing with unique artistic expression. To enable consumers to more intuitively understand the characteristics of clothing design, clothing display has become a crucial component of the apparel industry. Especially with the rapid development of e-commerce platforms, fashion shows, and virtual fitting environments in recent years, the efficient and comprehensive display of clothing design elements is crucial for enhancing consumer experience and commercial value.

[0003] In the related art, in order to fully display the design elements of clothing, the following two main methods are usually adopted: the first is to use real models or mannequins for dynamic clothing display, and during the display process, the different design features of the clothing are highlighted through the model's movements or posture adjustments, such as showing the flow of the skirt and the three-dimensional effect of the cuff cut; the second is to use static flat display, such as through the use of clothing flat-lay pictures, local close-up pictures and multi-angle pictures to show the overall and detailed design of the clothing. These methods can be achieved through actual shooting or computer rendering technology. The above methods can meet the user's cognitive needs for clothing design elements to a certain extent through real or virtual display forms.

[0004] However, while dynamic display methods in related technologies can intuitively showcase some of a garment's design features, the time constraints of the action display make it difficult to manually adjust all of the garment's design elements in a short period of time. Static, flat-panel display methods, while able to showcase garment designs from multiple angles and with detailed images, lack dynamic effects and struggle to truly recreate the garment's expressiveness in actual wear. Therefore, under related technologies, the display of garment design elements suffers from incomplete coverage and a single, single-effect presentation, making it difficult to meet users' demand for a more sophisticated display of designed garments. Summary of the Invention

[0005] The present application provides an AI-based customized clothing display method, system, storage medium and program product for addressing the problem of how to display all clothing design elements more comprehensively and effectively under the premise of limiting the length of time.

[0006] In a first aspect, the present application provides an AI-based customized clothing display method, which is applied to a clothing display system, and the method comprises:

[0007] Obtaining customization requirements corresponding to customized clothing and multiple clothing design elements included in the customized clothing, wherein the customization requirements correspond to a preset number of life scenes;

[0008] Matching one or more display actions corresponding to each clothing design element from a preset database to obtain a display action set corresponding to each clothing design element, wherein the display action has the smallest occlusion ratio on the clothing design element compared to other actions in the preset database;

[0009] Selecting a target display action with the highest display action repetition rate among other display action sets from each display action set to obtain a target display action set;

[0010] Dividing each target display action in the target display action set into a corresponding life scene according to the matching degree, to obtain a first display action subset corresponding to each life scene;

[0011] If the difference between the standard quantity value and the quantity value corresponding to the target display action is greater than a preset quantity threshold, the target display actions corresponding to the clothing design elements are repeatedly added to different first display action subsets in descending order of priority of the clothing design elements to obtain a second display action subset corresponding to each life scene, the standard quantity value is positively correlated with the limited duration of the display video, and the target display actions in the same second display action subset are different from each other;

[0012] Based on the AI ​​model, the plurality of life scenes and the target display actions of the second display action subset corresponding to each life scene are converted into text descriptions, wherein the text description includes the execution sequence and execution time of each target display action in each life scene;

[0013] A dynamic display video of the customized clothing is generated based on the text description.

[0014] Through the above-mentioned embodiments, the system combines clothing design elements with display actions, selecting the display actions with the smallest occlusion ratio, thereby maximizing the display of design element details. Furthermore, by dividing the target display actions into corresponding life scenarios and using text descriptions to generate the execution sequence and time of coherent actions, dynamic display videos with different time limits are ultimately generated. This method effectively addresses the problem of incomplete coverage of design elements in traditional dynamic clothing displays due to time constraints, allowing the characteristics of clothing to be fully and efficiently displayed within a limited time.

[0015] In some embodiments, after the step of dividing each target display action in the target display action set into a corresponding life scene according to the matching degree to obtain a first display action subset corresponding to each life scene, the method further includes:

[0016] If the difference between the standard quantity value and the quantity value corresponding to the target display action is less than or equal to the preset quantity threshold, the first display action subset and the second display action subset corresponding to each of the life scenes are the same;

[0017] If the target display action corresponding quantity value is greater than the standard quantity value, the clothing design elements are reduced in order of priority from low to high until the target display action corresponding quantity value is less than or equal to the standard quantity value.

[0018] Through the above embodiment, when the number of target display actions is insufficient, the system supplements the number by repeatedly adding actions for high-priority elements; when the number of target display actions is excessive, the system maintains a balance by reducing the display actions for low-priority elements. This dynamic adjustment mechanism ensures that the display content covers core design elements while also meeting the video length limit, thereby enhancing the comprehensiveness and efficiency of the display and avoiding the problem of too much or too little content.

[0019] In some embodiments, the step of repeatedly adding the target display actions corresponding to the clothing design elements to different first display action subsets in descending order of priority of the clothing design elements to obtain the second display action subset corresponding to each of the life scenes specifically includes:

[0020] Repeatedly adding the first target display action corresponding to the clothing design element with the highest priority to different first display action subsets to obtain a second display action subset corresponding to each of the life scenes;

[0021] Calculate the preliminary video duration according to the execution sequence and execution time corresponding to each target display action in the second display action subset in the text description;

[0022] If the preliminary video duration is less than the limited duration, the second target display action corresponding to the clothing design element with the second highest priority is repeatedly added to different first display action subsets and the new preliminary video duration is calculated again for judgment.

[0023] Through the above embodiment, the system prioritizes high-priority design elements and gradually adds the next-highest priority display actions when the video is running short, maximizing the use of the limited duration. This approach effectively addresses the problem of traditional presentations being unable to incorporate all design elements due to insufficient duration, while ensuring that core design elements are prioritized, making the video presentation both comprehensive and focused.

[0024] In some embodiments, the step of positively correlating the number of standards with the limited duration of displaying the video specifically includes:

[0025] Obtaining a standard display duration of each display action in the target display action set and an average duration of all display actions;

[0026] Calculating the first action quantity according to the difference between the limit duration and the standard total duration of the target display action set and the average duration;

[0027] The standard number is obtained by adding the first number of actions and the second number of actions corresponding to the target display action set.

[0028] Through the above-described embodiment, the system calculates the standard and average duration of display actions and dynamically adjusts the number of display actions based on the time limit, ensuring that display actions are neither exceeded nor missed. This dynamic adjustment mechanism based on duration and number of actions not only avoids the problem of insufficient display due to time limits, but also improves the integrity of the display content and the comprehensiveness of the effect, providing a precise time control solution for clothing displays.

[0029] In some embodiments, before the step of converting the plurality of life scenarios and the target display actions of the second display action subset corresponding to each life scenario into text descriptions based on the AI ​​model, the method further includes:

[0030] Obtaining an action start feature and an action end feature of each target display action in the second display action subset corresponding to each life scene;

[0031] Calculating a first similarity between the action ending feature corresponding to each of the life scenes and the action starting features corresponding to other life scenes;

[0032] Determining the first life scene and the second life scene corresponding to the action start feature and the action end feature with the greatest first similarity as two life scenes for continuous display in the dynamic display video, wherein the second life scene in the dynamic display video is displayed before the first life scene;

[0033] Determining, based on the sorting result of the first similarity, the display order of different life scenes in the dynamic display video and the switching display actions corresponding to the different life scenes at the switching points;

[0034] In each life scene, the execution sequence of other target display actions is determined based on the second similarity between the action ending feature of the target display action corresponding to the scene switching point and the other action starting features.

[0035] Through the above embodiment, the system calculates the similarity between the ending feature of each life scene and the starting feature of other scene actions, determines the feature pair with the greatest similarity, and then arranges the corresponding life scenes in a continuous display order. Subsequently, the display order of different life scenes is determined based on the similarity sorting results, and the connection effect between different scenes is coordinated through the target display action at the scene switching. Within the scene, taking the target display action at the scene switching as the starting point, the execution timing of the target display action is dynamically adjusted through the similarity between the ending feature and the starting feature, thereby ensuring the fluency of the content displayed within the scene.

[0036] In some embodiments, it is characterized in that the step of generating a dynamic display video of the customized clothing based on the text description specifically includes:

[0037] generating a preliminary demonstration video based on the text description;

[0038] Counting one or more displayed angle intervals corresponding to each clothing design element in the preliminary display video to obtain a displayed angle interval set;

[0039] determining one or more non-displayed angle intervals based on a preset display angle interval set and the displayed angle interval set to obtain a non-displayed angle interval set, wherein the preset display angle interval set includes all display angle intervals of the custom garment, and the display angle intervals are direction angles of the custom garment relative to the virtual camera lens in the preliminary video;

[0040] Determine the target clothing design element with the highest priority corresponding to each undisplayed angle interval and the target display action corresponding to the target clothing design element in the preliminary display video;

[0041] Under the premise of not changing the execution time of the target display action, the non-displayed angle interval is added to the displayed angle interval by continuously changing the direction angle of the virtual lens to obtain a dynamic display video.

[0042] Through the above-described embodiment, the system counts the set of displayed angle intervals in the preliminary display video and determines the set of undisplayed angle intervals by comparing it with the preset set of displayed angle intervals. Then, based on the highest-priority design element in the undisplayed angle interval set and its corresponding target display action, the direction angle of the virtual lens is dynamically adjusted without changing the execution time, gradually replenishing the undisplayed angle intervals. While ensuring the limited length of the preliminary video, this technology further improves the angle coverage of the dynamic display video, solves the problems of single angles and incomplete expression of design elements in traditional display methods, and improves the comprehensiveness of the displayed content.

[0043] In some embodiments, after the step of generating a dynamic display video of the customized clothing based on the text description, the method further includes:

[0044] Determining modification parameter values ​​of clothing design elements according to a user's modification operation on the customized clothing;

[0045] The modified parameter value is added to and subtracted from a preset parameter to obtain an upper limit and a lower limit of the parameter disturbance interval, wherein the value of the preset parameter is positively correlated with the number of times the user modifies the customized garment;

[0046] generating corresponding candidate customized garments respectively from a plurality of sets of candidate modification values ​​generated at preset intervals within the parameter disturbance interval, the plurality of sets of candidate modification values ​​including the modification parameter values;

[0047] The candidate customized clothing with the lowest similarity to the customized clothing modified by the user history among the plurality of candidate customized clothing is determined as the target customized clothing, and the target customized clothing is used to generate a new dynamic display video.

[0048] Through the above-described embodiment, the system analyzes the user's modification requirements, generates multiple sets of candidate modification values, and selects the candidate garment with the lowest similarity to the user's historical modification as the target, generating a new dynamic display video. This method achieves personalized optimization of the display of clothing design elements within a limited time frame, while ensuring the comprehensive display of core design elements, meeting the user's diverse needs for display effects and improving the adaptability and practical effects of clothing displays.

[0049] In a second aspect, the present application provides a clothing display system, the clothing display system comprising: one or more processors and a memory;

[0050] The memory is coupled to the one or more processors, and the memory is used to store computer program code, and the computer program code includes computer instructions. The one or more processors call the computer instructions so that the clothing display system can implement an AI-based customized clothing display method provided in the above embodiment, which will not be repeated here.

[0051] In a third aspect, the present application provides a computer-readable storage medium comprising instructions. When the instructions are executed on a clothing display system, the clothing display system can implement an AI-based customized clothing display method provided in the above embodiment, which will not be described in detail here.

[0052] In a fourth aspect, the present application provides a computer program product. When the computer program product is run on a clothing display system, the clothing display system can implement an AI-based customized clothing display method provided in the above embodiment, which will not be repeated here.

[0053] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0054] 1. By matching the display action set with the smallest occlusion ratio from a preset database, the target display action with the highest repetition rate is prioritized, and the matching degree is combined to allocate them according to the real-life scenario. The display actions of high-priority design elements are dynamically adjusted to ensure that the clothing design elements are fully covered in the dynamic display video within the limited duration. This method effectively addresses the problem of incomplete design element coverage in traditional dynamic displays due to time constraints, allowing the characteristics of clothing to be fully and efficiently displayed within a limited time.

[0055] 2. By prioritizing design elements, we determine the allocation and order of target display actions, ensuring that high-priority design elements are highlighted in the video while dynamically adjusting the display content of lower-priority elements within a limited duration, thereby achieving a balance between display efficiency and coverage. Furthermore, by calculating the similarity of action features between scenes, we optimize the sequence and coherence of scene transitions, ensuring high-quality dynamic display videos in terms of content fluidity and natural scene quality.

[0056] 3. By dynamically adjusting the virtual camera's direction and angle to supplement the undisplayed angles, the system ensures comprehensive coverage of the clothing design elements without changing the display execution time, thus enhancing the completeness of the displayed content. Furthermore, parameter perturbations are used to generate candidate customized garments. Based on the user's historical modification behavior, the target customized garment that best meets the user's personalized needs is selected to generate a new display video, achieving dynamic optimization based on user feedback. This method not only meets the user's personalized needs, but also enhances the diversity and customization of the display videos. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a flow chart of an AI-based customized clothing display method in an embodiment of the present application;

[0058] Figure 2 This is another flowchart of an AI-based customized clothing display method in an embodiment of the present application;

[0059] Figure 3 This is a schematic diagram of an exemplary scenario in which a clothing display system in an embodiment of the present application dynamically displays video screening and display actions;

[0060] Figure 4 This is a schematic diagram of an exemplary scenario in which the clothing display system in the embodiment of the present application matches display actions to each life scene in the dynamic display video;

[0061] Figure 5This is a schematic diagram of the physical device structure of the clothing display system in the embodiment of the present application. DETAILED DESCRIPTION

[0062] The terms used in the following examples of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the specification and appended claims of this application, the singular expressions "a," "an," "said," "above," "the," and "this" are intended to include plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in this application refers to any or all possible combinations comprising one or more of the listed items.

[0063] In the following, the terms "first" and "second" are used for descriptive purposes only and should not be understood to imply or suggest relative importance or implicitly indicate the number of the technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of this application, unless otherwise specified, "plurality" means two or more.

[0064] It should be noted that, for ease of understanding, in the embodiment of the present application, the "clothing display system" can be simply referred to as the "system", and the two refer to the same target; the "clothing design element" can also be simply referred to as the "design element", and the two refer to the same target; the "display action" can also be simply referred to as the "action", and the two refer to the same target.

[0065] For ease of understanding, the method provided in this implementation is described in the following flow. Figure 1 , which is a flow chart of an AI-based customized clothing display method in an embodiment of the present application.

[0066] S101. Obtain customization requirements corresponding to the customized clothing and multiple clothing design elements included in the customized clothing.

[0067] Specifically, the clothing display system obtains the user's customization requirements for custom clothing during the design phase through a user input interface, order management system, or other data interfaces. After the designer generates the custom clothing based on the customization requirements, they upload the custom clothing to be displayed to the user to the clothing display system. Before generating the specific clothing display video, the system first extracts the specific clothing design elements contained in the custom clothing. Optionally, the clothing display system can obtain design information for custom clothing through multi-source data access. Combining manual input, file parsing, and image recognition technologies, it can identify and extract specific clothing design elements from the design plan. The system can also parse design data in different formats (such as image files, text descriptions, and 3D models) to classify and organize various types of design elements.

[0068] Custom clothing, specifically designed by designers based on user needs, is the subject of the display system's processing. Custom requirements refer to specific user requirements for custom clothing, encompassing size, style, applicable scenarios, and functionality. These requirements correspond to a pre-set number of life scenarios (e.g., work, leisure, sports, and formal occasions). Clothing design elements are the core design elements that comprise custom clothing. These include specific design features such as fabric material (e.g., silk, wool), color schemes (e.g., contrasting red and black, Morandi colors), tailoring techniques (e.g., draping, high-waist designs), textures and patterns (e.g., plaid, floral prints), and decorative accessories (e.g., buttons, embroidery, and tassels).

[0069] Optionally, the clothing display system can extract the style tags corresponding to each clothing design element in the customized clothing, count the types and quantities of each tag and calculate the proportion, determine the tag whose proportion reaches a preset threshold as the first style tag, and at the same time obtain the second style tag corresponding to each life scene from the preset database; then convert the first and second style tags into vector representations, use algorithms such as cosine similarity to calculate the matching degree between the two, and determine the life scenes with a similarity higher than a preset threshold as candidate scenes. If there are multiple candidates, the one with the highest similarity is selected. If not, the top N scenes with relatively high similarity are selected to determine the life scenes used to generate dynamic display videos.

[0070] S102: Match one or more display actions corresponding to each clothing design element from a preset database to obtain a display action set corresponding to each clothing design element.

[0071] The preset database refers to a database pre-established and stored within the clothing display system that contains a large amount of display action data. Display actions refer to specific movements or gestures used to showcase clothing design elements. These movements are performed by models or avatars when displaying clothing, and they highlight the design features of the garment. The occlusion ratio indicates the degree to which a display action obscures a specific clothing design element. A display action with the smallest occlusion ratio of a clothing design element compared to other actions in the preset database indicates that the action maximizes the display of the clothing design element while minimizing occlusion.

[0072] This step is performed after the system obtains the clothing design elements of the custom clothing. It finds the most suitable display action for each design element, so as to subsequently generate a comprehensive and effective dynamic display video. Specifically, the clothing display system searches and matches each clothing design element in a preset database to find a display action that is compatible with the design element. Optionally, the system can filter out one or more suitable display actions from the database based on preset matching rules, such as the occlusion ratio of the action to the design element and the degree of fit between the action and the design element, to form a display action set corresponding to each design element.

[0073] Specific as Figure 3 The figure is a schematic diagram of an exemplary scene of a clothing display system for dynamically displaying video screening and displaying actions in an embodiment of the present application. For ease of understanding, Figure 3 Take only three clothing design elements as an example. Figure 3 In a corresponding specific embodiment, the system can filter out the display actions corresponding to clothing design element 1, clothing design element 2, and clothing design element 3 from a preset database based on the occlusion ratio: clothing design element 1 corresponds to action 1, action 2, and action 3; clothing design element 2 corresponds to action 2 and action 4; clothing design element 3 corresponds to action 5. Thus, the action display sets corresponding to each clothing design element are obtained: set 1 {action 1, action 2, action 3}, set 2 {action 2, action 4}, and set 3 {action 5}.

[0074] S103 : Select a target display action from each display action set, which has the highest repetition rate with the display actions in other display action sets, to obtain a target display action set.

[0075] The repetition rate represents the frequency with which a particular display action appears within a set of other display actions—that is, the ratio of the number of times the action appears in the display action sets of multiple design elements to the total number of sets. This step is performed after the system generates a display action set for each clothing design element. By selecting actions with a high repetition rate, the goal is to improve the consistency and coherence of the display actions, making them easier to distribute and display in different life scenarios.

[0076] Specifically, the clothing display system traverses the display action sets of each clothing design element. For each display action in each set, the system calculates the number of times it appears in all other display action sets to obtain the repetition rate of the action. Then, the display action with the highest repetition rate from each set is selected as the target display action. All selected target display actions are combined to form a target display action set.

[0077] Optionally, the system can create an action frequency statistics table, with rows and columns representing the actions in all displayed action sets. Then, it iterate over each displayed action set and, for each action in the set, increment the count of its occurrences in other sets in the statistics table. After the statistics are complete, for each set, the number of repetitions of each action is checked, and the action with the highest number of repetitions is selected as the target displayed action.

[0078] Optionally, the system can extract all actions from the display action set to form an action list, and then use a hash table to count the number of times each action appears in the list (excluding the number of times in the current set). For each display action set, iterate over the actions in it, query the hash table for the number of times it appears, and select the action with the most repetitions as the target display action.

[0079] Continue with Figure 3 For example, after determining the display action sets (set 1 {action 1, action 2, action 3}, set 2 {action 2, action 4} and set 3 {action 5}), the system can count the distribution data of each action (such as action 1) in all display action sets.

[0080] For example, action 1: set 1;

[0081] Action 2: Set 1, Set 2;

[0082] Action 3: Set 1;

[0083] Action 4: Set 2;

[0084] Action 5: Set 3;

[0085] From this we can see that in Figure 3 In the corresponding embodiment, the same action 2 (i.e., repeated action) appears in both set 1 and set 2, and only action 5 exists in combination 3, so action 2 and action 5 are the target display actions with the highest display action repetition rate. The combination of action 2 and action 5 obtains the target display action set: {action 2, action 5}.

[0086] S104 . Divide each target display action in the target display action set into a corresponding life scene according to the matching degree, and obtain a first display action subset corresponding to each life scene.

[0087] The matching degree indicates the degree of adaptability between the target display action and the real-life scenario. This refers to the degree to which the action style fits the scene's atmosphere, and the degree to which the action function matches the scene's requirements. For example, relaxed and natural display actions are more suitable for casual scenes. This step is executed after the system obtains the target display action set. It aims to assign the actions to appropriate real-life scenarios, preparing for the subsequent generation of display action subsets for each scenario.

[0088] Specifically, the clothing display system calculates the matching degree of each action in the target display action set with each life scene, and divides the actions into corresponding life scenes according to the matching degree (different life scenes correspond to different display actions). Each life scene collects the divided actions to form the corresponding first display action subset.

[0089] Optionally, the system can pre-set matching rules for each life scenario, such as formal and dignified actions for work scenarios and relaxed and natural movements for leisure scenarios. For each target action, its characteristics (such as movement amplitude and posture type) are extracted and compared with the rules for each scenario. The matching degree is calculated and the action is assigned to the scenario with the highest matching degree.

[0090] Optionally, the system can also create an action-scenario matching table, recording the appropriate life scenarios for each display action and the corresponding matching scores. For each action in the target display action set, the matching table is queried to obtain its matching scores for each scenario. The action is then assigned to the scenario with the highest matching score, forming the first display action subset.

[0091] S105: Whether the quantity value corresponding to the target display action is greater than the standard quantity value.

[0092] After determining the target display action set in step S103, the system counts the number of target display actions in the target display action set to obtain a corresponding number value for the target display actions. It then obtains a standard number value related to the dynamic display video duration limit and compares the two to determine whether the corresponding number value for the target display actions is greater than the standard number value. If so, it is determined that the dynamic display video generated with the current number of actions may have timed out, and the process proceeds to step S106 to reduce the corresponding number value for the target display actions. Otherwise, the process proceeds to step S107 to further determine whether the dynamic display video generated with the current number of actions has a display duration that is too short.

[0093] The standard quantity value refers to a reasonable number of display actions determined based on the limited duration of the dynamic display video, and is used to indicate that the number is positively correlated with the limited duration. The longer the duration, the larger the standard quantity value.

[0094] In addition, the clothing display system can traverse each display action in the target display action set, extract the standard display duration corresponding to each action from the preset action attribute database, and accumulate the standard display durations of all actions and divide them by the number of actions to calculate the average duration of all display actions; then calculate the standard total duration of the target display action set, subtract the standard total duration from the limit duration to obtain the duration difference, and then divide the difference by the average duration to obtain the number of first actions that need to be added or reduced; finally, obtain the adjusted second action number of the target display action set, perform arithmetic operations on it and the first action number to obtain a standard number that is positively correlated with the limit duration, which reflects the reasonable total number of display actions under the current duration limit.

[0095] S106. Reduce clothing design elements in descending order of priority.

[0096] Specifically, when the number of target display actions exceeds the standard value, the clothing display system removes the target display actions corresponding to the lower-priority elements in ascending order of priority. After each removal, the system recounts the number of target display actions until the number is less than or equal to the standard value. The system must ensure that when reducing actions, it prioritizes retaining the display of higher-priority elements.

[0097] Optionally, the system can prioritize each garment design element, for example, assigning high priority to fabric material and tailoring techniques, and low priority to decorative accessories. The system then sorts the design elements from low to high priority, removing the target display actions corresponding to the low-priority elements one by one. After each removal, the system checks whether the required number of target display actions is met.

[0098] Optionally, the system can also establish a design element priority table to record the priority level of each element. When it is necessary to reduce actions, start with the element with the lowest priority and remove its corresponding target display actions until the number of target display actions meets the standard. Alternatively, the system determines the priority based on the user's attention to the design elements, and the elements with low user attention have low priority. By analyzing the user's historical data or current customization needs, the priority of each element is determined, and then the actions corresponding to the elements are reduced in order from low to high, which is not limited here.

[0099] S107: Whether the difference between the standard quantity value and the quantity value corresponding to the target display action is greater than a preset quantity threshold.

[0100] Specifically, the clothing display system calculates the difference between the standard quantity value and the quantity value corresponding to the target display action, compares this difference with a preset quantity threshold, and determines whether the difference is greater than the preset quantity threshold. If it is greater, the system proceeds to step S108, determining that the quantity needs to be replenished by repeatedly adding target display actions of high-priority design elements. If it is less than or equal to the preset quantity, the system proceeds to step S109, determining that the quantity value corresponding to the current target display action just meets the time limit requirement, and no additional quantity is needed or the first display action subset is directly used.

[0101] S108. Repeatedly add target display actions corresponding to clothing design elements to different first display action subsets in descending order of priority of clothing design elements to obtain a second display action subset corresponding to each life scene.

[0102] This step is performed when the difference between the standard quantity value and the target number of display actions is greater than a preset threshold. The purpose is to make the total number of actions meet the standard by supplementing the display actions corresponding to high-priority design elements, ensuring that the display video duration is sufficient. Specifically, the clothing display system repeatedly adds the target display actions corresponding to each element to the first display action subset of different life scenes in order of clothing design element priority from high to low. After each addition, it is checked whether the actions in the same subset are repeated to ensure uniqueness. After the addition is completed, each life scene forms a second display action subset containing more actions to meet the video duration requirements.

[0103] Specific as Figure 4 The figure is a schematic diagram of an exemplary scene in which the clothing display system in the embodiment of the present application matches the display action for each life scene in the dynamic display video. For ease of understanding, Figure 4 Only 3 life scenes and 7 target display actions are used as examples to construct an embodiment for explanation. Figure 4 In the corresponding specific embodiment, the system first divides the target display actions according to the matching degree of each target display action (action 2, action 5, action 7, action 8, action 9, action 10 and action 11) with each life scene (scene 1, scene 2 and scene 3), and obtains one or more target display actions corresponding to each life scene. For example, if the matching degree of action 2 with scene 1, scene 2 and scene 3 is calculated to be 80%, 66% and 40% respectively, then action 2 is assigned to scene 1. Similarly, the life scenes corresponding to other target display actions can be divided to obtain the first display action subset corresponding to each life scene. For example, in Figure 4 In the corresponding embodiment, the first display action subset corresponding to scene 1 is {action 2, action 5, action 7}; the first display action subset corresponding to scene 2 is {action 8, action 9}; the first display action subset corresponding to scene 3 is {action 10, action 11}; then, if the difference between the number of actions in the target display action set (7) and the standard number (15) is greater than a preset threshold (such as 2), the target display action corresponding to the clothing design element with the highest priority in the customized clothing (the target display action needs to be an action in the target display action set, such as action 2) is first added to the other first display action subsets repeatedly to obtain the second display action subset. Figure 4 As shown, the second display action subset corresponding to scene 1 is {action 2, action 5, action 7}; the second display action subset corresponding to scene 2 is {action 2, action 8, action 9}; and the second display action subset corresponding to scene 3 is {action 2, action 10, action 11}.

[0104] If the difference between the number of target display actions in the updated second display action subset and the standard number value is still greater than the preset number threshold, the target display action corresponding to the clothing design element with the second priority will be used to continue updating and judging the second display action subset, which will not be repeated here.

[0105] S109: The first display action subset and the second display action subset corresponding to each life scene are the same.

[0106] Specifically, after the system calculates the difference and determines that there is no need to adjust the number of actions, it directly uses the first display action subset as the second display action subset for each life scene without performing any addition or deletion operations to ensure that the number of actions and video length meet the requirements.

[0107] S110 , based on the AI ​​model, convert multiple life scenes and the target display actions of the second display action subset corresponding to each life scene into text descriptions.

[0108] Specifically, the clothing display system inputs each life scene and its corresponding second display action subset into the AI ​​model. The model analyzes the action characteristics and scene requirements and generates a text description containing the execution sequence and time.

[0109] Optionally, the clothing display system divides the display priority of each target display action under different life scenes according to the matching degree between each target display action and the corresponding life scene (see step S104 for the specific calculation process of the matching degree). That is, the higher the matching degree between the target display action and the corresponding life scene, the higher the priority of the target display action in the corresponding life scene, thereby determining the display order of multiple target display actions in each life scene.

[0110] At the same time, the clothing display system matches the standard display duration corresponding to each target display action from the database, such as action A (10 seconds), action B (8 seconds), etc.

[0111] Next, the clothing display system inputs the display order of multiple target display actions in each life scenario and the standard display duration corresponding to each target display action into the AI ​​model, and outputs a text description containing the execution sequence and time.

[0112] Optionally, the AI ​​model of the present application can essentially be a pre-trained natural language processing (NLP) model. Specifically, the clothing display system can randomly select life scenes from the database. Then select multiple display actions whose matching degree with the life scenes is higher than a preset matching degree threshold, and construct a display action sequence in order of matching degree from large to small. Structured training sample data is constructed as model input based on life scenes, display action sequences and the standard display duration corresponding to each display action, and each set of input training sample data is equipped with a corresponding target text description as model output. By adopting a sequence-to-sequence natural language processing model (such as a Transformer-based neural network), the model is trained using supervised learning so that it can learn the mapping relationship between input structure and output text. After sufficient training, the AI ​​model can automatically generate logically clear and semantically accurate text descriptions based on any given life scenario, action sequence and action duration. For example: input "Scene: Leisure; Action: Turn, Walk, Waving; Duration: 3, 5, 2", and the output is "In the leisure scenario, the model first turns around for 3 seconds, then walks for 5 seconds, and finally waves for 2 seconds."

[0113] S111. Generate a dynamic display video of the customized clothing based on the text description.

[0114] This step is performed after obtaining the text description and is the last step of the entire process. Its purpose is to convert the text description into a visual dynamic video to show the user the effect of customized clothing.

[0115] Specifically, the clothing display system analyzes the timing and time information in the text descriptions, combines them with 3D clothing models and virtual models, and generates dynamic display videos. Based on the sequence and duration of actions in different scenes in the text, the system controls the virtual model to perform corresponding display actions, adjusting camera angles and switching timing to ensure a smooth video that fully displays all design elements.

[0116] Alternatively, the system can use professional 3D modeling software (such as Maya or Blender) to import the custom clothing's 3D model and avatar, set action keyframes based on text descriptions, adjust the execution time and sequence of each action, and then generate a video based on scene rendering. For example, set action A to execute between 0 and 10 seconds, and action B between 12 and 22 seconds, with camera switching in between.

[0117] Alternatively, the system uses a real-time rendering engine (such as Unity or Unreal Engine) to convert text descriptions into commands that the engine can recognize, driving the avatar's movements in real time. The engine automatically generates smooth transitions based on timing and time, adds lighting and material effects, and outputs the video.

[0118] Optionally, the system can also employ parametric video generation methods, converting the timing and time in text descriptions into video parameters to control camera movement, model movements, and lighting changes in the virtual scene, generating personalized presentation videos. For example, this could adjust the camera angle based on the text or focus on specific design elements during the execution of an action, though this is not limited here.

[0119] In the above embodiment, the system uses AI intelligent matching strategies to combine clothing design elements with display actions, selecting the display action with the smallest occlusion ratio, thereby maximizing the display of design element details. At the same time, by dividing the target display action into matching life scenes and using text descriptions to generate the execution sequence and time of coherent actions, a dynamic display video is finally generated. This method effectively addresses the problem of incomplete coverage of design elements due to time limitations in traditional dynamic displays, allowing the characteristics of clothing to be fully and efficiently displayed within a limited time.

[0120] In addition, after the clothing display system generates a dynamic display video and sends it to the user terminal for display, it can receive the user's modification instructions for customized clothing through the user interaction interface of the user terminal, and determine the modification parameter value of the clothing design element by parsing the specific modification content in the instruction and converting it into the corresponding design element parameters; then query the cumulative number of modifications made by the user to the current customized clothing, determine the preset parameter size according to the preset "number of modifications-preset parameter" mapping rule, add and subtract the preset parameter to the modification parameter value, and obtain the upper and lower limits of the parameter disturbance interval; then, based on the upper and lower limits of the parameter disturbance interval and the preset interval, generate a sequence of equally spaced candidate modification values ​​within the interval, and substitute each candidate modification value into the 3D model of the customized clothing. The model or design document is used to replace the parameters of the corresponding design elements to generate the corresponding candidate customized clothing; finally, all the modified and confirmed customized clothing data are extracted from the user's historical orders to build a historical clothing feature library, and the design feature vector is extracted for each candidate customized clothing and the similarity is calculated with the clothing in the historical feature library. The candidate customized clothing with the lowest average similarity is selected as the target customized clothing to generate a new dynamic display video, thereby realizing personalized optimization of the display of clothing design elements within a limited time, while ensuring the comprehensive display of core design elements, meeting the user's diverse needs for display effects, and improving the adaptability and actual effect of clothing display.

[0121] After determining the second display action subset corresponding to each life scene, it is necessary to determine the order of different life scenes in the dynamic display video and the execution order of the display actions corresponding to each life scene. Figure 2As shown, it is another flow chart of an AI-based customized clothing display method in an embodiment of the present application, which is used to determine the execution timing of life scenes and display actions in life scenes before converting life scenes and target display actions into text descriptions based on the AI ​​model.

[0122] S201: Obtain the action start feature and action end feature of each target display action in the second display action subset corresponding to each life scene.

[0123] The action start feature refers to the posture feature of the target at the beginning of the action, which is used to represent key information such as joint angles and body posture at the beginning of the action. The action end feature refers to the posture feature of the target at the end of the action, which is used to represent key information such as joint angles and body posture at the end of the action.

[0124] Specifically, the clothing display system extracts key posture features at the start and end of each target display action in the second display action subset corresponding to each life scenario. By analyzing the 3D model or video frames of the action, the system obtains information such as joint positions, angles, and body orientation at the start and end of the action, generating corresponding feature vectors as the start and end features of the action, respectively.

[0125] Optionally, the system uses a motion capture device to obtain three-dimensional motion data of the target's displayed movements, performs key point detection on the start and end frames of each movement, extracts features such as joint coordinates and rotation angles, and generates vector representations of movement start features and movement end features.

[0126] Optionally, the system pre-processes the video of the target demonstrating the action, extracts the image features of the start frame and the end frame through a convolutional neural network (CNN), and then combines it with a posture estimation algorithm to obtain the posture feature vectors at the beginning and end of the action.

[0127] S202: Calculate the first similarity between the action ending feature corresponding to each life scene and the action starting features corresponding to other life scenes.

[0128] Specifically, the clothing display system traverses all life scenes. For each scene, it extracts the ending features of all target display actions. It then calculates similarity with the starting features of all target display actions in other life scenes. Using appropriate similarity calculation methods (such as cosine similarity or Euclidean distance), the system determines the first similarity between each scene's ending features and the starting features of other scenes, forming a similarity matrix.

[0129] Optionally, the system can represent the action ending features of each life scene and the action starting features of other life scenes as multidimensional vectors, and use the cosine similarity formula to calculate the cosine value of the angle between the vectors as the first similarity. The larger the value, the higher the similarity.

[0130] Optionally, the system may also use Euclidean distance to calculate the distance between the action end feature vector and the action start feature vector, and use the reciprocal of the distance as the first similarity. The smaller the distance, the higher the similarity.

[0131] Optionally, the system may normalize the action features and use Manhattan distance to calculate the difference between feature vectors, taking the inverse of the difference as the first similarity. The smaller the difference, the higher the similarity.

[0132] S203: Determine the first life scene and the second life scene corresponding to the action start feature and the action end feature with the largest first similarity as two life scenes for continuous display in the dynamic display video.

[0133] Specifically, after obtaining the action feature similarity matrix between all life scenes, the clothing display system searches for the maximum first similarity value, finds the action start and end features corresponding to this maximum value, and then determines the first and second life scenes corresponding to these two features. The system identifies these two life scenes as scenes that need to be displayed consecutively in the dynamic display video, and stipulates that the second life scene should be displayed before the first life scene to utilize the high similarity of action features between the two to achieve a smooth scene transition.

[0134] Optionally, the system can organize the first similarity calculation results into a two-dimensional matrix, use a matrix search algorithm to find the position of the maximum value, determine the indexes of the corresponding first life scene and second life scene based on the position, and thus determine the scene pairs to be displayed continuously.

[0135] Optionally, the system can also convert the similarity matrix into a one-dimensional vector, find the element index corresponding to the maximum value through a sorting algorithm, and then map it to the corresponding life scene to determine the scene pairs to be displayed continuously and the display order.

[0136] S204: Determine the display order of different life scenes in the dynamic display video and the switching display actions corresponding to the different life scenes at the switching points based on the sorting result of the first similarity.

[0137] Specifically, the clothing display system prioritizes the most similar life scenes based on the first similarity ranking results, prioritizing them for a continuous display sequence. The system then gradually expands the sequence to include other scenes, forming a complete display sequence chain, with the scene pair with the highest first similarity as the core. Furthermore, a corresponding transition action is selected for each scene transition. This action is typically the target display action with the highest similarity between the ending features of the previous scene and the starting features of the next scene, ensuring a smooth transition.

[0138] Optionally, the system constructs life scenarios as nodes in a directed graph, using the first similarity as the edge weight. The system determines the display order by finding the maximum-weight path in the graph (e.g., the longest path algorithm), ensuring that the edge weights (similarity) between adjacent nodes (scenarios) are maximized. The switching action selects the target display action with the highest similarity in action features between the nodes at both ends of the edge.

[0139] Optionally, the system uses the minimum spanning tree (MST) algorithm to construct a scene connection graph, with similarity as the edge weight. The order of edges in the tree is the scene display order, and the action pairs corresponding to the tree edges are used as switching actions to ensure the optimal overall switching smoothness.

[0140] S205 , determining the execution sequence of other target display actions in each life scene based on the second similarity between the action ending feature of the target display action corresponding to the scene switching point and the other action starting features.

[0141] Specifically, for each life scene, the clothing display system uses the ending feature of the target display action at the transition between the current scene and the previous scene as a benchmark, and calculates the second similarity between this ending feature and the starting features of other target display actions in the current scene. The system arranges these actions in descending order of second similarity, forming an execution sequence within the scene and ensuring a smooth transition between the features of adjacent actions. If the scene is the first display scene, the system uses the starting feature of the first target display action in the scene as a benchmark, and calculates and ranks the second similarities with other actions.

[0142] Optionally, the system can start from the target display action at the scene switch (or the first action of the scene), and select the unexecuted action with the second highest similarity to the ending feature of the current action as the next action each time, until all actions are arranged to form an execution sequence.

[0143] Optionally, the system can also calculate the second similarity between the start features of all target display actions in the scene and the end features of the switching action, generate a similarity list, sort the actions from high to low by similarity, obtain the execution sequence, and ensure that the features of adjacent actions are the most similar.

[0144] Optionally, the system can also introduce an action priority correction factor. During greedy sorting, if high-similarity actions also have high design element priority, they will be prioritized to balance coherence and element importance and generate the final execution sequence.

[0145] In the above embodiment, the system calculates the similarity between the ending feature of each life scene and the starting feature of other scenes, determines the feature pair with the greatest similarity, and then arranges the corresponding life scenes in a continuous display order. Subsequently, the display order of different life scenes is determined based on the similarity sorting results, and the connection effect between different scenes is coordinated through the target display action at the scene switching. Within the scene, taking the target display action at the scene switching as the starting point, the execution timing of the target display action is dynamically adjusted by the similarity between the ending feature and the starting feature, thereby ensuring the smoothness of the content displayed within the scene.

[0146] S206. Based on the AI ​​model, multiple life scenes and the target display actions of the second display action subset corresponding to each life scene are converted into text descriptions, and a preliminary display video is generated based on the text descriptions.

[0147] This step is the same as step S110 and will not be repeated here.

[0148] S207 , counting one or more displayed angle intervals corresponding to each clothing design element in the preliminary display video to obtain a displayed angle interval set.

[0149] The displayed angle range refers to the direction angle range of the customized clothing relative to the virtual lens in the preliminary video, and is used to indicate the viewing angle range of the clothing that has been displayed in the video.

[0150] This step is performed after the initial presentation video is generated based on the AI ​​model. Its purpose is to count the displayed angles of each design element in the video, providing a basis for subsequent supplementary angles. Specifically, the clothing presentation system analyzes the initial presentation video frame by frame. For each clothing design element, it tracks its display in the video, detects changes in the virtual camera's orientation relative to the clothing, and records the angle range in which each element appears in the video. The system then merges multiple angle ranges for the same design element to form a set of displayed angle ranges for each design element.

[0151] Optionally, the system can perform frame sampling on the preliminary display video, extract key frames, use a posture estimation algorithm to detect the posture of the virtual model and the orientation of the clothing in each frame, calculate the direction angle of the clothing relative to the virtual lens, record the angle value of each design element in each key frame, and merge continuous angle intervals.

[0152] Optionally, the system can also pre-define a region of interest (ROI) for each clothing design element. During video playback, it tracks pixel changes within the ROI in real time, calculates the lens movement trajectory through the optical flow algorithm, converts it into an angle change interval, and summarizes it to obtain a set of displayed angles.

[0153] Optionally, the system can also use 3D reconstruction technology to reconstruct the three-dimensional model movement trajectory of the clothing based on multiple frames of images in the video, calculate the rotation angle range of the virtual lens in the three-dimensional space, and correspond to the displayed angle range of each design element.

[0154] S208: Determine one or more non-displayed angle intervals based on the preset display angle interval set and the displayed angle interval set to obtain a non-displayed angle interval set.

[0155] The undisplayed angle interval refers to the angle range in the preset set that is not covered by the displayed set and needs to be supplemented and displayed in subsequent steps.

[0156] Specifically, the clothing display system compares a preset set of display angle intervals with a set of displayed angle intervals. Using set operations (such as set difference), it identifies angle intervals in the preset set that are not covered by the displayed set. The system then merges and removes duplicates from these uncovered angle intervals, ultimately forming a set of undisplayed angle intervals, ensuring that every undisplayed angle is clearly identified.

[0157] Optionally, the system represents both the preset display angle interval and the displayed angle interval as an ordered list of angle ranges, sorts them by angle size, compares the interval overlaps one by one, and calculates the portion of the preset interval that is not covered by the displayed interval as the non-displayed angle interval.

[0158] S209: Determine the target clothing design element with the highest priority corresponding to each undisplayed angle interval and the target display action corresponding to the target clothing design element in the preliminary display video.

[0159] Specifically, for each undisplayed angle interval, the clothing display system determines the design element most important to display at that angle (i.e., the element with the highest priority) based on pre-set design element priority rules. The system then searches for the target display action corresponding to that design element in the initial display video, ensuring that the same action is used when displaying the supplementary angle interval to maintain consistency and coherence.

[0160] S210 , without changing the execution time of the target display action, by continuously changing the direction angle of the virtual lens, adding a non-displayed angle interval on the basis of the displayed angle interval, to obtain a dynamic display video.

[0161] The virtual lens refers to the virtual camera used to shoot custom clothing in a 3D virtual scene, controlling the viewing angle. The direction angle refers to the direction of the virtual lens relative to the custom clothing. Adjusting this angle allows you to show different sides of the clothing.

[0162] This step is performed after determining the undisplayed angle intervals and the corresponding design elements and actions. The purpose is to supplement the undisplayed angles by adjusting the virtual lens angle without changing the original action time. Specifically, for each undisplayed angle interval, during the execution of the corresponding target display action, the clothing display system controls the virtual lens to smoothly transition from the displayed angle to the undisplayed angle, thereby expanding the angle interval. The system ensures that the change in lens angle is continuous and does not change the execution timing and duration of the target display action. Ultimately, all supplemented angle intervals are integrated into the preliminary video to generate a complete dynamic display video.

[0163] Optionally, the system can determine the starting and ending angle key frames of the undisplayed angle interval within the execution time period of the target display action, insert the key frames of the undisplayed angle based on the key frames of the displayed angle, calculate the lens angle of the intermediate frames through linear interpolation, and generate a continuous lens motion trajectory.

[0164] Optionally, the system can also analyze the movement characteristics of the target display action, such as movement amplitude and rhythm, and adjust the lens movement speed according to the undisplayed angle interval to ensure that the lens angle change matches the movement rhythm and complete the angle expansion without changing the action execution time.

[0165] Optionally, the system may also split the undisplayed angle interval into multiple sub-intervals, and insert lens angle changes of the sub-intervals at different stages of the target display action, ensuring that the addition of each sub-interval does not affect the normal execution time of the action.

[0166] In the above-described embodiment, the system counts the set of displayed angle intervals in the preliminary display video and determines the set of undisplayed angle intervals by comparing it with a preset set of displayed angle intervals. Then, based on the highest-priority design element in the undisplayed angle interval set and its corresponding target display action, the virtual lens's direction angle is dynamically adjusted, gradually replenishing the undisplayed angle intervals without changing the execution time. While maintaining the limited duration of the preliminary video, this technology further improves the angle coverage of the dynamic display video, resolving the issues of single angles and incomplete design element expression in traditional display methods and enhancing the comprehensiveness of the displayed content.

[0167] The clothing display system of the embodiment of the present invention is applied to electronic equipment. Figure 5 A schematic diagram of the architecture of an electronic device suitable for implementing an embodiment of the present invention is shown.

[0168] It should be noted that Figure 5 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention.

[0169] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments can be accomplished by instructions (computer programs) or by controlling related hardware through instructions (computer programs), and the instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. The electronic device of this embodiment includes a storage medium and a processor, wherein the storage medium stores a plurality of instructions, which can be loaded by the processor to execute any step of the method provided in the embodiment of the present invention.

[0170] Specifically, the storage medium and the processor are electrically connected directly or indirectly to achieve data transmission or interaction. For example, these elements can be electrically connected to each other via one or more signal lines. The storage medium stores computer-executable instructions for implementing the data access control method, including at least one software function module that can be stored in the storage medium in the form of software or firmware. The processor executes various functional applications and data processing by running the software programs and modules stored in the storage medium. The storage medium can be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc. The storage medium is used to store programs, and the processor executes the programs after receiving the execution instructions.

[0171] Furthermore, the software programs and modules in the above-mentioned storage medium may also include an operating system, which may include various software components and / or drivers for managing system tasks (such as memory management, storage device control, power management, etc.), and may communicate with various hardware or software components to provide an operating environment for other software components. The processor may be an integrated circuit chip having signal processing capabilities. The above-mentioned processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc., which may implement or execute the various methods, steps, and logic flow diagrams disclosed in this embodiment. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0172] Since the instructions stored in the storage medium can execute the steps of any method provided in the embodiments of the present invention, the beneficial effects of any method provided in the embodiments of the present invention can be achieved. Please refer to the previous embodiments for details and will not be repeated here.

[0173] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. An AI-based customized clothing display method, applied to a clothing display system, characterized in that: The method comprises: Obtaining customization requirements corresponding to customized clothing and multiple clothing design elements included in the customized clothing, wherein the customization requirements correspond to a preset number of life scenes; Matching one or more display actions corresponding to each clothing design element from a preset database to obtain a display action set corresponding to each clothing design element, wherein the display action has the smallest occlusion ratio on the clothing design element compared to other actions in the preset database; Selecting a target display action with the highest display action repetition rate among other display action sets from each display action set to obtain a target display action set; Dividing each target display action in the target display action set into a corresponding life scene according to the matching degree, to obtain a first display action subset corresponding to each life scene; If the difference between the standard quantity value and the quantity value corresponding to the target display action is greater than a preset quantity threshold, the target display actions corresponding to the clothing design elements are repeatedly added to different first display action subsets in descending order of priority of the clothing design elements to obtain a second display action subset corresponding to each life scene, the standard quantity value is positively correlated with the limited duration of the display video, and the target display actions in the same second display action subset are different from each other; Based on the AI ​​model, the plurality of life scenes and the target display actions of the second display action subset corresponding to each life scene are converted into text descriptions, wherein the text description includes the execution sequence and execution time of each target display action in each life scene; A dynamic display video of the customized clothing is generated based on the text description.

2. The method according to claim 1, characterized in that After the step of dividing each target display action in the target display action set into a corresponding life scene according to the matching degree to obtain a first display action subset corresponding to each life scene, the method further includes: If the difference between the standard quantity value and the quantity value corresponding to the target display action is less than or equal to the preset quantity threshold, the first display action subset and the second display action subset corresponding to each of the life scenes are the same; If the target display action corresponding quantity value is greater than the standard quantity value, the clothing design elements are reduced in order of priority from low to high until the target display action corresponding quantity value is less than or equal to the standard quantity value.

3. The method according to claim 1, characterized in that The step of repeatedly adding the target display actions corresponding to the clothing design elements to different first display action subsets in descending order of priority of the clothing design elements to obtain a second display action subset corresponding to each of the life scenes specifically includes: Repeatedly adding the first target display action corresponding to the clothing design element with the highest priority to different first display action subsets to obtain a second display action subset corresponding to each of the life scenes; Calculate the preliminary video duration according to the execution sequence and execution time corresponding to each target display action in the second display action subset in the text description; If the preliminary video duration is less than the limited duration, the second target display action corresponding to the clothing design element with the second highest priority is repeatedly added to different first display action subsets and the new preliminary video duration is calculated again for judgment.

4. The method according to claim 1, wherein The steps of positively correlating the number of standards with the limited duration of displaying the video specifically include: Obtaining a standard display duration of each display action in the target display action set and an average duration of all display actions; Calculating the first action quantity according to the difference between the limit duration and the standard total duration of the target display action set and the average duration; The standard number is obtained by adding the first number of actions and the second number of actions corresponding to the target display action set.

5. The method according to claim 1, wherein Before the step of converting the plurality of life scenes and the target display action of the second display action subset corresponding to each life scene into a text description based on the AI ​​model, the method further includes: Obtaining an action start feature and an action end feature of each target display action in the second display action subset corresponding to each life scene; Calculating a first similarity between the action ending feature corresponding to each of the life scenes and the action starting features corresponding to other life scenes; Determining the first life scene and the second life scene corresponding to the action start feature and the action end feature with the greatest first similarity as two life scenes for continuous display in the dynamic display video, wherein the second life scene in the dynamic display video is displayed before the first life scene; Determining, based on the sorting result of the first similarity, the display order of different life scenes in the dynamic display video and the switching display actions corresponding to the different life scenes at the switching points; In each life scene, the execution sequence of other target display actions is determined based on the second similarity between the action ending feature of the target display action corresponding to the scene switching point and the other action starting features.

6. The method according to claim 1, characterized in that The step of generating a dynamic display video of the customized clothing based on the text description specifically includes: generating a preliminary demonstration video based on the text description; Counting one or more displayed angle intervals corresponding to each clothing design element in the preliminary display video to obtain a displayed angle interval set; determining one or more non-displayed angle intervals based on a preset display angle interval set and the displayed angle interval set to obtain a non-displayed angle interval set, wherein the preset display angle interval set includes all display angle intervals of the custom garment, and the display angle intervals are direction angles of the custom garment relative to the virtual camera lens in the preliminary video; Determine the target clothing design element with the highest priority corresponding to each undisplayed angle interval and the target display action corresponding to the target clothing design element in the preliminary display video; Under the premise of not changing the execution time of the target display action, the non-displayed angle interval is added to the displayed angle interval by continuously changing the direction angle of the virtual lens to obtain a dynamic display video.

7. The method according to claim 1, characterized in that After the step of generating a dynamic display video of the customized clothing based on the text description, the method further includes: Determining modification parameter values ​​of clothing design elements according to a user's modification operation on the customized clothing; The modified parameter value is added to and subtracted from a preset parameter to obtain an upper limit and a lower limit of the parameter disturbance interval, wherein the value of the preset parameter is positively correlated with the number of times the user modifies the customized garment; generating corresponding candidate customized garments respectively from a plurality of sets of candidate modification values ​​generated at preset intervals within the parameter disturbance interval, the plurality of sets of candidate modification values ​​including the modification parameter values; The candidate customized clothing with the lowest similarity to the customized clothing modified by the user history among the plurality of candidate customized clothing is determined as the target customized clothing, and the target customized clothing is used to generate a new dynamic display video.

8. A clothing display system, characterized in that: The clothing display system includes: one or more processors and memories; The memory is coupled to the one or more processors, and is configured to store computer program codes, wherein the computer program codes include computer instructions. The one or more processors call the computer instructions to enable the clothing display system to execute the method according to any one of claims 1 to 7.

9. A computer-readable storage medium comprising instructions, characterized in that: When the instruction is executed on a clothing display system, the clothing display system is caused to execute the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product is run on a clothing display system, the clothing display system is caused to execute the method according to any one of claims 1 to 7.

Images