Customized clothing display method based on artificial intelligence

Abstract

The invention relates to the technical field of artificial intelligence clothing display, and discloses a customized clothing display method based on artificial intelligence. The method comprises the step of generating a personalized clothing model by obtaining body measurement data of a customer and customization requirements including design styles and display time. The best display speed is intelligently calculated by evaluating the durability measurement of model display and combining with a display time range set by a user, so that prospective optimization of the display performance is realized; and deploying a real-time monitoring system, dynamically detecting model state deviation in a rendering process, and automatically triggering a parameter adjustment mechanism. According to the scheme, the problems of display lagging and unstable frame rate caused by a complex structure and system resource fluctuation of a customized garment three-dimensional model are solved, the smoothness and stability of visual output in the whole specified display period are ensured through dynamic speed optimization and real-time adaptive adjustment, and the real-time performance and reality sense of virtual fitting are improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence clothing display technology, specifically to a method for displaying customized clothing based on artificial intelligence. Background Technology

[0002] Current online custom clothing display technologies typically rely on pre-built, generic 3D clothing models. After a customer inputs their body measurements, the system uses simple linear scaling or deformation algorithms to adapt the basic clothing model to the customer's body shape. The core of this display method lies in the static accuracy of model generation, while subsequent display processes are often simplified. Display systems generally use fixed rendering parameters and animation speeds, and their configuration depends on the baseline performance of the hardware, without considering the complexity of specific clothing models or their actual resource consumption over long periods. Existing solutions lack forward-looking assessments of model performance throughout the entire display cycle, making it difficult to guarantee a stable and smooth experience from start to finish.

[0003] Existing technology has shortcomings. Fixed display settings cannot adapt to the varying performance requirements of personalized models with different complexities. Displaying a complex clothing model customized for a specific body type versus a simple model under the same settings will result in the former experiencing stuttering and latency, while the latter fails to fully utilize hardware performance for optimal performance. A more significant problem is the system's inability to handle dynamic changes during the display process. When the model begins display, if hardware temperature rises leading to frequency reduction, or if other background programs consume resources, the display quality will unpredictably decline, resulting in frame rate drops or model rendering errors. This can only be temporarily alleviated by manually refreshing or reloading, severely disrupting the continuity and immersion of the user experience.

[0004] Predicting display performance based on the unique attributes of each customized garment model and intelligently optimizing display parameters accordingly is a problem to be solved. Meanwhile, maintaining stable display quality in a dynamically changing system environment and achieving automatic fault tolerance and optimization without manual intervention are key challenges in improving the quality of online customized clothing services. Summary of the Invention

[0005] The purpose of this invention is to provide a customized clothing display method based on artificial intelligence to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a customized clothing display method based on artificial intelligence, the method comprising: Obtain body measurement data and clothing customization requirements submitted by customers. The body measurement data includes detailed size information and body type identification, while the clothing customization requirements include design style requirements and display time range. Standardize body measurement data by standardizing its format and correcting outliers. Select a basic clothing model using standardized body data and design style requirements; The integrated body type category identifier is used to adjust the outline of the basic clothing model to create a personalized clothing model; Configure display environment parameters, including rendering precision and refresh rate limits; Based on the personalized clothing model and display environment parameters, evaluate the durability measure of the model display; By combining persistence metrics and display time range, the optimal display speed can be calculated; Deploy a real-time monitoring system to observe changes in the model's state during the display process; When the model state changes beyond the allowable deviation, the automatic adjustment process is activated; The automatic adjustment process modifies display settings to maintain smooth display performance.

[0007] Preferably, the method involves standardizing the format of body measurement data and correcting outliers to generate standardized body data, the method comprising: The system analyzes detailed dimensional information from body measurement data, extracts multiple dimensional values, detects outliers among these values, corrects them using interpolation, converts the corrected dimensional values ​​into standard units of measurement, normalizes the converted dimensional values ​​according to body type category identifiers, verifies the integrity of the normalized data, and outputs standardized body data.

[0008] Preferably, a basic clothing model is selected using standardized body data and design style requirements. The method includes: Load the pre-stored basic clothing model library, which contains multiple clothing styles and corresponding size ranges; The matching degree of standardized body data with the size range of multiple clothing styles is calculated. The matching degree calculation results are filtered according to the design style requirements to select candidate clothing models. An artificial intelligence matching mechanism is applied to prioritize the candidate clothing models. The base clothing model with the highest matching degree is selected based on the priority ranking results.

[0009] Preferably, the method involves adjusting the outline of a basic clothing model using integrated body type category identifiers to create a personalized clothing model, the method comprising: Identify the body shape feature parameters corresponding to the body shape category identifier, obtain the geometric structure data of the basic clothing model, map the body shape feature parameters to the geometric structure data, and perform vertex displacement calculation; Adjust the mesh shape of the basic clothing model based on the vertex displacement calculation results; A texture-fitting algorithm is applied to the surface of the adjusted mesh shape to generate a personalized clothing model.

[0010] Preferably, the method for configuring display environment parameters, including rendering precision and refresh rate limits, includes: The rendering accuracy level is determined based on the complexity of the personalized clothing model, and the refresh rate limit threshold is set according to the display time range. Query system hardware capability data, adjust rendering precision level and refresh rate limit threshold, generate display environment parameter configuration file, and verify the compatibility of display environment parameter configuration file with personalized clothing model.

[0011] Preferably, the method for evaluating the durability of the model display based on personalized clothing models and display environment parameters includes: The rendering process of a personalized clothing model under display environment parameters is simulated. Frame rate stability and image clarity data are collected during the simulation. The decay trend of frame rate stability and image clarity data is calculated. Combining the decay trend and refresh rate limit threshold, persistence metric values ​​are derived and stored for subsequent optimization.

[0012] Preferably, the method for calculating the optimal display speed by combining persistence metrics and display time range includes: Establish a correlation model between persistence metric and display speed. Input the display time range into the correlation model to obtain multiple candidate display speeds. Test the impact of multiple candidate display speeds on the personalized clothing model. Select the candidate display speed with the least impact as the optimal display speed. Calibrate the matching degree between the optimal display speed and the system hardware capability data.

[0013] Preferably, a real-time monitoring system is deployed to observe changes in the model's state during the display process. The method includes: Initialize the sensor data stream, capture the model rendering data during display, analyze the abnormal fluctuation patterns in the model rendering data, and compare the abnormal fluctuation patterns with the allowable deviation threshold. When the abnormal fluctuation pattern continues to exceed the allowable deviation threshold, mark the model state change event and record the timestamp and severity of the model state change event.

[0014] Preferably, when the model state change exceeds the allowable deviation, an automatic adjustment process is activated, the method comprising: Retrieve relevant data on model state change events; The type of adjustment is determined based on relevant data, including parameter fine-tuning or model regeneration; Execute the operation sequence corresponding to the adjustment type, verify whether the adjusted model state has returned to the normal range, and if it has not returned to the normal range, iterate the adjustment process until the condition is met.

[0015] Preferably, the method of automatically adjusting display settings to maintain display smoothness includes: Adjust the rendering precision and refresh rate limits in the display environment parameters, update the display attributes of the personalized clothing model, synchronously modify the optimal display speed to adapt to the new settings, monitor the adjusted display smoothness indicators, and dynamically optimize the triggering conditions of the automatic adjustment process based on the monitoring results.

[0016] Compared with the prior art, the beneficial effects of the present invention are: By evaluating the persistence metrics of model display and calculating the optimal display speed, the display settings are transformed from static configuration to dynamic optimization. This technology analyzes the relationship between the geometric complexity and texture information of the personalized clothing model and the rendering accuracy and refresh rate limitations of the current display environment to generate a quantitative metric to predict the model's performance under continuous operation. Combined with the user-specified display time range, the system can reverse-calculate the optimal animation or rotation speed to maintain stable output within the given time. This approach makes the display speed no longer a fixed value, but a dynamic variable closely related to the model's own characteristics, the environment's carrying capacity, and the display duration. It avoids the phenomenon of smooth initial display followed by stuttering due to accumulated resource consumption, ensuring consistent and smooth visual output throughout the entire expected display period, and achieving a fine match between resource allocation and display requirements.

[0017] Deploying a real-time monitoring system and activating the automatic adjustment process introduces a closed-loop feedback control mechanism to the display system. This technology continuously tracks key performance indicators such as frame generation time, memory usage, and model mesh deformation errors during the display process. Once the monitored data deviates from the preset stable range, the system immediately triggers adjustment logic instead of waiting for user complaints. The adjustment process dynamically modifies underlying settings such as rendering detail levels, shadow quality, or physical simulation accuracy. This real-time response mechanism transforms the display process from an open-loop, passive playback state to a closed-loop, proactive maintenance state. It effectively offsets instantaneous performance degradation caused by system background tasks, hardware performance fluctuations, or driver compatibility issues, promptly corrects model display anomalies, and restores smoothness before users perceive stuttering or distortion, improving the robustness of the display system and the reliability of the user experience. Attached Figure Description

[0018] Figure 1 This is a schematic diagram illustrating the working principle of the AI-based customized clothing display method described in this invention. Figure 2 A flowchart for standardizing body data processing; Figure 3 Flowchart generated for personalized clothing models; Figure 4 Multi-dimensional comparison chart of the effects of adjusting clothing models for different body types; Figure 5 Comparison chart showing the effects of adjusting strategies for different load scenarios. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 This invention provides an AI-based method for showcasing customized clothing. The method includes: acquiring body measurement data and clothing customization requirements submitted by the customer; the body measurement data includes detailed size information and body type category identifiers; and the clothing customization requirements include design style requirements and a display time range. The acquisition process is completed through a user interface or network interface, and the data is stored in a structured format for easy subsequent processing. Subsequently, the body measurement data is standardized in format and outlier corrections are performed to generate standardized body data, ensuring data consistency and reliability. Using the standardized body data and design style requirements, the system selects a basic clothing model and filters the most suitable style from a pre-stored model library using a matching algorithm. The basic clothing model is then adjusted in silhouette using integrated body type category identifiers to form a personalized clothing model. This adjustment involves modifying the geometric structure to fit the customer's body characteristics. Display environment parameters are configured, including rendering accuracy and refresh rate limits. These parameters are optimized based on system hardware capabilities and display requirements. Based on the personalized clothing model and display environment parameters, a persistence metric for model display is evaluated, reflecting the model's stability during long-term display. Combining the persistence metric and the display time range, the optimal display speed is calculated to ensure a smooth and efficient display process. A real-time monitoring system is deployed to observe changes in the model's state during the demonstration. Monitoring data includes frame rate and image quality metrics. When changes in the model's state exceed allowable deviations, an automatic adjustment process is activated. This system detects anomalies in real time and triggers a correction mechanism. The automatic adjustment process modifies the demonstration settings to maintain smoothness. Adjustments involve dynamic parameter updates and model regeneration to cope with changing conditions.

[0021] Example 1: See Figure 2In its implementation, the AI-based customized clothing display method involves standardizing body measurement data through format unification and outlier correction to generate standardized body data. This data includes detailed size information and body type category identifiers. The system receives body measurement data submitted by clients via a data interface, with various formats such as JSON, XML, or CSV. When parsing the detailed size information, parsing algorithms are used to identify key fields such as chest circumference, waist circumference, and height. Multiple size values ​​are extracted and stored in a temporary buffer. Outliers among these values ​​are detected using statistical methods such as the Z-score algorithm to calculate the deviation of each value from the dataset. Values ​​exceeding a preset threshold are marked as outliers. When correcting outliers using interpolation, a linear interpolation algorithm is applied based on adjacent data points to fill in the gaps. During outlier correction, the system first iterates through the size value sequence using a sliding window, using the median within the window as a baseline value. Values ​​deviating from the baseline value are then marked as data to be corrected. For regions with multiple consecutive data points to be corrected, historical normal size intervals within the same body type category are retrieved for interval verification. Once verification is successful, interpolation correction is performed to ensure that the corrected data conforms to the physiological size logic of the corresponding body type. The corrected dimensions are converted to standard units of measurement using a unit conversion module, such as converting inches to centimeters or pounds to kilograms. The converted dimensions are then normalized based on body type identifiers using a min-max scaling algorithm to map the values ​​to the 0-1 range. The integrity of the normalized data is verified by checking for missing values ​​or logical conflicts, such as waist circumference being greater than chest circumference. Standardized body data is then output and stored in a structured format in a database. In some embodiments, the parsing process uses regular expressions to match the dimension field, extracting the values ​​and verifying data type consistency. It is understood that the normalization process relies on predefined body type parameters to ensure data suitability for subsequent model selection.

[0022] In practice, a basic clothing model is selected using standardized body data and design style requirements. Design style requirements include attributes such as color, pattern, or cut type. A pre-stored basic clothing model library is loaded, and model files are read from the server or local storage. The basic clothing model library contains multiple clothing styles and corresponding size ranges. Style information is associated with size ranges in the form of metadata. The standardized body data is matched with the size ranges of multiple clothing styles. The Euclidean distance formula is applied to compare the differences between the customer's size and the model size. The matching degree calculation results generate a similarity score list. The matching degree calculation results are filtered according to the design style requirements. A rule engine is used to exclude models that do not conform to the style, such as filtering formal styles for casual styles. Candidate clothing models are selected based on score thresholds, and the top-k candidates are selected. An artificial intelligence matching mechanism is applied to prioritize the candidate clothing models. A neural network model is used to analyze historical selection data to optimize the ranking. Based on the priority ranking results, the basic clothing model with the highest matching degree is selected, and the model identifier is output.

[0023] In practice, the format standardization and outlier correction of body measurement data are further refined. When parsing body measurement data, multi-source data such as manual input or sensor acquisition are processed. After extracting multiple size values, data type verification is performed to ensure that the data is numerical. Outliers are detected by using box plots to identify values ​​outside the upper and lower quartiles. When using interpolation methods for correction, spline interpolation algorithms are applied to smooth the data curve for consecutive missing values. The corrected multiple size values ​​are converted to standard units of measurement and batch processing is achieved by querying a unit conversion table. The converted size values ​​are normalized according to the body type category identifier, and the data distribution is adjusted using the Z-score normalization algorithm. The integrity of the normalized data is verified by performing integrity constraint checks such as non-empty verification and range verification. The standardized body data is output as a serialized JSON format for subsequent use. The basic clothing model selection process involves multiple stages of screening. When loading the pre-stored basic clothing model library, the index model files are accessed to accelerate the process. The matching degree between standardized body data and the size range of multiple clothing styles is calculated using a multi-dimensional weighted distance formula to balance the importance of different sizes. The matching degree calculation results are filtered according to the design style requirements, and Boolean logic expressions are used to evaluate the style conformity. Candidate clothing models are selected to generate a candidate set and intermediate results are cached. An artificial intelligence matching mechanism is applied to prioritize the candidate clothing models and train a decision tree model based on feature importance scoring. Based on the priority ranking results, the basic clothing model with the highest matching degree is selected, and the selection log is recorded.

[0024] In practical implementation, body measurement data processing and model selection work in tandem. After parsing the body measurement data, an anomaly detection pipeline is triggered. When extracting multiple dimensional values, array or list data structures are processed. Outliers among the multiple dimensional values ​​are detected using density clustering algorithms such as DBSCAN to identify isolated points. When using interpolation methods for correction, Lagrange interpolation is applied to marginal values ​​to maintain data continuity. The corrected multiple dimensional values ​​are converted to standard units of measurement and integrated into the International System of Units (SI). Based on body type identification, the converted dimensional values ​​are normalized using decimal scaling to simplify calculations. The integrity of the normalized data is verified using hash verification to check data consistency. After outputting standardized body data, a ready event is published. In practical implementation, outlier detection can be combined with time series analysis to process dynamic measurement data. The basic clothing model selection optimizes matching accuracy. When loading the pre-stored basic clothing model library, a caching mechanism is used to reduce IO latency. The matching degree of standardized body data with the size range of multiple clothing styles is calculated, and Mahalanobis distance is applied to handle relevant size dimensions. The matching degree calculation results are filtered according to design style requirements, and natural language processing is integrated to parse style descriptions. Candidate clothing models are screened, and a heap data structure is used to maintain the candidate queue. An artificial intelligence matching mechanism is applied to prioritize the candidate clothing models, and a support vector machine model is deployed to classify the candidate models. Based on the priority ranking results, the basic clothing model with the highest matching degree is selected, and the model usability is verified.

[0025] Example 2: See Figure 3 In practical implementation, the system integrates body type category identifiers to adjust the outline of the basic clothing model to form a personalized clothing model. The system receives body type category identifiers and basic clothing model data from the upstream process. Body type category identifiers are string codes, such as "APPLE" or "PEAR," used to indicate the customer's body type classification. Identifying the body type feature parameters corresponding to the body type category identifier is accomplished by querying a predefined body type feature parameter mapping table. This mapping table is stored in key-value pairs, where the key is the body type category identifier and the value is a set of floating-point body type feature parameters, such as shoulder width ratio, waist-to-hip ratio, or torso length coefficient. The system obtains the geometric structure data of the basic clothing model by reading vertex coordinates, triangle patch indices, normal vectors, and texture coordinates from the model file. Geometric structure data is typically organized in memory as an array. Mapping body type feature parameters to geometric structure data involves constructing a parameterized deformation function. This function takes the body type feature parameters as input and outputs a set of geometric transformation instructions. Vertex displacement calculation is the core step. The calculation is based on a deformation model grounded in physics, and the three-dimensional displacement of each vertex is determined by its mesh region and related body type feature parameters. Vertex displacement calculation can be quantitatively described using the following formula: ; Where: symbol Represents the three-dimensional displacement vector of the i-th vertex, denoted by . Represents the total number of body shape characteristic parameters, symbol This represents the weighting factor by which the i-th vertex is affected by the j-th body shape feature parameter. The weighting factor is calculated based on the distance between the vertex and the key anatomical points of the model. This represents the specific value of the j-th parameter obtained from the body shape feature parameter mapping table, with the symbol... This represents a 4x4 affine transformation matrix, generated by the overall scaling and rotation rules corresponding to the body type identifier. The symbol... This represents the coordinate vector of the i-th vertex in its local coordinate system.

[0026] In practice, the mesh shape of the basic clothing model is adjusted based on the vertex displacement calculation results. The system creates a new mesh container, which combines the original vertex coordinates with the calculated displacement vectors. Vertex-by-vertex addition generates a new array of vertex positions. This addition operation directly modifies the 3D spatial coordinates of the vertices. The adjustment process must maintain the topological connectivity of the mesh, i.e., the triangular facet index array must remain unchanged, to ensure the integrity of the mesh model. In practice, a texture bonding algorithm is applied to surface-process the adjusted mesh shape. The texture bonding algorithm mainly addresses the texture stretching or distortion problems that occur after mesh deformation. The algorithm input includes the adjusted new mesh geometry data and the original texture image and UV mapping coordinates of the base clothing model. The texture bonding algorithm first detects UV coordinate distortion by calculating the change in the area ratio or side length ratio of each triangular facet in the texture space (UV space) and the 3D model space to quantify the degree of distortion. For 3D model facests with distortion exceeding a threshold, the algorithm performs corresponding local corrections in the texture image space. This correction includes smooth repositioning of the affected UV coordinates or non-uniform scaling of specific regions of the texture image, with the goal of minimizing visual texture distortion. Generating a personalized clothing model means packaging the final surface-processed mesh geometry data, the corrected UV coordinates, and the texture image resources into a complete 3D model asset.

[0027] In practice, identifying the body type category identifier and its corresponding body feature parameters is a data retrieval and interpretation process. The system parses the received body type category identifier string and performs an exact match query in the body feature parameter database. The database records contain a set of predefined anatomical proportion parameters related to a specific body type category. Obtaining the geometric structure data of the basic clothing model typically means deserializing graphics data such as vertex buffers and index buffers from persistent storage to the application's working memory. Mapping body feature parameters to geometric structure data requires establishing association rules between the parameter space and the vertex geometry space, such as determining which body feature parameter primarily affects the vertices of the torso region and which affects the vertices of the sleeve region. Vertex displacement calculation is a computationally intensive step, requiring traversal of the entire vertex array and application of formulas for calculation. This calculation process can utilize parallel computing frameworks to improve efficiency.

[0028] In practical implementation, after adjusting the mesh shape of the basic clothing model based on the vertex displacement calculation results, the mesh needs further smoothing to avoid unnatural sharp edges. The system can apply Laplacian smoothing or hyperbolic smoothing algorithms to fine-tune the vertex positions after displacement. The smoothing algorithm optimizes local curvature while maintaining the overall shape change. When applying texture bonding algorithms to perform surface processing on the adjusted mesh shape, in addition to handling texture stretching, it also includes the process of recalculating normal vectors. Because vertex displacement changes the orientation of the model surface, the system needs to recalculate the normal vector of each vertex based on the new mesh geometry data to ensure the correctness of lighting rendering. The generated personalized clothing model is the output of this embodiment, and its data format needs to be compatible with downstream rendering engines, usually exported in a common 3D file format.

[0029] Example 3: In specific implementation, the configuration of display environment parameters includes rendering precision and refresh rate limits. The system receives personalized clothing model data generated from the upstream process. The personalized clothing model data includes the number of mesh vertices, texture map resolution, and material complexity information. The rendering precision level is determined based on the complexity of the personalized clothing model. The complexity is calculated using a quantization function. The function inputs are the number of vertices, texture size, and number of shader instructions, and the output is an integer rendering precision level, categorized as, for example, L0 (low), L1 (medium), and L2 (high). A refresh rate limit threshold is set based on the display time range, which is a time interval specifying the duration of the display. The refresh rate limit threshold is set to the maximum allowed frame rate, such as 60 frames per second or 30 frames per second, to prevent hardware overheating or power consumption exceeding limits. System hardware capability data is queried by calling the system information interface to obtain the currently available GPU memory capacity, CPU core frequency, and system free memory size. Adjusting the rendering precision level and refresh rate limit threshold is a dynamic process. The calculated initial rendering precision level is compared with system hardware capabilities. If the hardware capabilities are insufficient to support the initial level, the rendering precision level is progressively reduced until it matches the desired level. Simultaneously, the refresh rate limit threshold is scaled proportionally to the length of the display time range. When querying system hardware capabilities, the system obtains GPU memory bandwidth, CPU single-core load, and memory read / write latency data through a dedicated hardware monitoring interface, while also recording hardware resource fluctuation curves over the past ten minutes. When adjusting the rendering precision level, priority is given to reducing the rendering precision of non-core visual layers, such as first lowering the shadow map resolution and ambient occlusion sampling rate, and then adjusting the model's mesh subdivision level to ensure that the visual presentation of the clothing is not significantly affected. A configuration file for generating display environment parameters serializes the final determined rendering precision level, refresh rate limit threshold, and other relevant parameters such as anisotropic filtering level and shadow quality into a structured configuration file. The file format can be JSON or YAML. Verify the compatibility between the display environment parameter configuration file and the personalized clothing model by starting a lightweight rendering verification process. Load the personalized clothing model into memory and apply the parameters in the configuration file to perform a single-frame rendering test to detect whether there are model display errors, missing textures, or performance bottlenecks.

[0030] In practical implementation, the persistence metric of the model's display is evaluated based on the personalized clothing model and display environment parameters. The rendering process of the personalized clothing model under the display environment parameters is simulated in an isolated rendering sandbox environment, which simulates the graphics API behavior of the target hardware. Frame rate stability and image sharpness data are collected during the simulation. Frame rate stability is evaluated by recording the time taken for consecutive rendering frames and calculating their standard deviation and mean. Image sharpness data is quantified by sampling the output image after rendering and calculating its peak signal-to-noise ratio (PSNR) compared to the ideal reference image. The decay trend of frame rate stability and image sharpness data is calculated using time series analysis. Linear regression fitting is performed on the frame rate stability and image sharpness data sequences recorded in time windows during the simulation, and the slope of the fitted line is the decay rate. The persistence metric is derived by combining the decay trend and refresh rate limit threshold. The persistence metric is a comprehensive indicator used to predict the maximum duration for which the personalized clothing model can maintain acceptable display quality under given display environment parameters. The following formula can be used to estimate: ; Where: symbol The calculated persistence metric value is represented by the symbol. This represents the minimum allowable threshold for frame rate stability preset based on refresh rate limits and user experience requirements. (Symbol: ...) This represents the frame rate stability measurement value during the initial stage of the simulated rendering process, with the symbol... The slope of the frame rate stability data sequence obtained through linear regression is represented by the symbol. This indicates the preset minimum allowable threshold for image sharpness, symbol [symbol missing]. This represents the image sharpness measurement value at the initial stage of the simulated rendering process, with the symbol... This represents the attenuation slope of the image sharpness data sequence. (Symbol) It is a weighting factor between 0 and 1, used to balance the relative importance of frame rate stability and image sharpness in the persistence metric calculation. Storing the persistence metric for subsequent optimization means that the calculated persistence metric... Write it to the performance database or cache along with the corresponding display environment parameter configuration.

[0031] In some embodiments, the calculation of persistence metrics can employ a more complex nonlinear decay model, such as an exponential decay model, to more accurately describe the performance degradation pattern. In specific implementations, querying system hardware capability data involves continuously monitoring changes in hardware resource utilization to dynamically adjust rendering precision levels and refresh rate limit thresholds, achieving resource-adaptive parameter configuration. After generating the display environment parameter configuration file, the system generates a unique hash checksum for the configuration version to be quickly identified in subsequent verifications. When verifying the compatibility of the display environment parameter configuration file with the personalized clothing model, if a serious incompatibility issue is detected, such as the model failing to load, the system will fall back to a set of predefined safe default parameters and restart the configuration process. Simulating the rendering process of the personalized clothing model under the display environment parameters uses Level of Detail (LOD) technology, employing model copies of varying complexity at different viewing distances for a more realistic simulation. Collecting frame rate stability and image sharpness data during the simulation process requires high-precision timers and image quality assessment algorithms.

[0032] In practical implementation, when calculating the decay trend of frame rate stability and image sharpness data, piecewise linear regression or polynomial fitting can be used to improve the accuracy of trend description for non-linear decay patterns. The process of deriving persistence metrics by combining decay trends and refresh rate limit thresholds essentially predicts when frame rate stability or image sharpness data will decay to an unacceptable lower limit threshold. Storing persistence metrics for subsequent optimization means that this historical performance data can serve as prior knowledge when configuring display environment parameters for similarly complex personalized clothing models in the future, thereby progressively optimizing configuration strategies.

[0033] Example 4: In specific implementation, the optimal display speed is calculated by combining persistence metric and display time range. The system reads the previously calculated persistence metric value and the user-specified display time range from storage. The persistence metric value is a time length value, and the display time range includes a start timestamp and an end timestamp, or a total duration value. Establishing a correlation model between persistence metric and display speed is a mathematical modeling process. The correlation model uses persistence metric as an indicator of system stability and defines display speed as the number of animation loops or viewpoint changes completed by the model per unit time. The correlation model can be represented as a functional relationship, indicating which display speeds the system can stably support under a given persistence metric value. The display time range is input into the correlation model, and the correlation model adjusts its internal parameters according to the length of the display time range. For example, a longer display time range tends to choose a more conservative display speed to ensure the reliability of the system under long-term operation. The correlation model outputs multiple candidate display speeds, which are discrete or sampled values ​​within a continuous range. The impact of multiple candidate display speeds on personalized clothing models was tested by deploying personalized clothing models in a real-world operating environment or a high-fidelity simulator and applying each candidate display speed. Key performance indicators were collected during the testing process. The candidate display speed with the least impact was selected as the optimal display speed based on a comprehensive analysis of the collected performance indicators. Minimal impact means that the display speed has the least negative impact on system resource consumption and model rendering quality. The matching degree between the optimal display speed and system hardware capability data was calibrated by comparing the initially selected optimal display speed with real-time system hardware capability data such as current CPU utilization and available memory. Fine-tuning was performed as necessary to ensure matching. After calculating the optimal display speed, the system performed a secondary adaptation of this speed with the body shape feature parameters of the personalized clothing model. For the geometric complexity of the clothing model corresponding to different body shape categories, the animation frame interval and viewpoint switching step size of the display speed were fine-tuned. Simultaneously, the system synchronized the optimal display speed parameters to the benchmark parameter library of the real-time monitoring system as a speed-based benchmark threshold for subsequent judgment of whether the model state is abnormal, ensuring accurate identification of performance fluctuations caused by improper display speed adaptation during monitoring. See Table 1 for the candidate display speed test results. Table 1 Comparison of Candidate Display Speed ​​Test Indicators: In practical implementation, deploying a real-time monitoring system to observe model state changes during the display process involves initializing sensor data streams. This means starting a series of software detectors or hardware counters to collect system performance data. These sensors continuously run and generate time-series data streams. When deploying the real-time monitoring system, an independent buffer is set up for each sensor data stream. The buffer uses a circular storage structure to cache the most recent model rendering data in real time. Simultaneously, multiple sets of model state benchmark curves are preset. The real-time collected data is compared point-by-point with the corresponding benchmark curves. Only when multiple consecutive sampling points deviate from the benchmark curve is it considered an abnormal fluctuation, avoiding false triggering of anomaly flags due to instantaneous data jitter. Capturing model rendering data during the display includes real-time data retrieved or queried from the graphics API. This data includes rendering time for each frame, number of triangles drawn, texture memory usage, etc. Analyzing abnormal fluctuation patterns in the model rendering data involves applying anomaly detection algorithms to the captured time-series data. The algorithm first establishes a baseline level for performance indicators and then identifies data points or sequences that significantly deviate from the baseline. Comparing abnormal fluctuation patterns with allowable deviation thresholds involves comparing the detected abnormal fluctuation amplitude with preset thresholds. The allowable deviation threshold is an acceptable fluctuation range predefined based on historical data and user experience standards. When an abnormal fluctuation pattern persists beyond the allowable deviation threshold, a model state change event is marked. The persistence criterion is that the abnormal fluctuation must appear consecutively across multiple sampling periods. The timestamp and severity of each model state change event are recorded. The timestamp records the precise time the event was first detected, and the severity is graded based on the fluctuation amplitude and duration.

[0034] A preliminary mapping can be established using a function of the following form to model the relationship between persistence metrics and display speed: ; Where: symbol This represents the theoretically calculated display speed of a candidate generated by the association model, with the symbol... It is a scaling factor whose value is positively correlated with the system hardware performance benchmark, and its sign is... It is a persistence metric, symbol It is a smoothing constant used to avoid issues with persistence metrics. Logarithmic calculation problems at very small scales, symbols It is a unit time constant, symbol It is the total duration of the display time range.

[0035] When testing the impact of candidate display speeds, load testing can be introduced to simulate performance under high concurrency scenarios. In practice, selecting the candidate display speed with the least impact as the optimal display speed is a multi-objective decision-making process. The decision requires comprehensive consideration of multiple indicators, such as the average frame rate, CPU utilization, memory usage fluctuations, and model rendering quality score shown in Table 1. The system assigns weights to each indicator and calculates a comprehensive score for each candidate display speed, ultimately selecting the candidate display speed with the highest comprehensive score as the optimal display speed. Calibrating the optimal display speed to match the system hardware capability data involves slightly lowering or raising the value of the optimal display speed to ensure that the utilization of the main hardware resources during the display process remains within a safe target range. When initializing the sensor data stream, the sampling frequency and data format of each sensor need to be configured to ensure data stream consistency. The model rendering data captured during the display is typically stored in memory in a circular buffer to support efficient historical data backtracking analysis. Statistical methods such as Z-Score, moving average deviation, or abrupt change detection are used to analyze abnormal fluctuation patterns in the model rendering data.

[0036] In practical implementation, when comparing abnormal fluctuation patterns with allowable deviation thresholds, different thresholds can be set for different performance indicators. For example, the frame rate fluctuation threshold can be set to fluctuate within 10% relative to the baseline, while the memory usage fluctuation threshold can be set to a fixed value in megabytes. When an abnormal fluctuation pattern continuously exceeds the allowable deviation threshold, a model state change event is marked. The event log includes a timestamp, severity, the performance indicator ID that triggered the event, the specific value of the abnormal fluctuation, and related process information. The data recording the model state change event is written to the system log database for subsequent automatic adjustment or manual analysis. In some embodiments, the initialization of the sensor data stream can dynamically load and unload different sensor modules as needed to optimize resource usage. Optionally, the model rendering quality score can be objectively calculated by comparing the structural similarity (SSIM) of the rendered output with a reference image.

[0037] See Figure 4In the personalized clothing model generation stage, the model adjustment effect for different body types was quantitatively evaluated through four dimensions: contour matching accuracy, texture fitting quality, display smoothness score, and adjustment time. Specifically, after contour adjustment, the contour matching accuracy (purple bars) of the basic clothing models corresponding to each body type remained above 96%, with the standard and petite types reaching over 98%, demonstrating an effective mapping from body type feature parameters to the geometric structure of the basic model. Texture fitting quality (yellow bars) and display smoothness score (brown bars) showed differentiated performance depending on body type complexity; the full-figured type had relatively lower scores in both indicators due to the larger mesh adjustment range. Adjustment time (blue broken line) was positively correlated with the vertex displacement calculation amount during body type adjustment; the petite type took only about 2.7 seconds due to the smaller mesh modification range, while the full-figured type took 4.7 seconds. In parameter correlation, contour matching accuracy and adjustment time showed a significant positive correlation, reflecting that high-precision contour adjustment requires a longer computation time.

[0038] Example 5: In specific implementation, when the model state change exceeds the allowable deviation, the automatic adjustment process is activated. The system continuously monitors the event stream from the real-time monitoring system and retrieves relevant data for model state change events by querying the event log database. This relevant data includes the event's unique identifier, timestamp, severity level, triggered performance indicator name, and its specific deviation from the normal value. Determining the adjustment type based on this data is a classification decision process. Adjustment types include parameter fine-tuning or model regeneration. Parameter fine-tuning involves modifying display environment parameters such as rendering precision or refresh rate limits, while model regeneration requires reverting to the personalized clothing model generation stage and re-executing the contour adjustment algorithm. The decision logic is based on a predefined rule set, which considers the severity of the model state change event, the frequency of the event, and root cause analysis. For example, the rule stipulates that if the severity is "high" and similar events have occurred frequently recently, the model regeneration adjustment type is preferred; if the severity is "low" or "medium" and it is an isolated event, the parameter fine-tuning adjustment type is selected. Executing the operation sequence corresponding to the adjustment type means initiating a pre-programmed series of steps for a specific adjustment type. For parameter fine-tuning adjustment types, the operation sequence includes reading the current display environment parameter configuration file, adjusting specific parameter values ​​by a predetermined step size, verifying the validity of the new parameters, and then applying the new parameters. For model regeneration adjustment types, the operation sequence includes calling the application programming interface of the personalized clothing model generation module, passing in the necessary input data, receiving the newly generated model data, and then replacing the currently displayed model. Verifying whether the adjusted model state has returned to the normal range is done by restarting or continuing to monitor the system's data collection and analysis in real time. The system compares the performance indicators collected after adjustment with the preset normal range threshold. If it has not returned to the normal range, the adjustment process is iteratively executed until the conditions are met. The iterative process involves selecting different adjustment types, trying different parameter adjustment ranges within the parameter fine-tuning adjustment type, or trying different generation parameters in the model regeneration adjustment type.

[0039] In practice, the automatic adjustment process modifies display settings to maintain smoothness. Adjusting rendering precision and refresh rate limits in the display environment parameters is the core operation of parameter fine-tuning. Lowering the rendering precision level can quickly reduce GPU load, while increasing the refresh rate limit threshold can improve screen smoothness when the CPU has spare capacity. Updating the display attributes of personalized clothing models includes modifying the model's level of detail (LOD), disabling unnecessary post-processing effects, or adjusting the model's animation skeletal update frequency. Simultaneously modifying the optimal display speed to adapt to the new settings requires recalculating a new speed value that matches the adjusted display environment parameters and personalized clothing model display attributes. Monitoring the adjusted display smoothness metrics means that the automatic adjustment process is not a one-time operation but rather a closed loop of monitoring, evaluation, and readjustment. The system continuously tracks key metrics such as frame generation time, input latency, and screen tearing. Dynamically optimizing the triggering conditions of the automatic adjustment process based on monitoring results is based on a feedback learning mechanism. The system records the triggering reason, executed operation, and final effect of each automatic adjustment process, using this historical data to optimize the judgment threshold or rules for deciding when to re-trigger the automatic adjustment process, making it more accurate and predictive. Adjusting rendering precision and refresh rate limits in the display environment parameters can be done using an incremental strategy, such as decreasing the rendering precision by one level each time (e.g., from L2 High to L1 Medium), or increasing the refresh rate limit threshold by 5 frames per second. Updating the display attributes of personalized clothing models typically involves dynamically modifying shader parameters, texture sampling settings, or model level of detail by calling interface functions provided by the graphics engine. Synchronously adjusting the optimal display speed to accommodate new settings can be done using a simplified speed-performance model to ensure that the new speed does not immediately overload the system again.

[0040] The specific actions taken to modify display settings during the automatic adjustment process can be aided by a decision weighting formula. This formula helps select the one with the best expected effect from multiple feasible adjustment options: ; Where: symbol This represents the decision weight score of a candidate adjustment operation; a higher score indicates that the operation is selected with higher priority. (Symbol) , and These are predefined weighting coefficients, used to balance the importance of different factors. (Symbol) This represents the actual value of the abnormal performance indicator currently being monitored. (Symbol) This indicates the expected normal target value to which the performance metric is to be restored. (Symbol) This indicates a score representing the expected effectiveness of the adjustment in resolving the current anomaly. The score is based on historical data or a rule base. (Symbol) This represents the system overhead or side effects assessment value resulting from performing the adjustment operation itself, such as the degree of image quality degradation or the adjustment time. The system can calculate decision weight scores for each of the several optional operations under the parameter fine-tuning adjustment type. Then, select the operation with the highest score to execute.

[0041] In some embodiments, verifying whether the adjusted model state has returned to the normal range can be achieved by setting a short, stable observation period. The adjustment is considered successful only if the performance metrics remain normal throughout the observation period. In specific implementations, retrieving relevant data on model state change events involves correlation analysis, grouping multiple related events occurring simultaneously to identify common root causes. When determining the adjustment type based on relevant data, if the system cannot definitively determine the type, a conservative strategy can be adopted to prioritize fine-tuning parameters with lower overhead. Executing the operation sequence corresponding to the adjustment type is atomic; either all steps complete successfully, or the system rolls back to the pre-adjustment state upon failure. Verifying whether the adjusted model state has returned to the normal range needs to be completed quickly to avoid excessively long display interruptions. If the normal range has not been returned, during iterative execution of the adjustment process, each iteration records the attempted adjustment measures and their results; this information is used to optimize subsequent decisions.

[0042] In practice, adjusting rendering precision and refresh rate limits in the display environment parameters requires obtaining a graphics rendering context lock to ensure no rendering errors occur during parameter modification. If updating the display attributes of personalized clothing models is dynamic, synchronization with the rendering frames is crucial to avoid visual artifacts. Synchronizing the optimal display speed to adapt to new settings typically involves updating a speed variable controlling animation playback or camera movement. Monitoring the smoothness of the adjusted display should focus on metrics directly impacting user experience, such as frame rate stability and interactive responsiveness. Dynamically optimizing the triggering conditions for the automatic adjustment process based on monitoring results can be viewed as a parameter self-tuning process; for example, automatically raising or lowering alarm thresholds for certain performance metrics based on historical successful adjustments. In some embodiments, for model regeneration adjustment types, the regeneration task can be executed asynchronously in a background thread, with hot-switching performed only after the new model is ready, minimizing interference with the foreground display.

[0043] See Figure 5In the analysis of the adjustment strategies for an AI-powered customized clothing display system under different load scenarios, the effectiveness (%) of three adjustment strategies—parameter fine-tuning, model regeneration, and hybrid strategies—was quantified under normal load, light overload, moderate overload, heavy overload, and extreme overload scenarios. Specifically: Under normal load scenarios, the hybrid strategy's effectiveness (nearly 98%) was significantly higher than parameter fine-tuning (approximately 95%) and model regeneration (80%). As the load level increased, the effectiveness of parameter fine-tuning showed a significant downward trend, reaching only about 20% under extreme overload scenarios. Model regeneration, on the other hand, remained relatively stable after moderate overload (85%-90%), while the hybrid strategy's effectiveness gradually decreased with increasing load but remained consistently higher than parameter fine-tuning. The core value of this figure lies in clarifying the adaptability of adjustment strategies under different load scenarios: for light and lower loads, the hybrid strategy can be prioritized to ensure high effectiveness; in moderate and higher overload scenarios, the model regeneration strategy exhibits better stability; and parameter fine-tuning is more suitable for isolated performance fluctuation scenarios with low load and low severity.

[0044] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0045] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A customized clothing display method based on artificial intelligence, characterized in that, The method includes: Obtain body measurement data and clothing customization requirements submitted by customers. The body measurement data includes detailed size information and body type identification, while the clothing customization requirements include design style requirements and display time range. Standardize body measurement data by standardizing its format and correcting outliers. Select a basic clothing model using standardized body data and design style requirements; The integrated body type category identifier is used to adjust the outline of the basic clothing model to create a personalized clothing model; Configure display environment parameters, including rendering precision and refresh rate limits; Based on the personalized clothing model and display environment parameters, evaluate the durability measure of the model display; By combining persistence metrics and display time range, the optimal display speed can be calculated; Deploy a real-time monitoring system to observe changes in the model's state during the display process; When the model state changes beyond the allowable deviation, the automatic adjustment process is activated; The automatic adjustment process modifies display settings to maintain smooth display performance.

2. The artificial intelligence-based customized clothing display method as described in claim 1, characterized in that, The method involves standardizing and correcting outliers in body measurement data to generate standardized body data, the method comprising: The system analyzes detailed dimensional information from body measurement data, extracts multiple dimensional values, detects outliers among these values, corrects them using interpolation, converts the corrected dimensional values ​​into standard units of measurement, normalizes the converted dimensional values ​​according to body type category identifiers, verifies the integrity of the normalized data, and outputs standardized body data.

3. The artificial intelligence-based customized clothing display method as described in claim 2, characterized in that, The method for selecting a basic clothing model using standardized body data and design style requirements includes: Load the pre-stored basic clothing model library, which contains multiple clothing styles and corresponding size ranges; The matching degree of standardized body data with the size range of multiple clothing styles is calculated. The matching degree calculation results are filtered according to the design style requirements to select candidate clothing models. An artificial intelligence matching mechanism is applied to prioritize the candidate clothing models. The base clothing model with the highest matching degree is selected based on the priority ranking results.

4. The artificial intelligence-based customized clothing display method as described in claim 3, characterized in that, The method of integrating body type category identifiers to adjust the outline of a basic clothing model to form a personalized clothing model includes: Identify the body shape feature parameters corresponding to the body shape category identifier, obtain the geometric structure data of the basic clothing model, map the body shape feature parameters to the geometric structure data, and perform vertex displacement calculation; Adjust the mesh shape of the basic clothing model based on the vertex displacement calculation results; A texture-fitting algorithm is applied to the surface of the adjusted mesh shape to generate a personalized clothing model.

5. The artificial intelligence-based customized clothing display method as described in claim 4, characterized in that, Configuring display environment parameters, including rendering precision and refresh rate limits, the method includes: The rendering accuracy level is determined based on the complexity of the personalized clothing model, and the refresh rate limit threshold is set according to the display time range. Query system hardware capability data, adjust rendering precision level and refresh rate limit threshold, generate display environment parameter configuration file, and verify the compatibility of display environment parameter configuration file with personalized clothing model.

6. The artificial intelligence-based customized clothing display method as described in claim 5, characterized in that, The method for evaluating the durability of a model's display based on personalized clothing models and display environment parameters includes: The rendering process of a personalized clothing model under display environment parameters is simulated. Frame rate stability and image clarity data are collected during the simulation. The decay trend of frame rate stability and image clarity data is calculated. Combining the decay trend and refresh rate limit threshold, persistence metric values ​​are derived and stored for subsequent optimization.

7. The artificial intelligence-based customized clothing display method as described in claim 6, characterized in that, The method for calculating the optimal display speed by combining persistence metrics and display time range includes: Establish a correlation model between persistence metric and display speed. Input the display time range into the correlation model to obtain multiple candidate display speeds. Test the impact of multiple candidate display speeds on the personalized clothing model. Select the candidate display speed with the least impact as the optimal display speed. Calibrate the matching degree between the optimal display speed and the system hardware capability data.

8. The artificial intelligence-based customized clothing display method as described in claim 7, characterized in that, Deploying a real-time monitoring system to observe changes in the model's state during the display process, the method includes: Initialize the sensor data stream, capture the model rendering data during display, analyze the abnormal fluctuation patterns in the model rendering data, and compare the abnormal fluctuation patterns with the allowable deviation threshold. When the abnormal fluctuation pattern continues to exceed the allowable deviation threshold, mark the model state change event and record the timestamp and severity of the model state change event.

9. The artificial intelligence-based customized clothing display method as described in claim 8, characterized in that, When the model state changes beyond the allowable deviation, an automatic adjustment process is activated, the method including: Retrieve relevant data on model state change events; The type of adjustment is determined based on relevant data, including parameter fine-tuning or model regeneration; Execute the operation sequence corresponding to the adjustment type, verify whether the adjusted model state has returned to the normal range, and if it has not returned to the normal range, iterate the adjustment process until the condition is met.

10. The artificial intelligence-based customized clothing display method as described in claim 9, characterized in that, The method of automatically adjusting display settings to maintain smooth display includes: Adjust the rendering precision and refresh rate limits in the display environment parameters, update the display attributes of the personalized clothing model, synchronously modify the optimal display speed to adapt to the new settings, monitor the adjusted display smoothness indicators, and dynamically optimize the triggering conditions of the automatic adjustment process based on the monitoring results.

Images