Three-dimensional model parameter optimization method, system, electronic device and storage medium
By training machine learning models and building automated rendering processes, the problem of unpredictable rendering time of three-dimensional model is solved, precise prediction of rendering time and maximum performance, and reduced labor costs and labor hours.
Patent Information
- Application Number
- CN202410122846.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-01-30
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-01-30
AI Technical Summary
During the rendering of existing three-dimensional models, the rendering time is unpredictable, resulting in high time cost and the inability to accurately control the best rendering parameters, affecting the rendering performance.
By training machine learning models, learn the rendering configuration parameters and time relationships of three-dimensional models, build an automated rendering process and visual detection model, predict rendering time and adjust parameters to ensure rendering quality.
Accurate prediction and performance of 3D model rendering time is achieved, reducing labor costs and labor hours, and automated execution improves efficiency.
Smart Images

Figure CN117953122B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of three-dimensional modeling, and in particular to a three-dimensional model parameter optimization method, system, electronic equipment and storage medium. Background Art
[0002] With the advent of the digital age, 3D virtual animation has penetrated into many industries, from movies, advertising to video games. In these fields, character design and modeling have become the decisive factors for the quality of animation.
[0003] In the design of 3D models, modeling and rendering are key steps, which directly affect the appearance and performance of the final virtual character.
[0004] The existing 3D model rendering process often involves inputting the configuration parameters that need to be rendered, and then waiting for the software to render. However, because the rendering time automatically generated by the software is inaccurate, the time cost of rendering is unpredictable. In addition, for some model rendering tasks that are completed within a limited time, it is impossible to accurately control the optimal model rendering parameters, which often causes the 3D model rendering time to be greatly exceeded.
[0005] Therefore, it is necessary to develop a method to obtain the actual time of rendering the 3D model more accurately, so that the time spent on rendering the 3D model can be concretized, and by planning the rendering time of the 3D model, specific rendering configuration parameters can be reversely generated to maximize the rendering performance of the 3D model.
[0006] In view of this, the present invention proposes a three-dimensional model parameter optimization method, system, electronic device and storage medium. Summary of the invention
[0007] The technical problem solved by the present invention is: the present invention trains a machine learning model based on existing 3D model rendering data, learns the relationship between the theoretical rendering time and the actual rendering time of the 3D model, and constructs a visual inspection model for detecting whether the quality of the 3D model is qualified.
[0008] To achieve the above-mentioned objectives, the present invention provides the following technical solutions: obtaining rendering data of a 3D model that has been rendered; training a machine learning model to learn and predict the optimal 3D model rendering configuration parameters; building an automated rendering process for the 3D model; establishing a visual inspection model to inspect the 3D model that has been rendered, and performing secondary rendering on the 3D model that does not meet the requirements.
[0009] Preferably, the machine learning model includes a first machine learning model and a second machine learning model; the first machine learning model is used to learn the relationship between the first type of data and the second type of data, and the second machine learning model is used to learn the relationship between the second type of data and the third type of data;
[0010] The first category of data is the rendering configuration parameters of the three-dimensional model and the theoretical rendering time of the modeling software to complete the rendering configuration parameters; the second category of data is the theoretical rendering time generated by the modeling software and the actual completion amount of the rendering configuration parameters; the third category of data is the actual rendering time and rendering configuration parameters of the three-dimensional model.
[0011] Preferably, the machine learning model includes a first machine learning model and a second machine learning model; the first machine learning model is used to learn the relationship between the first type of data and the second type of data, and the second machine learning model is used to learn the relationship between the second type of data and the third type of data;
[0012] The first category of data is the rendering configuration parameters of the three-dimensional model and the theoretical rendering time of the modeling software to complete the rendering configuration parameters; the second category of data is the theoretical rendering time generated by the modeling software and the actual completion amount of the rendering configuration parameters; the third category of data is the actual rendering time and rendering configuration parameters of the three-dimensional model.
[0013] Preferably, the machine learning model is a reversible neural network model, and the training process of the first machine learning model is as follows:
[0014] According to the acquired rendering data, the first type of data and the second type of data of each completed rendered three-dimensional model are matched;
[0015] Integrate the first and second types of data into a data set, which includes a training set, a validation set, and a test set;
[0016] The first machine learning model is trained through the training set to learn the mapping relationship between the first type of data and the corresponding second type of data of a single model; after the training is completed, the first machine learning model is verified and tested through the validation set and the test set. When the model performance of the first machine is greater than expected, the training of the first machine learning model is completed;
[0017] The training process of the second machine learning model is as follows:
[0018] According to the acquired rendering data, the first type of data, the second type of data and the third type of data of each completed rendered 3D model are matched,
[0019] Integrate the first type of data, the second type of data and the third type of data into a data set, which includes a training set, a validation set and a test set;
[0020] The second machine learning model is trained through the training set to learn the mapping relationship between the first category of data, the corresponding second category of data, and the corresponding third category of data of a single model; after the training is completed, the second machine learning model is verified and tested through the validation set and the test set. When the model performance of the second machine is greater than expected, the training of the second machine learning model is completed.
[0021] Preferably, the automatic rendering process is as follows:
[0022] When setting the rendering configuration parameters first, a preliminary three-dimensional model and rendering configuration parameters are input into the modeling software to obtain a theoretical rendering time, the theoretical rendering time and the rendering configuration parameters are integrated into a first type of data, the second type of data is output through the first machine learning model, the first and second types of data are input into the second machine learning model to obtain a third type of data, and the actual rendering time in the third type of data is obtained; if the rendering time does not exceed the expected time, rendering is started; if the rendering time exceeds the expected time, the actual rendering time is manually reduced, the first type of data is reversely output, the rendering configuration parameters in the first type of data are obtained, and the modeling software performs rendering according to the newly generated rendering configuration parameters;
[0023] When setting the rendering time first, the rendering time is set in the modeling software to obtain the rendering configuration parameters; the rendering time and the rendering configuration parameters are integrated as the third type of data, and input into the second machine learning model; the second machine learning model outputs the rendering configuration parameters that can actually be completed within the rendering time, and the output rendering configuration parameters are imported into the modeling software for rendering;
[0024] All steps can be automated by building automation scripts.
[0025] Preferably, the visual detection model is a CNN neural network model, and the training steps of the visual detection model are as follows:
[0026] Classify the styles of the 3D models and obtain image data of the 3D models of each style respectively;
[0027] The image data is a full-angle image of a three-dimensional model, and the image labels are manually divided. The image labels include style labels and digital labels. The style labels are composed of letters, and the quality labels are composed of numbers. The digital label 1 indicates that it meets the rendering requirements, and the digital label 0 indicates that it does not meet the rendering requirements.
[0028] Configure the CNN neural network model, build the structure of convolutional layer, pooling layer and fully connected layer, and select activation function and loss function;
[0029] Preprocess the collected images and normalize the pixel values; divide the image data into data sets, and the data sets are divided into training sets, validation sets, and test sets;
[0030] The training set is used as the input of the CNN neural network model, and the CNN neural network model is trained to learn the feature data of qualified image labels and the feature data of unqualified label images; minimizing the loss function is used as the training goal, and the training is stopped when the loss function value is less than a preset value.
[0031] Preferably, the CNN neural network model is used to detect whether the rendered 3D model meets the requirements, collect multi-angle images of the 3D model in multiple rendering styles, assign style type labels to each image, input the collected images into the trained CNN neural network model, and the CNN neural network extracts and recognizes feature data in the image, and determines whether the rendered model meets the requirements based on the feature data of the 3D model in each style;
[0032] The model images that do not meet the requirements are transmitted to humans, who manually detect the image data after the 3D model is rendered. When the detection result is consistent with the CNN neural network, the rendering configuration parameters of the current 3D model are adjusted, and the 3D model is rendered a second time until the rendering results of the 3D model at all angles meet the requirements.
[0033] The three-dimensional model parameter optimization system is implemented based on the three-dimensional model parameter optimization method.
[0034] Preferably, the system includes a data acquisition module, a machine learning module, an automatic rendering module, and a visual inspection module;
[0035] The data acquisition module is used to acquire rendering data of the rendered three-dimensional model;
[0036] The machine learning module is used to train the machine learning model to learn and predict the optimal 3D model rendering configuration parameters;
[0037] The automated rendering module is used to construct an automated process rendering process of a three-dimensional model;
[0038] The visual inspection module is used to establish a visual inspection model, inspect the rendered three-dimensional model, and perform secondary rendering on the three-dimensional model that does not meet the requirements.
[0039] An electronic device, characterized in that it includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of a three-dimensional model parameter optimization method are implemented.
[0040] A readable storage medium, characterized in that the readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor to execute the steps of a three-dimensional model parameter optimization method.
[0041] Beneficial effects of the present invention: The present invention can achieve relatively accurate prediction of the rendering time of the 3D model through fewer data types of the 3D model, so that the time cost consumed in the 3D model rendering process can be predicted, and can also maximize the rendering performance of the 3D model by inputting the rendering time and outputting the optimal rendering parameters of the 3D model;
[0042] The method provided by the present invention is automatically executed in the 3D modeling process, and humans only need to input rendering requirements and recheck model images that do not meet the requirements, thus saving man-hours and labor costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0043] Figure 1 A flow chart of the three-dimensional model parameter optimization method provided by the present invention;
[0044] Figure 2 A structural diagram of a three-dimensional model parameter optimization system provided by the present invention;
[0045] Figure 3 It is a schematic diagram of the structure of an electronic device provided by the present invention;
[0046] Figure 4 It is a schematic diagram of the structure of a computer-readable storage medium provided by the present invention. DETAILED DESCRIPTION
[0047] In order to better understand the present application, a more detailed description will be made of various aspects of the present application with reference to the accompanying drawings. It should be understood that these detailed descriptions are only descriptions of exemplary embodiments of the present application, and are not intended to limit the scope of the present application in any way. Throughout the specification, the same reference numerals refer to the same elements. The expression "and / or" includes any and all combinations of one or more of the associated listed items.
[0048] In the accompanying drawings, the size, dimensions and shapes of the elements have been slightly adjusted for ease of illustration. The drawings are for illustration only and are not strictly drawn to scale. As used herein, the terms "substantially", "approximately" and similar terms are used as terms of approximation, not as terms of degree, and are intended to illustrate the inherent deviations in measurements or calculations that will be recognized by those of ordinary skill in the art. In addition, in this application, the order in which the steps are described does not necessarily represent the order in which these processes occur in actual operation, unless otherwise specified or can be derived from the context.
[0049] It should also be understood that expressions such as "include", "including", "have", "contain" and / or "comprising" are open rather than closed expressions in this specification, which indicate the presence of the stated features, elements and / or components, but do not exclude the presence of one or more other features, elements, components and / or combinations thereof. In addition, when expressions such as "at least one of..." appear after a list of listed features, they modify the entire list of features rather than just the individual elements in the list. In addition, when describing embodiments of the present application, "may" is used to mean "one or more embodiments of the present application". And, the term "exemplary" is intended to refer to an example or illustration.
[0050] Unless otherwise defined, all words (including engineering terms and scientific and technological terms) used in this article have the same meaning as those commonly understood by ordinary technicians in the field to which this application belongs. It should also be understood that unless there is a clear explanation in this application, words defined in commonly used dictionaries should be interpreted as having the same meaning as their meaning in the context of the relevant technology, and should not be interpreted in an idealized or overly formal sense.
[0051] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0052] Example 1
[0053] Reference Figure 1 ,The first embodiment of the present invention provides a three-dimensional model parameter optimization method.
[0054] S1: Obtain rendering data of a 3D model that has been rendered.
[0055] The rendering data of the three-dimensional model includes rendering time and rendering configuration parameters; the rendering time includes theoretical rendering time and actual rendering time; the rendering configuration parameters include configurations of geometry processing, rasterization, shading, texture mapping and lighting calculation of the three-dimensional model.
[0056] The theoretical rendering time is predicted by the modeling software and does not represent the actual rendering time of the model. There is a large error between the theoretical rendering time and the actual rendering time. The actual rendering time is generally much longer than the theoretical rendering time.
[0057] The rendering configuration parameters also include the completion amount of the rendering configuration parameters within the theoretical rendering time.
[0058] The machine learning model includes a first machine learning model and a second machine learning model; the first machine learning model is used to learn the relationship between the first category of data and the second category of data, and the second machine learning model is used to learn the relationship between the second category of data and the third category of data.
[0059] The first type of data is the rendering configuration parameters of the 3D model and the theoretical rendering time of the modeling software to complete the rendering configuration parameters: ,in is the theoretical rendering time, They are the rendering configuration parameters that the modeling software needs to complete within the theoretical rendering time for 3D model geometry processing, rasterization, shading, texture mapping and lighting calculation.
[0060] The second type of data is the actual completion amount of theoretical rendering time and rendering configuration parameters generated by the modeling software; ,in is the theoretical rendering time, They are the rendering configuration parameters for the modeling software to complete 3D model geometry processing, rasterization, shading, texture mapping and lighting calculation within the theoretical rendering time.
[0061] The third type of data is the actual rendering time and rendering configuration parameters of the 3D model: , is the actual rendering time.
[0062] S2: Train the machine learning model to learn and predict the optimal 3D model rendering configuration parameters.
[0063] Specifically, the machine learning model is a reversible neural network model, and the configuration of the reversible neural network includes an input layer, a reversible layer and an output layer.
[0064] The training process of the first machine learning model is as follows:
[0065] According to the acquired rendering data, the first type of data and the second type of data of each completed rendered three-dimensional model are matched.
[0066] The first type of data and the second type of data are integrated into a data set, which includes a training set, a validation set, and a test set.
[0067] The first machine learning model is trained through the training set to learn the mapping relationship between the first category of data and the corresponding second category of data of a single model; after the training is completed, the first machine learning model is verified and tested through the validation set and the test set. When the model performance of the first machine learning model is greater than expected, the training of the first machine learning model is completed.
[0068] The performance of the first machine learning model is measured by the loss function. When the loss function value is less than the predetermined value or converges, the model training is completed, and the mean square error MSE is used as the loss function.
[0069] The training process of the second machine learning model is as follows:
[0070] According to the acquired rendering data, the first type of data, the second type of data and the third type of data of each completed rendered three-dimensional model are matched.
[0071] The first, second and third types of data are integrated into a data set, which includes a training set, a validation set and a test set.
[0072] The second machine learning model is trained through the training set to learn the mapping relationship between the first category of data, the corresponding second category of data, and the corresponding third category of data of a single model; after the training is completed, the second machine learning model is verified and tested through the validation set and the test set. When the model performance of the second machine is greater than expected, the training of the second machine learning model is completed.
[0073] The performance of the first machine learning model is measured by the loss function. When the loss function value is less than the predetermined value or converges, the model training is completed, and the mean square error MSE is used as the loss function.
[0074] Due to the nonlinear variation of rendering performance requirements, the higher the degree of model rendering completion, that is, the later the model rendering time, the greater the rendering time variation error. Therefore, it is impossible to directly train a single machine learning model to directly derive the third category of data from the first category of data.
[0075] Therefore, the present invention adopts two machine learning models, trains the first machine learning model to derive the second type of data from the first type of data, and then derives the final third type of data from the first type of data and the second type of data. This can greatly reduce the time error and the correlation between the rendering time and the rendering configuration parameters can be more accurately reflected.
[0076] The actual rendering time output is a predicted value, which still has a certain error compared to the actual rendering time. However, compared with the error between the theoretical rendering time and the actual rendering time, the error of the predicted value is negligible.
[0077] S3: Build an automated rendering process for 3D models.
[0078] The automated rendering process is as follows:
[0079] When setting the rendering configuration parameters first, a preliminary three-dimensional model and rendering configuration parameters will be established and input into the modeling software to obtain the theoretical rendering time, and the theoretical rendering time and rendering configuration parameters will be integrated into a first type of data. The second type of data will be output through the first machine learning model, and the first and second types of data will be input into the second machine learning model to obtain the third type of data, and the actual rendering time in the third type of data will be obtained; if the rendering time does not exceed the expected time, rendering will be started; if the rendering time exceeds the expected time, the actual rendering time will be manually reduced, and the first type of data will be reversely output to obtain the rendering configuration parameters in the first type of data, and the modeling software will render according to the newly generated rendering configuration parameters.
[0080] When setting the rendering time first, set the rendering time in the modeling software to obtain the rendering configuration parameters; integrate the rendering time and the rendering configuration parameters as the third type of data, and reversely input them into the second machine learning model. The second machine learning model outputs the rendering configuration parameters that can actually be completed within the rendering time, and the output rendering configuration parameters are imported into the modeling software for rendering.
[0081] All steps can be automated by building automation scripts.
[0082] S4: Establish a visual inspection model, inspect the rendered 3D model, and perform secondary rendering on the 3D model that does not meet the requirements.
[0083] The visual detection model is a CNN neural network model, and the training steps of the visual detection model are as follows:
[0084] The 3D models are classified by style, and the image data of the 3D models of each style are obtained respectively.
[0085] The image data is a full-angle image of a three-dimensional model, and the image labels are manually divided. The image labels include style labels and digital labels. The style labels are composed of letters, and the quality labels are composed of numbers. The digital label 1 indicates that it meets the rendering requirements, and the digital label 0 indicates that it does not meet the rendering requirements.
[0086] The image features of 3D models that do not meet the requirements include image misalignment, excessive number of graphic aliasing, bright spots and dark spots in parts of the model, etc.
[0087] Configure the CNN neural network model, build the structure of convolutional layer, pooling layer and fully connected layer, and select activation function and loss function.
[0088] The collected images are preprocessed and the pixel values are normalized; the image data are divided into data sets, and the data sets are divided into training sets, validation sets, and test sets.
[0089] The training set is used as the input of the CNN neural network model, and the CNN neural network model is trained to learn the feature data of qualified image labels and the feature data of unqualified label images; minimizing the loss function is used as the training goal, and the training is stopped when the loss function value is less than a preset value.
[0090] The CNN neural network model is used to detect whether the 3D model after rendering meets the requirements. Multi-angle images of the 3D model in various rendering styles are collected, and a style type label is assigned to each image. The collected images are input into the trained CNN neural network model. The CNN neural network extracts and recognizes feature data in the image, and determines whether the rendered model meets the requirements based on the feature data of the 3D model in each style.
[0091] The model images that do not meet the requirements are transmitted to humans, who manually detect the image data after the 3D model is rendered. When the detection result is consistent with the CNN neural network, the rendering configuration parameters of the current 3D model are adjusted based on the defects in the 3D model image, and the 3D model is rendered a second time until the rendering results of the 3D model at all angles meet the requirements.
[0092] Example 2
[0093] Reference Figure 2 ,The second embodiment of the present invention provides a three-dimensional model parameter optimization system.
[0094] The system includes a data acquisition module, a machine learning module, an automatic rendering module, and a visual detection module.
[0095] The data acquisition module is used to acquire rendering data of the rendered three-dimensional model.
[0096] The rendering data of the three-dimensional model includes rendering time and rendering configuration parameters; the rendering time includes theoretical rendering time and actual rendering time; the rendering configuration parameters include configurations of geometry processing, rasterization, shading, texture mapping and lighting calculation of the three-dimensional model.
[0097] The rendering configuration parameters also include the completion amount of the rendering configuration parameters within the theoretical rendering time.
[0098] The machine learning module is used to train the machine learning model to learn and predict the optimal 3D model rendering configuration parameters.
[0099] The machine learning model includes a first machine learning model and a second machine learning model; the first machine learning model is used to learn the relationship between the first category of data and the second category of data, and the second machine learning model is used to learn the relationship between the second category of data and the third category of data.
[0100] The first category of data is the rendering configuration parameters of the three-dimensional model and the theoretical rendering time of the modeling software to complete the rendering configuration parameters; the second category of data is the theoretical rendering time generated by the modeling software and the actual completion amount of the rendering configuration parameters; the third category of data is the actual rendering time and rendering configuration parameters of the three-dimensional model.
[0101] The automatic rendering module is used to construct an automatic rendering process of the three-dimensional model.
[0102] The automated rendering process is executed by an automated script.
[0103] The visual inspection module is used to establish a visual inspection model, inspect the rendered three-dimensional model, and perform secondary rendering on the three-dimensional model that does not meet the requirements.
[0104] The visual detection model is a CNN neural network model. The CNN neural network model is configured and trained. After the training is completed, it is used to detect whether the rendering result of the three-dimensional model meets the requirements.
[0105] When it is detected that the image of the rendered 3D model does not meet the requirements, the model image that does not meet the requirements is transmitted to humans, who perform a secondary inspection on the image data after the 3D model is rendered. When the manual inspection result is consistent with the CNN neural network, the rendering configuration parameters of the current 3D model are adjusted, and the 3D model is rendered a second time until the rendering results of the 3D model at all angles meet the requirements.
[0106] Example 3
[0107] Figure 3 1 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 3 As shown, according to another aspect of the present application, an electronic device 500 is also provided. The electronic device 500 may include one or more processors and one or more memories. The memories store computer readable codes, and when the computer readable codes are executed by one or more processors, the method for reducing energy consumption and improving efficiency of fine burning process based on multivariate analysis as described above may be executed.
[0108] The method or system according to the embodiment of the present application can also be used Figure 3 The electronic device architecture shown in FIG. Figure 3 As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to a network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as ROM 503 or hard disk 507, may store the method for reducing energy consumption and improving efficiency of the fine firing process based on multivariate analysis provided in the present application. The method for reducing energy consumption and improving efficiency of the fine firing process based on multivariate analysis may, for example, include: obtaining rendering data of a three-dimensional model that has been rendered; training a machine learning model to learn and predict the optimal three-dimensional model rendering configuration parameters; constructing an automated rendering process for the three-dimensional model; establishing a visual detection model to detect the three-dimensional model that has been rendered, and performing secondary rendering on the three-dimensional model that does not meet the requirements.
[0109] Furthermore, the electronic device 500 may also include a user interface 508. Figure 3 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 3 One or more components of an electronic device are shown.
[0110] Example 4
[0111] Figure 4 Schematic diagram of a computer-readable storage medium structure provided by an embodiment of the present application. Figure 4 As shown, a computer-readable storage medium 600 according to an embodiment of the present application is shown. Computer-readable instructions are stored on the computer-readable storage medium 600. When the computer-readable instructions are executed by the processor, the method for reducing energy consumption and improving efficiency of the fine burning process based on multivariate analysis according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium 600 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.
[0112] In addition, according to the implementation of the present application, the process described with reference to the flowchart above can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, the non-transitory machine-readable storage medium stores machine-readable instructions, the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application, such as: obtaining rendering data of a three-dimensional model that has been rendered; training a machine learning model to learn and predict the optimal three-dimensional model rendering configuration parameters; building an automated rendering process for a three-dimensional model; establishing a visual inspection model to inspect the rendered three-dimensional model, and performing secondary rendering on the three-dimensional model that does not meet the requirements.
[0113] The methods, apparatuses, and devices of the present application may be implemented in many ways. For example, the methods, apparatuses, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers recording media storing programs for executing the method according to the present application.
[0114] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.
[0115] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. 3D model parameter optimization method, including: Obtaining rendering data of the rendered three-dimensional model; Train machine learning models to learn and predict optimal 3D model rendering configuration parameters; Build an automated rendering process for 3D models; Establish a visual inspection model to inspect the rendered 3D models, and perform secondary rendering on the 3D models that do not meet the requirements; The rendering data of the three-dimensional model includes rendering time and rendering configuration parameters; the rendering time includes theoretical rendering time and actual rendering time; The rendering configuration parameters include configurations of geometry processing, rasterization, shading, texture mapping and lighting calculation of the three-dimensional model; The rendering configuration parameters also include the completion amount of the rendering configuration parameters within the theoretical rendering time; The machine learning model includes a first machine learning model and a second machine learning model; the first machine learning model is used to learn the relationship between the first type of data and the second type of data, and the second machine learning model is used to learn the relationship between the second type of data and the third type of data; The first type of data is the rendering configuration parameters of the three-dimensional model and the theoretical rendering time of the modeling software to complete the rendering configuration parameters; The second type of data is the actual completion amount of the theoretical rendering time and rendering configuration parameters generated by the modeling software; the third type of data is the actual rendering time and rendering configuration parameters of the three-dimensional model; The machine learning model is a reversible neural network model, and the training process of the first machine learning model is as follows: According to the acquired rendering data, the first type of data and the second type of data of each completed rendered three-dimensional model are matched; Integrate the first and second types of data into a data set, which includes a training set, a validation set, and a test set; Training the first machine learning model through the training set to learn the mapping relationship between the first type of data of a single model and the corresponding second type of data; After the training is completed, the first machine learning model is verified and tested through the verification set and the test set. When the model performance of the first machine is greater than expected, the training of the first machine learning model is completed; The training process of the second machine learning model is as follows: According to the acquired rendering data, the first type of data, the second type of data and the third type of data of each completed rendered 3D model are matched, Integrate the first type of data, the second type of data and the third type of data into a data set, which includes a training set, a validation set and a test set; Training the second machine learning model through the training set to learn the mapping relationship between the first type of data, the corresponding second type of data, and the corresponding third type of data of a single model; After the training is completed, the second machine learning model is verified and tested through the verification set and the test set. When the model performance of the second machine is greater than expected, the training of the second machine learning model is completed; The automated rendering process is as follows: When setting the rendering configuration parameters first, a preliminary three-dimensional model and the rendering configuration parameters are input into the modeling software to obtain a theoretical rendering time, the theoretical rendering time and the rendering configuration parameters are integrated into a first type of data, the second type of data is output through a first machine learning model, the first and second types of data are input into a second machine learning model to obtain a third type of data, and the actual rendering time in the third type of data is obtained; If the rendering time does not exceed the expected time, rendering is started; if the rendering time exceeds the expected time, the actual rendering time is manually reduced, the first type of data is reversely output, the rendering configuration parameters in the first type of data are obtained, and the modeling software performs rendering according to the newly generated rendering configuration parameters; When setting the rendering time first, the rendering time is set in the modeling software to obtain the rendering configuration parameters; the rendering time and the rendering configuration parameters are integrated as the third type of data, and input into the second machine learning model; the second machine learning model outputs the rendering configuration parameters that can actually be completed within the rendering time, and the output rendering configuration parameters are imported into the modeling software for rendering; All steps can be automatically implemented by building automation scripts; The visual detection model is a CNN neural network model, and the training steps of the visual detection model are as follows: Classify the styles of the 3D models and obtain image data of the 3D models of each style respectively; The image data is a full-angle image of a three-dimensional model, and the image labels are manually divided. The image labels include style labels and digital labels. The style labels are composed of letters, and the quality labels are composed of numbers. The digital label 1 indicates that it meets the rendering requirements, and the digital label 0 indicates that it does not meet the rendering requirements. Configure the CNN neural network model, build the structure of convolutional layer, pooling layer and fully connected layer, and select activation function and loss function; Preprocess the collected images and normalize the pixel values; divide the image data into data sets, and the data sets are divided into training sets, validation sets, and test sets; The training set is used as the input of the CNN neural network model, and the CNN neural network model is trained to learn the feature data of qualified image labels and the feature data of unqualified label images; minimizing the loss function is used as the training goal, and the training is stopped when the loss function value is less than a preset value.
2. The three-dimensional model parameter optimization method according to claim 1, characterized in that: The CNN neural network model is used to detect whether the rendered 3D model meets the requirements, collect multi-angle images of the 3D model in various rendering styles, classify each image into a style type label, input the collected images into the trained CNN neural network model, and the CNN neural network extracts and recognizes feature data in the image, and judges whether the rendered model meets the requirements based on the feature data of the 3D model in each style; The model images that do not meet the requirements are transmitted to humans, who manually detect the image data after the 3D model is rendered. When the detection result is consistent with the CNN neural network, the rendering configuration parameters of the current 3D model are adjusted, and the 3D model is rendered a second time until the rendering results of the 3D model at all angles meet the requirements.
3. A three-dimensional model parameter optimization system, which is implemented based on the three-dimensional model parameter optimization method according to any one of claims 1 and 2, characterized in that: Including data acquisition module, machine learning module, automatic rendering module, and visual inspection module; The data acquisition module is used to acquire rendering data of the rendered three-dimensional model; The machine learning module is used to train the machine learning model to learn and predict the optimal 3D model rendering configuration parameters; The automated rendering module is used to construct an automated process rendering process of a three-dimensional model; The visual inspection module is used to establish a visual inspection model, inspect the rendered three-dimensional model, and perform secondary rendering on the three-dimensional model that does not meet the requirements.
4. An electronic device, characterized in that: include: A processor and a memory, wherein the memory stores a computer program that can be called by the processor; The processor executes the three-dimensional model parameter optimization method according to any one of claims 1 and 2 by calling the computer program stored in the memory.
5. A computer-readable storage medium, characterized in that: Instructions are stored, and when the instructions are executed on a computer, the computer is caused to execute the three-dimensional model parameter optimization method as claimed in any one of claims 1 and 2.
Citation Information
Patent Citations
Marine auxiliary boiler virtual control simulation system based on force feedback
CN102707989A
Rendered-image acquiring method, device and system and storage medium
CN108876764A
Virtual character model processing method and device and readable storage medium
CN111739134A
Digital human rendering method and system under condition of low-configuration display card
CN115471599A
Web platform for interactive design, synthesis and delivery of 3D character motion data
US20100259547A1
Cited By
Automatic optimization method and system for AI exhibition hall multi-fold screen rendering parameters
CN122289487A
