Three-dimensional model rendering duration prediction method and system

Abstract

The invention relates to a three-dimensional model rendering duration prediction method and system, and belongs to the technical field of three-dimensional models, and the method comprises the steps: obtaining a first rendering parameter related to the rendering duration in file data of a three-dimensional model in a rendering cloud platform through a parameter obtaining script, obtaining a second rendering parameter in the database, and integrating the second rendering parameter with the first rendering parameter to obtain an initial data set; training the linear regression model and deploying the linear regression model to a server of a rendering cloud platform; the rendering cloud platform extracts a first rendering parameter of the target three-dimensional model based on the target three-dimensional model, obtains the number of rendering nodes selected and matched for the target three-dimensional model, predicts the rendering duration by using a linear regression model, and outputs a prediction result of the rendering duration; the appropriate number of rendering nodes can be selected according to the rendering duration, prediction errors are reduced, resource waste is avoided, the rendering efficiency is improved, the predicted rendering duration is integrated into the rendering cloud platform, the platform function is enriched, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional models, and in particular to a method and system for predicting the rendering duration of a three-dimensional model. Background Art

[0002] With the rapid development of 3D modeling and animation technology, 3D models are increasingly used in film and television production, game development, industrial design and other fields. However, the rendering process of 3D models often takes a long time and is affected by many factors such as model complexity, material properties, light source settings, computing resources, etc. In large-scale rendering projects, how to accurately predict the rendering time, select a reasonable number of nodes, and reasonably arrange rendering tasks has become a key issue to improve rendering efficiency and reduce costs.

[0003] In the current prior art, when rendering a model, it is sufficient to ensure that the rendering time is within a preset time range. For example, patent CN117953122A discloses a three-dimensional model parameter optimization method, system, electronic device and storage medium. In this scheme, 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, and the theoretical rendering time and rendering configuration parameters are integrated into a first type of data. The second type of data is output through a first machine learning model, and 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, the 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, and the rendering configuration parameters in the first type of data are obtained, and the modeling software renders according to the newly generated rendering configuration parameters.

[0004] However, the above method has the problem that when the model rendering time is unknown, it is difficult to choose the number of rendering nodes. At this time, the model rendering will inevitably take more time and cost more than expected. Summary of the invention

[0005] In view of the shortcomings existing in the related art, the purpose of the present invention is to provide a 3D model rendering time prediction method and system to solve the technical problem that when the model rendering time is unknown, it is difficult to choose the number of rendering nodes, and the model rendering will inevitably take time and cost more than expected.

[0006] The present invention provides a method for predicting the rendering duration of a three-dimensional model, comprising the following steps: Data acquisition step: obtaining a first rendering parameter related to rendering time in the file data of the three-dimensional model in the rendering cloud platform through a parameter acquisition script, and obtaining a second rendering parameter in a database; Data processing step: integrating the first rendering parameter and the second rendering parameter to obtain an initial data set, and saving the initial data set into a model data file; Model training step: training the linear regression model according to the initial data set, and deploying the trained linear regression model to the server of the rendering cloud platform; Model prediction step: the rendering cloud platform extracts a first rendering parameter of the target three-dimensional model based on the target three-dimensional model, obtains the number of rendering nodes selected for the target three-dimensional model, uses the linear regression model to predict the rendering time based on the first rendering parameter of the target three-dimensional model and the number of rendering nodes, and outputs a prediction result of the rendering time; The embodiment of the present invention can more accurately predict the rendering time of a three-dimensional model on a Sunway supercomputing cluster through training of a linear regression model, can predict the rendering time according to the number of rendering nodes, and select an appropriate number of rendering nodes according to the rendering time, thereby reducing prediction errors, helping users to reasonably select the number of rendering nodes according to the prediction results, avoiding waste of resources, and improving rendering efficiency. The predicted rendering time is integrated into the rendering cloud platform, enriching the platform functions and improving the user experience.

[0007] In some embodiments of the present invention, the model training step is specifically: Deploy the deep learning framework to the Sunway supercomputing cluster, and train and deploy the linear regression model through the deep learning framework; Extracting batch samples of the training set in the initial data set, calculating the model prediction value by forward propagation, calculating the loss function value by the model parameters, updating the model parameters by iteration, evaluating the performance of the linear regression model by the difference between the model prediction value and the loss function value, validating the linear regression model by the validation set in the initial data set, and testing the linear regression model by the test set in the initial data set; Wherein, the linear regression model includes: An input layer, used to obtain the first rendering parameter and the number of rendering nodes of the initial data set and / or the target three-dimensional model; The output layer performs calculations on the cores of the nodes in the Sunway supercomputing cluster and outputs the prediction results of the rendering time.

[0008] The embodiment of the present invention uses a linear regression model to predict parameters with linear characteristics, thereby reducing the amount of calculation, saving computing resources, and achieving a more accurate prediction effect.

[0009] In some embodiments of the present invention, the first rendering parameters include: number of vertices, number of edges, number of faces, diffuse reflection, number of samples, light visibility, noise reduction parameters, number of light bounces, specular reflection, glossiness, transmission, subsurface scattering, volume scattering, shadow, ambient occlusion, motion blur, depth of field, clamping, light sampling, noise threshold, number of blocks, x Direction pixels, y Direction pixels and total number of frames; The second rendering parameters include: the number of rendering nodes and the rendering time; The initial data set includes a plurality of rendering data, each of which includes a first rendering parameter and a second rendering parameter. After shape adjustment is performed on each rendering data, the shape of the initial data set after shape adjustment is obtained: , and split the initial data set after shape adjustment into training set, validation set and test set; among them, N The number of rendered data.

[0010] The embodiment of the present invention obtains rendering parameters related to rendering time and adjusts the shape of rendering data so that the weights between different features can be distributed more reasonably, avoiding excessive influence of a feature on the result due to excessively large or small values, improving the training speed and convergence of the model, and contributing to a deep understanding of the interaction between various parameters in the rendering process.

[0011] Some embodiments of the present invention further provide a 3D model rendering duration prediction system, comprising: Data acquisition module: obtains the first rendering parameter related to the rendering time in the file data of the three-dimensional model in the rendering cloud platform through the parameter acquisition script, and obtains the second rendering parameter in the database; Data processing module: integrating the first rendering parameter and the second rendering parameter to obtain an initial data set, and saving the initial data set into a model data file; Model training module: training the linear regression model according to the initial data set, and deploying the trained linear regression model to the server of the rendering cloud platform; Model prediction module: The rendering cloud platform extracts the first rendering parameter of the target three-dimensional model based on the target three-dimensional model, and obtains the number of rendering nodes selected for the target three-dimensional model, uses the linear regression model to predict the rendering time based on the first rendering parameter of the target three-dimensional model and the number of rendering nodes, and outputs the prediction result of the rendering time.

[0012] Among them, the model training module is specifically: Deploy the deep learning framework to the Sunway supercomputing cluster, and train and deploy the linear regression model through the deep learning framework; Extract batch samples of the training set in the initial data set, calculate the model prediction value through forward propagation, calculate the loss function value through the model parameters, update the model parameters through iteration, evaluate the performance of the linear regression model through the difference between the model prediction value and the loss function value, verify the linear regression model through the validation set in the initial data set, and test the linear regression model through the test set in the initial data set.

[0013] Wherein, the linear regression model includes: An input layer, used to obtain the first rendering parameter and the number of rendering nodes of the initial data set and / or the target three-dimensional model; The output layer performs calculations on the cores of the nodes in the Sunway supercomputing cluster and outputs the prediction results of the rendering time.

[0014] The first rendering parameters include: number of vertices, number of edges, number of faces, diffuse reflection, number of samples, light visibility, denoising parameters, number of light bounces, specular reflection, glossiness, transmission, subsurface scattering, volume scattering, shadow, ambient occlusion, motion blur, depth of field, clamping, light sampling, noise threshold, number of blocks, x Direction pixels, y Direction pixels and total number of frames; The second rendering parameters include: the number of rendering nodes and the rendering time.

[0015] The initial data set includes a plurality of rendering data, each of which includes a first rendering parameter and a second rendering parameter. After shape adjustment is performed on each rendering data, the shape of the initial data set after shape adjustment is obtained: , and split the initial data set after shape adjustment into training set, validation set and test set; among them, N The number of rendered data.

[0016] The embodiment of the present invention can more accurately predict the rendering time of a three-dimensional model on a Sunway supercomputing cluster through training of a linear regression model, can predict the rendering time according to the number of rendering nodes, and select an appropriate number of rendering nodes according to the rendering time, thereby reducing prediction errors, helping users to reasonably select the number of rendering nodes according to the prediction results, avoiding waste of resources, and improving rendering efficiency. The predicted rendering time is integrated into the rendering cloud platform, enriching platform functions, improving user experience, and helping to promote the deep integration of three-dimensional rendering technology and high-performance computing technology, and promoting technological innovation and development in related fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the specific embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work. Figure 1 A flowchart of a method for predicting rendering duration of a three-dimensional model provided by an embodiment of the present invention; Figure 2 An application flow chart of a 3D model rendering duration prediction method provided by an embodiment of the present invention; Figure 3 A schematic diagram of the structure of a 3D model rendering duration prediction system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application. It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other. The technical solution of the present invention is described in detail below in conjunction with specific embodiments and the accompanying drawings.

[0019] As attached Figure 1 -Attached Figure 2 As shown, the present invention provides a method for predicting the rendering duration of a three-dimensional model, comprising the following steps: Data acquisition step S1: obtaining a first rendering parameter related to rendering time in the file data of the three-dimensional model in the rendering cloud platform through a parameter acquisition script, and obtaining a second rendering parameter in the database; optionally, the parameter acquisition script is a get_para.py script in the API provided by Blender; Data processing step S2: integrating the first rendering parameter and the second rendering parameter to obtain an initial data set, and saving the initial data set into a model data file; Model training step S3: training the linear regression model according to the initial data set, and deploying the trained linear regression model to the server of the rendering cloud platform; Model prediction step S4: The rendering cloud platform extracts the first rendering parameter of the target three-dimensional model based on the target three-dimensional model, and obtains the number of rendering nodes selected for the target three-dimensional model, and uses a linear regression model to predict the rendering time based on the first rendering parameter of the target three-dimensional model and the number of rendering nodes, and outputs the prediction result of the rendering time.

[0020] Through the above method, training the linear regression model can more accurately predict the rendering time of the three-dimensional model on the Sunway supercomputing cluster. It can predict the rendering time according to the number of rendering nodes, and select the appropriate number of rendering nodes according to the rendering time, thereby reducing the prediction error and helping users to reasonably select the number of rendering nodes according to the prediction results, avoid resource waste, improve rendering efficiency, integrate the predicted rendering time into the rendering cloud platform, enrich the platform functions, and improve the user experience.

[0021] Among them, the model training step S3 is specifically as follows: Deploy the deep learning framework to the Sunway supercomputing cluster, and train and deploy the linear regression model through the deep learning framework; Optionally, the linear regression model can be trained and deployed through the deep learning framework swpytorch. The Sunway supercomputing cluster has rich computing resources and adopts a heterogeneous architecture. Each node integrates a 6-core group, each core group contains a master core and 64 slave cores, and the computing resources are mainly concentrated on the slave cores. The deep learning framework swpytorch is optimized based on the unique architecture of the Sunway supercomputing cluster, with better acceleration effect and strong scalability. The deep learning framework swpytorch can be used for faster training and improved efficiency. Extract batch samples of the training set in the initial data set, calculate the model prediction value through forward propagation, calculate the loss function value through the model parameters, update the model parameters through iteration, evaluate the performance of the linear regression model through the difference between the model prediction value and the loss function value, verify the linear regression model through the validation set in the initial data set, and test the linear regression model through the test set in the initial data set; Among them, the linear regression model includes: An input layer, used to obtain the initial data set and / or the first rendering parameter and the number of rendering nodes of the target three-dimensional model; The output layer performs calculations on the cores of the nodes in the Shenwei supercomputing cluster and outputs the prediction results of the rendering time. Optionally, the model parameters are model weights and bias items; the linear regression model can better fit the training data by adjusting the model weights and bias items. The model weights determine the degree of influence of the input features on the output, while the bias items determine the baseline value of the model output. The goal of the model is to minimize the prediction error, that is, to reduce the difference between the model prediction value and the true value. The model weights and bias items are adjusted by, for example, the stochastic gradient descent method as an optimization algorithm to gradually reduce the above error. Appropriate adjustment of the model weights and bias items helps the linear regression model perform well on unseen data, that is, improves the generalization ability of the model; The performance of the linear regression model is evaluated by using the mean square error as the loss function. The model parameters are updated through iterative optimization using the stochastic gradient descent method as the optimizer. The loss function provides a quantitative indicator to measure the accuracy of the model prediction. By calculating the difference between the predicted value and the actual value, the performance of the model can be objectively evaluated. The loss function is the core of the optimization algorithm. The optimization algorithm calculates the gradient of the loss function relative to the model parameters to guide how to adjust these parameters to reduce the loss. Optionally, the validation set is used to adjust hyperparameters and prevent model overfitting, to select the best model configuration and adjust the complexity of the model. If multiple different models are trained or different hyperparameters are used, the validation set can be used to compare the performance of these models and select the best model. Although hyperparameters are usually adjusted using the validation set, the final model selection and hyperparameter settings can also be finally verified on the validation set. Hyperparameters include: batch size, number of iterations, and learning rate. The test set is used to evaluate the generalization ability of the model and simulate the performance of the model in actual applications. By evaluating the model on the test set, it can be tested whether the model is overfitting. If the model performs well on the training set but poorly on the test set, this may indicate that the model is too complex, causing it to learn the noise and details in the training data rather than the underlying data distribution. In order to make the weights between different features more reasonably distributed, avoid a certain feature from having excessive influence on the result due to excessively large or small values, improve the training speed and convergence of the model, and help to deeply understand the interaction between various parameters in the rendering process, the properties and types of the first rendering parameters and the second rendering parameters are obtained as shown in Table 1, where the first rendering parameters include: number of vertices, number of edges, number of faces, diffuse reflection, number of samples, light visibility, denoising parameters, number of light bounces, specular reflection, glossiness, transmission, subsurface scattering, volume scattering, shadow, ambient occlusion, motion blur, depth of field, clamping, light sampling, noise threshold, number of blocks, x Direction pixels, y Direction pixels and total number of frames; The second rendering parameters include: the number of rendering nodes and the rendering time; Table 1 Attributes and types of the first rendering parameter and the second rendering parameter

[0022] Preferably, the first rendering parameter is obtained through the get_para.py script in the API provided by Blender in the following steps: def check_frams(): info="Total frames:" frames=bpy.context.scene.frame_end - bpy.context.scene.frame_start + 1 return info+str(frams)+"\n" def check_fps(): info="Number of samples:" fps=bpy.context.scene.render.max_samples return info+str(fps)+"\n" def check_dpi(): info="Resolution:" dpi=str(bpy.context.scene.render.resolution_x)+" "+"x"" "+ str(bpy.context.scene.render.resolution_y) return info+str(dpi)+"\n" Among them, the resolution corresponds to the pixels in the x direction and the pixels in the y direction; The initial data set includes multiple rendering data, each of which includes a first rendering parameter and a second rendering parameter. After each rendering data is subjected to shape adjustment, the shape of the initial data set after shape adjustment is obtained: , and split the initial data set after shape adjustment into training set, validation set and test set; among them, N is the number of rendered data; optionally, 80% of the data in the initial data set is used as the training set, 10% of the data is used as the validation set, and 10% of the data is used as the test set; in order to solve the problem that due to the large difference in the numerical range of each attribute, direct training may cause floating point overflow or underflow during calculation, and will cause difficulties in the optimization process and greatly prolong the training time, the rendering parameters are normalized.

[0023] Reference Figure 3 As shown, an embodiment of the present invention further provides a 3D model rendering duration prediction system, comprising: Data acquisition module 1: obtaining a first rendering parameter related to rendering time in the file data of the three-dimensional model in the rendering cloud platform through a parameter acquisition script, and obtaining a second rendering parameter in the database; Data processing module 2: integrating the first rendering parameter and the second rendering parameter to obtain an initial data set, and saving the initial data set into a model data file; Model training module 3: Train the linear regression model based on the initial data set and deploy the trained linear regression model to the server of the rendering cloud platform; Model prediction module 4: The rendering cloud platform extracts the first rendering parameter of the target three-dimensional model based on the target three-dimensional model, and obtains the number of rendering nodes selected for the target three-dimensional model, uses a linear regression model to predict the rendering time based on the first rendering parameter of the target three-dimensional model and the number of rendering nodes, and outputs the prediction result of the rendering time.

[0024] Among them, the model training module 3 is specifically: Deploy the deep learning framework to the Sunway supercomputing cluster, and train and deploy the linear regression model through the deep learning framework; Extract batch samples of the training set in the initial data set, calculate the model prediction value through forward propagation, calculate the loss function value through the model parameters, update the model parameters through iteration, evaluate the performance of the linear regression model through the difference between the model prediction value and the loss function value, verify the linear regression model through the validation set in the initial data set, and test the linear regression model through the test set in the initial data set.

[0025] Among them, the linear regression model includes: An input layer, used to obtain the initial data set and / or the first rendering parameter and the number of rendering nodes of the target three-dimensional model; The output layer performs calculations on the cores of the nodes in the Sunway supercomputing cluster and outputs the prediction results of the rendering time.

[0026] Among them, the first rendering parameters include: number of vertices, number of edges, number of faces, diffuse reflection, number of samples, light visibility, denoising parameters, number of light bounces, specular reflection, glossiness, transmission, subsurface scattering, volume scattering, shadow, ambient occlusion, motion blur, depth of field, clamping, light sampling, noise threshold, number of blocks, x Direction pixels, y Direction pixels and total number of frames; The second rendering parameter includes: the number of rendering nodes and the rendering time.

[0027] The initial data set includes multiple rendering data, each of which includes a first rendering parameter and a second rendering parameter. After each rendering data is subjected to shape adjustment, the shape of the initial data set after shape adjustment is obtained: , and split the initial data set after shape adjustment into training set, validation set and test set; among them, N The number of rendered data.

[0028] It should be noted that the above is a reference for a method and system for predicting the rendering duration of a three-dimensional model, and the present invention is not limited thereto.

[0029] The embodiments of the present invention realize that the rendering time of a three-dimensional model on a Sunway supercomputing cluster can be more accurately predicted through training of a linear regression model, thereby reducing prediction errors, helping users to reasonably select the number of rendering nodes based on prediction results, avoiding waste of resources, and improving rendering efficiency. The predicted rendering time is integrated into the rendering cloud platform, enriching platform functions, and improving user experience. This solves the technical problem that when the prior art does not understand the model rendering time, it is difficult to select the number of rendering nodes, and at this time, the model rendering will inevitably be time-consuming and costly beyond expectations.

[0030] Finally, it should be noted that: the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other. The above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solution of the present invention, which should be included in the scope of the technical solution for protection of the present invention.

Claims

1. A method for predicting the rendering time of a three-dimensional model, characterized in that: The steps include: Data acquisition step: obtaining a first rendering parameter related to rendering time in the file data of the three-dimensional model in the rendering cloud platform through a parameter acquisition script, and obtaining a second rendering parameter in a database; Data processing step: integrating the first rendering parameter and the second rendering parameter to obtain an initial data set, and saving the initial data set into a model data file; Model training step: training the linear regression model according to the initial data set, and deploying the trained linear regression model to the server of the rendering cloud platform; Model prediction step: the rendering cloud platform extracts the first rendering parameter of the target three-dimensional model based on the target three-dimensional model, and obtains the number of rendering nodes selected for the target three-dimensional model, uses the linear regression model to predict the rendering time based on the first rendering parameter of the target three-dimensional model and the number of rendering nodes, and outputs the prediction result of the rendering time.

2. The 3D model rendering duration prediction method according to claim 1, characterized in that: The model training steps are specifically as follows: Deploy the deep learning framework to the Sunway supercomputing cluster, and train and deploy the linear regression model through the deep learning framework; Extract batch samples of the training set in the initial data set, calculate the model prediction value through forward propagation, calculate the loss function value through the model parameters, update the model parameters through iteration, evaluate the performance of the linear regression model through the difference between the model prediction value and the loss function value, verify the linear regression model through the validation set in the initial data set, and test the linear regression model through the test set in the initial data set.

3. The 3D model rendering duration prediction method according to claim 2, characterized in that: The linear regression model includes: An input layer, used to obtain the first rendering parameter and the number of rendering nodes of the initial data set and / or the target three-dimensional model; The output layer performs calculations on the cores of the nodes in the Sunway supercomputing cluster and outputs the prediction results of the rendering time.

4. The 3D model rendering duration prediction method according to claim 1, characterized in that: The first rendering parameters include: number of vertices, number of edges, number of faces, diffuse reflection, number of samples, light visibility, denoising parameters, number of light bounces, specular reflection, glossiness, transmission, subsurface scattering, volume scattering, shadow, ambient occlusion, motion blur, depth of field, clamping, light sampling, noise threshold, number of blocks, x Direction pixels, y Direction pixels and total number of frames; The second rendering parameters include: the number of rendering nodes and the rendering time.

5. The 3D model rendering duration prediction method according to claim 4, characterized in that: The initial data set includes a plurality of rendering data, wherein each rendering data includes a first rendering parameter and a second rendering parameter. After shape adjustment is performed on each rendering data, the shape of the initial data set after shape adjustment is obtained as follows: , and split the initial data set after shape adjustment into training set, validation set and test set; among them, N The number of rendered data.

6. A 3D model rendering time prediction system, characterized in that: include: Data acquisition module: obtains the first rendering parameter related to the rendering time in the file data of the three-dimensional model in the rendering cloud platform through the parameter acquisition script, and obtains the second rendering parameter in the database; Data processing module: integrating the first rendering parameter and the second rendering parameter to obtain an initial data set, and saving the initial data set into a model data file; Model training module: training the linear regression model according to the initial data set, and deploying the trained linear regression model to the server of the rendering cloud platform; Model prediction module: The rendering cloud platform extracts the first rendering parameter of the target three-dimensional model based on the target three-dimensional model, and obtains the number of rendering nodes selected for the target three-dimensional model, uses the linear regression model to predict the rendering time based on the first rendering parameter of the target three-dimensional model and the number of rendering nodes, and outputs the prediction result of the rendering time.

7. The 3D model rendering duration prediction system according to claim 6, characterized in that: The model training module is specifically: Deploy the deep learning framework to the Sunway supercomputing cluster, and train and deploy the linear regression model through the deep learning framework; Extract batch samples of the training set in the initial data set, calculate the model prediction value through forward propagation, calculate the loss function value through the model parameters, update the model parameters through iteration, evaluate the performance of the linear regression model through the difference between the model prediction value and the loss function value, verify the linear regression model through the validation set in the initial data set, and test the linear regression model through the test set in the initial data set.

8. The 3D model rendering duration prediction system according to claim 7, characterized in that: The linear regression model includes: An input layer, used to obtain the first rendering parameter and the number of rendering nodes of the initial data set and / or the target three-dimensional model; The output layer performs calculations on the cores of the nodes in the Sunway supercomputing cluster and outputs the prediction results of the rendering time.

9. The 3D model rendering duration prediction system according to claim 6, characterized in that: The first rendering parameters include: number of vertices, number of edges, number of faces, diffuse reflection, number of samples, light visibility, denoising parameters, number of light bounces, specular reflection, glossiness, transmission, subsurface scattering, volume scattering, shadow, ambient occlusion, motion blur, depth of field, clamping, light sampling, noise threshold, number of blocks, x Direction pixels, y Direction pixels and total number of frames; The second rendering parameters include: the number of rendering nodes and the rendering time.

10. The 3D model rendering duration prediction system according to claim 9, characterized in that: The initial data set includes a plurality of rendering data, wherein each rendering data includes a first rendering parameter and a second rendering parameter. After shape adjustment is performed on each rendering data, the shape of the initial data set after shape adjustment is obtained as follows: , and split the initial data set after shape adjustment into training set, validation set and test set; among them, N The number of rendered data.

Images