Interactive three-dimensional animation generation method

Abstract

The invention provides an interactive three-dimensional animation generation method, and relates to the technical field of interactive three-dimensional animations. The interactive three-dimensional animation generation method comprises the following steps: data preparation: collecting and preparing a three-dimensional animation data set for training a deep learning model, including a model, an animation sequence and user interaction data; deep learning model training: using the three-dimensional animation data set to train a deep learning model; interactive animation generation: automatically generating a three-dimensional animation sequence by using the trained deep learning model according to user input; and interactive control: adjusting and controlling the generation of the three-dimensional animation in real time through interactive data input by the user, and realizing the interaction between the user and the animation. By using the deep learning model, the three-dimensional animation can be automatically generated, the workload of manual modeling and animation production is reduced, meanwhile, the three-dimensional animation can be dynamically generated and adjusted according to real-time input of a user, high interactivity is provided, and the user can directly participate in the animation creation process.

Description

Technical Field

[0001] The invention relates to the technical field of interactive three-dimensional animation, in particular to an interactive three-dimensional animation generation method. Background Art

[0002] In the modern digital media and entertainment industry, interactive 3D animation has become an important form of expression, widely used in games, movies, virtual reality, education and simulation training. The traditional 3D animation production process usually involves complex modeling, animation production and post-processing, which requires a lot of manual labor and professional skills, which is not only time-consuming and labor-intensive, but also costly. With the development of technology, especially the progress in the fields of computer graphics and artificial intelligence, automated and intelligent animation generation methods have gradually become a hot topic of research.

[0003] Existing interactive 3D animation generation methods mainly rely on preset animation templates and scripts, and users create animations by selecting and combining these templates. However, the flexibility and creativity of this method are limited, and it is difficult to meet the user's needs for personalization and dynamic interaction. In addition, when dealing with complex scenes and dynamic interactions, the existing technology often requires high-performance hardware support, which limits the popularity and application of interactive 3D animation. Therefore, technicians in this field provide an interactive 3D animation generation method to solve the problems raised in the above background technology. Summary of the invention

[0004] 1. Technical issues to be resolved

[0005] In view of the deficiencies of the prior art, the present invention provides an interactive 3D animation generation method, which solves the problem that the traditional 3D animation generation method relies on manual modeling and animation production, and the process is cumbersome and time-consuming.

[0006] (II) Technical solution

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: an interactive three-dimensional animation generation method, characterized in that it includes:

[0008] Data preparation: collecting and preparing 3D animation datasets for training deep learning models, including models, animation sequences, and user interaction data;

[0009] Deep learning model training, using 3D animation datasets to train deep learning models;

[0010] Interactive animation generation, which automatically generates 3D animation sequences based on user input using a trained deep learning model;

[0011] Interactive control: through the interactive data input by the user, the generation of 3D animation can be adjusted and controlled in real time to realize the interaction between the user and the animation;

[0012] Animation optimization: optimize the generated 3D animation, including improving animation fluency, realism and interactive experience;

[0013] Animation output: output the optimized 3D animation to a display device or storage medium for users to watch and interact with.

[0014] Preferably, the data preparation comprises the following steps:

[0015] S1. Data collection: Collect 3D models from various resources, including open source model libraries, professional 3D model websites, and self-built models. Collect animation sequences of different styles and types, which can be existing animation clips or sequences produced by animation software. Record the historical data of users' interaction with 3D animations. This may need to be achieved through a specific interactive platform or application.

[0016] S2. Data cleaning: deleting damaged, incomplete or low-quality 3D models and animation sequences, ensuring that all data follows a unified format and standard for subsequent processing and model training;

[0017] S3. Data augmentation: increasing the diversity of 3D models by rotating, scaling, and translating them, and increasing the diversity of animation sequences by changing animation speed, adding noise, and adjusting lighting;

[0018] S4. Data annotation: annotate user interaction data in detail, including interaction type, timestamp, and interaction location; annotate animation sequences, and extract relevant features of key frames, action types, and expressions;

[0019] S5. Data segmentation: divide the data set into training set, validation set and test set, ensuring that the data in each set is evenly distributed and non-overlapping;

[0020] S6. Data preprocessing: preprocessing of 3D models, including mesh simplification, topology optimization and texture mapping; preprocessing of animation sequences, including key frame extraction, motion smoothing and time series normalization;

[0021] S7. Dataset construction: combine the preprocessed data into a training dataset for model training, and combine the preprocessed data into a validation and test dataset for model validation and performance evaluation;

[0022] S8. Data storage: select an appropriate format to store data, ensure that the data is easy to read and process, back up the data, and prevent data loss or damage.

[0023] Preferably, the deep learning model training comprises the following steps:

[0024] S1. Initialize model parameters. Before training begins, the model parameters need to be initialized. These parameters are usually initialized to small random values ​​to avoid gradient vanishing or exploding problems.

[0025] S2. Forward propagation, the input number passes through each layer of the model until the output layer, and the predicted value is calculated. The specific calculation formula is:

[0026] z=W·x+b

[0027] a=f(z)

[0028] Where W is the weight matrix, x is the input vector, b is the bias vector, f is the activation function, z is the weighted input, and a is the output after activation;

[0029] S3. Calculate the loss. The loss function is used to measure the difference between the model prediction value and the true value. For regression problems, the calculation formula of the mean square error loss function is:

[0030]

[0031] For classification problems, the calculation formula of the cross entropy loss function is:

[0032]

[0033] Among them, y is the true label, is the label predicted by the model, n is the sample data, and M is the number of categories;

[0034] S4. Back propagation, back propagation is to calculate the gradient of the loss function with respect to the model parameters, usually done using the chain rule. For a simple neural network layer, the gradient calculation formula of the parameter W is:

[0035]

[0036] in, is the gradient of the loss function with respect to the activation output, is the derivative of the activation function, is the gradient of the weighted input with respect to the weight;

[0037] S5. Parameter update, use gradient or its variant to update model parameters to reduce the value of the loss function. The calculation formula for parameter update is:

[0038]

[0039] Among them, W new is the updated weight, W oldis the current weight, α is the learning rate, is the gradient of the loss function with respect to the weights;

[0040] S6. Iterate the training and repeat the above steps until the performance of the model on the validation set no longer improves or the predetermined number of iterations is reached;

[0041] S7. Model evaluation, evaluate the performance of the model on the test set to determine the generalization ability of the model.

[0042] Preferably, the interactive animation generation comprises the following steps:

[0043] S1. User input processing, capturing user input, including touch, gestures, and voice commands, parsing user input, and converting it into a format that the model can understand;

[0044] S2. Animation generation, generates animations based on defined rules and user input, using machine learning models, generates animations based on user input, and calculates by using a recurrent neural network model:

[0045] h t =f(W h h t-1 +W x x t +b)

[0046] Among them, h t is the current hidden state, h t-1 is the hidden state at the previous moment, x t is the input at the current moment, W h and W x is the weight matrix, b is the bias term, and f is the activation function;

[0047] S3. Real-time response, write interactive logic, ensure that user input can correctly affect the generation and change of animation, adjust the generation of animation according to the user's real-time input, and achieve instant feedback;

[0048] S4. Animation optimization: optimize the frame rate and transition effect of animation to ensure the smoothness of animation, and enhance the realism of animation by adding lighting, shadow and material effects;

[0049] S5. Animation output: display the animation on the screen or save it as a file for users to download and share.

[0050] Preferably, the interactive control comprises the following steps:

[0051] S1. User input capture, prepare and configure input devices to capture user input and collect user input data in real time;

[0052] S2. Input data preprocessing, removing invalid or erroneous input data and converting the input data into a format that the model can process;

[0053] S3. Interaction logic implementation, defining the rules between user input and animation response, and writing code to handle user input events;

[0054] S4. Animation response, updating the state of the animation according to the user input, and adjusting the animation content to respond to the user input;

[0055] S5. Real-time feedback: design an animation feedback mechanism to generate corresponding feedback content based on user input;

[0056] S6. Interaction optimization: optimize the performance of interactive control, ensure the smoothness and real-time response of animation, optimize the interaction design and improve the user experience based on user feedback and behavior data;

[0057] S7. Interaction testing: testing whether the interactive functions work as expected, such as whether the input can correctly trigger the animation response. Invite users to participate in the test, collect feedback, and optimize the interactive design and animation content.

[0058] S8. Interaction integration: Integrate interactive control into the 3D animation system to ensure that user input can affect the animation content in real time, and publish the completed interactive animation to the target platform;

[0059] S9. Interaction monitoring and maintenance: monitor the interaction data between users and interactive animations, analyze user behaviors and preferences, and continuously update and optimize interactive controls and animation content based on monitoring data and user feedback.

[0060] Preferably, the animation optimization comprises the following steps:

[0061] S1. Performance analysis: analyze the CPU, GPU and memory usage of animation during rendering, monitor the frame rate of animation, and ensure that the animation runs smoothly without obvious lag;

[0062] S2. Optimize animation fluency, optimize the transition between key frames, ensure the smoothness of animation movements, adjust the animation timeline, and shorten the loading and execution time of animation;

[0063] S3. Improve visual effects, adjust lighting and shadows in the scene, enhance the realism and depth of animation, optimize the material and texture of the model, improve visual quality, and pay attention to maintaining reasonable resource consumption;

[0064] S4. Optimize the interactive experience, reduce the time from user interaction to animation response, improve the interactive experience, enhance visual and auditory feedback, and enable users to clearly perceive the interaction results;

[0065] S5. Code and resource optimization: optimize animation-related codes, improve code execution efficiency, compress images and audio used in animations, reduce file size, and speed up loading;

[0066] S6. Compatibility testing: testing animations on different devices and browsers to ensure compatibility and consistency, running animations on devices with different configurations to evaluate their performance;

[0067] S7. Collect user feedback: invite users to participate in the test, collect their feedback on the animation, analyze user behavior data, and understand users' usage habits and preferences for the animation;

[0068] S8. Continuously iterate and optimize animations based on performance analysis and user feedback, regularly release updated versions of animations, and introduce new optimizations and features;

[0069] S9. Documentation and guidelines: Write optimization documents, record the optimization process and results, provide reference for subsequent maintenance and updates, and provide user guides to help users better understand and use animations.

[0070] Preferably, the animation output comprises the following steps:

[0071] S1. Determine the output format. According to the target platform and user device, select the appropriate animation output format to ensure that the selected format has good compatibility and support on the target platform;

[0072] S2. Rendering the animation, setting the rendering parameters of the animation according to the requirements of the output format, using the animation software or rendering engine to perform the rendering process, and converting the animation sequence into a video file;

[0073] S3. Post-processing: editing and adjusting the rendered animation, compressing and optimizing the video file to reduce the file size while maintaining the image quality as much as possible;

[0074] S4. Audio synchronization: add background music and necessary sound effects according to the animation content to enhance the viewing experience and ensure that the audio is synchronized with the animation action to avoid the situation where the audio and the picture are out of sync;

[0075] S5. Output and test: export the processed animation to the final video file, and test the animation playback on different devices and browsers to ensure compatibility and smoothness;

[0076] S6. Publishing and distribution: Select a suitable publishing platform based on the target audience and usage scenarios of the animation, upload the animation to the selected platform, and ensure that the animation can be accessed and downloaded by users;

[0077] S7. Collect user feedback, encourage users to watch the animation and provide feedback, understand their viewing experience and opinions, analyze user feedback, and use it to improve and optimize future animation content;

[0078] S8. Animation update and maintenance: Based on user feedback and animation performance, regularly update and maintain animation content, provide necessary technical support, and solve problems encountered by users during viewing and interaction.

[0079] (III) Beneficial effects

[0080] The present invention provides an interactive three-dimensional animation generation method. It has the following beneficial effects:

[0081] 1. In the present invention, a deep learning model is used to automatically generate three-dimensional animation, reducing the workload of manual modeling and animation production. At the same time, the three-dimensional animation can be dynamically generated and adjusted according to the user's real-time input, providing a high degree of interactivity, allowing users to directly participate in the animation creation process.

[0082] 2. In the present invention, by optimizing algorithms and user interaction design, the efficiency of animation generation can be ensured, and it can run smoothly even on devices with limited resources. At the same time, the animation production process is simplified, the threshold for 3D animation production is lowered, and more people can participate in the creation of 3D animation. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0084] Figure 2 A schematic diagram of the process for preparing data in the present invention;

[0085] Figure 3 A schematic diagram of the process of deep learning model training in the present invention;

[0086] Figure 4 A schematic diagram of a process flow for interactive animation generation in the present invention;

[0087] Figure 5 It is a schematic diagram of the process of interactive control in the present invention;

[0088] Figure 6 A schematic diagram of the process of animation optimization in the present invention;

[0089] Figure 7 It is a schematic diagram of the flow of animation output in the present invention. DETAILED DESCRIPTION

[0090] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0091] Embodiment 1:

[0092] like Figure 1-7 As shown, an embodiment of the present invention provides an interactive three-dimensional animation generation method, comprising:

[0093] Data preparation: collecting and preparing 3D animation datasets for training deep learning models, including models, animation sequences, and user interaction data;

[0094] Deep learning model training, using 3D animation datasets to train deep learning models;

[0095] Interactive animation generation, which automatically generates 3D animation sequences based on user input using a trained deep learning model;

[0096] Interactive control: through the interactive data input by the user, the generation of 3D animation can be adjusted and controlled in real time to realize the interaction between the user and the animation;

[0097] Animation optimization: optimize the generated 3D animation, including improving animation fluency, realism and interactive experience;

[0098] Animation output: output the optimized 3D animation to a display device or storage medium for users to watch and interact with.

[0099] Data preparation includes the following steps:

[0100] S1. Data collection: Collect 3D models from various resources, including open source model libraries, professional 3D model websites, and self-built models. Collect animation sequences of different styles and types, which can be existing animation clips or sequences produced by animation software. Record the historical data of users' interaction with 3D animations. This may need to be achieved through a specific interactive platform or application.

[0101] S2. Data cleaning: deleting damaged, incomplete or low-quality 3D models and animation sequences, ensuring that all data follows a unified format and standard for subsequent processing and model training;

[0102] S3. Data augmentation: increasing the diversity of 3D models by rotating, scaling, and translating them, and increasing the diversity of animation sequences by changing animation speed, adding noise, and adjusting lighting;

[0103] S4. Data annotation: annotate user interaction data in detail, including interaction type, timestamp, and interaction location; annotate animation sequences, and extract relevant features of key frames, action types, and expressions;

[0104] S5. Data segmentation: divide the data set into training set, validation set and test set, ensuring that the data in each set is evenly distributed and non-overlapping;

[0105] S6. Data preprocessing: preprocessing of 3D models, including mesh simplification, topology optimization and texture mapping; preprocessing of animation sequences, including key frame extraction, motion smoothing and time series normalization;

[0106] S7. Dataset construction: combine the preprocessed data into a training dataset for model training, and combine the preprocessed data into a validation and test dataset for model validation and performance evaluation;

[0107] S8. Data storage: select an appropriate format to store data, ensure that the data is easy to read and process, back up the data, and prevent data loss or damage.

[0108] Deep learning model training consists of the following steps:

[0109] S1. Initialize model parameters. Before training begins, the model parameters need to be initialized. These parameters are usually initialized to small random values ​​to avoid gradient vanishing or exploding problems.

[0110] S2. Forward propagation, the input number passes through each layer of the model until the output layer, and the predicted value is calculated. The specific calculation formula is:

[0111] z=W·x+b

[0112] a=f(z)

[0113] Where W is the weight matrix, x is the input vector, b is the bias vector, f is the activation function, z is the weighted input, and a is the output after activation;

[0114] S3. Calculate the loss. The loss function is used to measure the difference between the model prediction value and the true value. For regression problems, the calculation formula of the mean square error loss function is:

[0115]

[0116] For classification problems, the calculation formula of the cross entropy loss function is:

[0117]

[0118] Among them, y is the true label, is the label predicted by the model, N is the sample data, and M is the number of categories;

[0119] S4. Back propagation, back propagation is to calculate the gradient of the loss function with respect to the model parameters, usually done using the chain rule. For a simple neural network layer, the gradient calculation formula of the parameter W is:

[0120]

[0121] in, is the gradient of the loss function with respect to the activation output, is the derivative of the activation function, is the gradient of the weighted input with respect to the weight;

[0122] S5. Parameter update, use gradient or its variant to update model parameters to reduce the value of the loss function. The calculation formula for parameter update is:

[0123]

[0124] Among them, W new is the updated weight, W old is the current weight, α is the learning rate, is the gradient of the loss function with respect to the weights;

[0125] S6. Iterate the training and repeat the above steps until the performance of the model on the validation set no longer improves or the predetermined number of iterations is reached;

[0126] S7. Model evaluation, evaluate the performance of the model on the test set to determine the generalization ability of the model.

[0127] Interactive animation generation includes the following steps:

[0128] S1. User input processing, capturing user input, including touch, gestures, and voice commands, parsing user input, and converting it into a format that the model can understand;

[0129] S2. Animation generation, generates animations based on defined rules and user input, using machine learning models, generates animations based on user input, and calculates by using a recurrent neural network model:

[0130] h t =f(W h h t-1 +W x x t +b)

[0131] Among them, h t is the current hidden state, h t-1 is the hidden state at the previous moment, x tis the input at the current moment, W h and W x is the weight matrix, b is the bias term, and f is the activation function;

[0132] S3. Real-time response, write interactive logic, ensure that user input can correctly affect the generation and change of animation, adjust the generation of animation according to the user's real-time input, and achieve instant feedback;

[0133] S4. Animation optimization: optimize the frame rate and transition effect of animation to ensure the smoothness of animation, and enhance the realism of animation by adding lighting, shadow and material effects;

[0134] S5. Animation output: display the animation on the screen or save it as a file for users to download and share.

[0135] Interactive control includes the following steps:

[0136] S1. User input capture, prepare and configure input devices to capture user input and collect user input data in real time;

[0137] S2. Input data preprocessing, removing invalid or erroneous input data and converting the input data into a format that the model can process;

[0138] S3. Interaction logic implementation, defining the rules between user input and animation response, and writing code to handle user input events;

[0139] S4. Animation response, updating the state of the animation according to the user input, and adjusting the animation content to respond to the user input;

[0140] S5. Real-time feedback: design an animation feedback mechanism to generate corresponding feedback content based on user input;

[0141] S6. Interaction optimization: optimize the performance of interactive control, ensure the smoothness and real-time response of animation, optimize the interaction design and improve the user experience based on user feedback and behavior data;

[0142] S7. Interaction testing: testing whether the interactive functions work as expected, such as whether the input can correctly trigger the animation response. Invite users to participate in the test, collect feedback, and optimize the interactive design and animation content.

[0143] S8. Interaction integration: Integrate interactive control into the 3D animation system to ensure that user input can affect the animation content in real time, and publish the completed interactive animation to the target platform;

[0144] S9. Interaction monitoring and maintenance: monitor the interaction data between users and interactive animations, analyze user behaviors and preferences, and continuously update and optimize interactive controls and animation content based on monitoring data and user feedback.

[0145] Animation optimization includes the following steps:

[0146] S1. Performance analysis: analyze the CPU, GPU and memory usage of animation during rendering, monitor the frame rate of animation, and ensure that the animation runs smoothly without obvious lag;

[0147] S2. Optimize animation fluency, optimize the transition between key frames, ensure the smoothness of animation movements, adjust the animation timeline, and shorten the loading and execution time of animation;

[0148] S3. Improve visual effects, adjust lighting and shadows in the scene, enhance the realism and depth of animation, optimize the material and texture of the model, improve visual quality, and pay attention to maintaining reasonable resource consumption;

[0149] S4. Optimize the interactive experience, reduce the time from user interaction to animation response, improve the interactive experience, enhance visual and auditory feedback, and enable users to clearly perceive the interaction results;

[0150] S5. Code and resource optimization: optimize animation-related codes, improve code execution efficiency, compress images and audio used in animations, reduce file size, and speed up loading;

[0151] S6. Compatibility testing: testing animations on different devices and browsers to ensure compatibility and consistency, running animations on devices with different configurations to evaluate their performance;

[0152] S7. Collect user feedback: invite users to participate in the test, collect their feedback on the animation, analyze user behavior data, and understand users' usage habits and preferences for the animation;

[0153] S8. Continuously iterate and optimize animations based on performance analysis and user feedback, regularly release updated versions of animations, and introduce new optimizations and features;

[0154] S9. Documentation and guidelines: Write optimization documents, record the optimization process and results, provide reference for subsequent maintenance and updates, and provide user guides to help users better understand and use animations.

[0155] Animation output includes the following steps:

[0156] S1. Determine the output format. According to the target platform and user device, select the appropriate animation output format to ensure that the selected format has good compatibility and support on the target platform;

[0157] S2. Rendering the animation, setting the rendering parameters of the animation according to the requirements of the output format, using the animation software or rendering engine to perform the rendering process, and converting the animation sequence into a video file;

[0158] S3. Post-processing: editing and adjusting the rendered animation, compressing and optimizing the video file to reduce the file size while maintaining the image quality as much as possible;

[0159] S4. Audio synchronization: add background music and necessary sound effects according to the animation content to enhance the viewing experience and ensure that the audio is synchronized with the animation action to avoid the situation where the audio and the picture are out of sync;

[0160] S5. Output and test: export the processed animation to the final video file, and test the animation playback on different devices and browsers to ensure compatibility and smoothness;

[0161] S6. Publishing and distribution: Select a suitable publishing platform based on the target audience and usage scenarios of the animation, upload the animation to the selected platform, and ensure that the animation can be accessed and downloaded by users;

[0162] S7. Collect user feedback, encourage users to watch the animation and provide feedback, understand their viewing experience and opinions, analyze user feedback, and use it to improve and optimize future animation content;

[0163] S8. Animation update and maintenance: Based on user feedback and animation performance, regularly update and maintain animation content, provide necessary technical support, and solve problems encountered by users during viewing and interaction.

[0164] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An interactive three-dimensional animation generation method, characterized in that: include: Data preparation: collecting and preparing 3D animation datasets for training deep learning models, including models, animation sequences, and user interaction data; Deep learning model training, using 3D animation datasets to train deep learning models; Interactive animation generation, based on user input, automatically generates 3D animation sequences using a trained deep learning model; Interactive control: through the interactive data input by the user, the generation of 3D animation is adjusted and controlled in real time to realize the interaction between the user and the animation; Animation optimization: optimize the generated 3D animation, including improving animation fluency, realism and interactive experience; Animation output: output the optimized 3D animation to a display device or storage medium for users to watch and interact with.

2. The interactive three-dimensional animation generation method according to claim 1, characterized in that: The data preparation includes the following steps: S1. Data collection: Collect 3D models from various resources, including open source model libraries, professional 3D model websites, and self-built models. Collect animation sequences of different styles and types, which can be existing animation clips or sequences produced by animation software. Record the historical data of users' interaction with 3D animations. This may need to be achieved through a specific interactive platform or application. S2. Data cleaning: deleting damaged, incomplete or low-quality 3D models and animation sequences, ensuring that all data follows a unified format and standard for subsequent processing and model training; S3. Data augmentation: increasing the diversity of 3D models by rotating, scaling, and translating them, and increasing the diversity of animation sequences by changing animation speed, adding noise, and adjusting lighting; S4. Data annotation: annotate user interaction data in detail, including interaction type, timestamp, and interaction location; annotate animation sequences, and extract relevant features of key frames, action types, and expressions; S5. Data segmentation: divide the data set into training set, validation set and test set, ensuring that the data in each set is evenly distributed and non-overlapping; S6. Data preprocessing: preprocessing of 3D models, including mesh simplification, topology optimization and texture mapping; preprocessing of animation sequences, including key frame extraction, motion smoothing and time series normalization; S7. Dataset construction: combine the preprocessed data into a training dataset for model training, and combine the preprocessed data into a validation and test dataset for model validation and performance evaluation; S8. Data storage: select an appropriate format to store data, ensure that the data is easy to read and process, back up the data, and prevent data loss or damage.

3. The interactive three-dimensional animation generation method according to claim 1, characterized in that: The deep learning model training includes the following steps: S1. Initialize model parameters. Before training begins, the model parameters need to be initialized. These parameters are usually initialized to small random values ​​to avoid gradient vanishing or exploding problems. S2. Forward propagation, the input number passes through each layer of the model until the output layer, and the predicted value is calculated. The specific calculation formula is: z=W·x+b a=f(z) Where W is the weight matrix, x is the input vector, b is the bias vector, f is the activation function, z is the weighted input, and a is the output after activation; S3. Calculate the loss. The loss function is used to measure the difference between the model prediction value and the true value. For regression problems, the calculation formula of the mean square error loss function is: For classification problems, the calculation formula of the cross entropy loss function is: Among them, y is the true label, is the label predicted by the model, N is the sample data, and M is the number of categories; S4. Back propagation, back propagation is to calculate the gradient of the loss function with respect to the model parameters, usually done using the chain rule. For a simple neural network layer, the gradient calculation formula of the parameter W is: in, is the gradient of the loss function with respect to the activation output, is the derivative of the activation function, is the gradient of the weighted input with respect to the weight; S5. Parameter update, use gradient or its variant to update model parameters to reduce the value of the loss function. The calculation formula for parameter update is: Among them, W new is the updated weight, W old is the current weight, α is the learning rate, is the gradient of the loss function with respect to the weights; S6. Iterate the training and repeat the above steps until the performance of the model on the validation set no longer improves or the predetermined number of iterations is reached; S7. Model evaluation, evaluate the performance of the model on the test set to determine the generalization ability of the model.

4. The interactive three-dimensional animation generation method according to claim 1, characterized in that: The interactive animation generation comprises the following steps: S1. User input processing, capturing user input, including touch, gestures, and voice commands, parsing user input, and converting it into a format that the model can understand; S2. Animation generation, generates animations based on defined rules and user input, using machine learning models, generates animations based on user input, and calculates by using a recurrent neural network model: h t =f(W h h t-1 +W x x t +b) Among them, h t is the current hidden state, h t-1 is the hidden state at the previous moment, x t is the input at the current moment, W h and W x is the weight matrix, b is the bias term, and f is the activation function; S3. Real-time response, write interactive logic, ensure that user input can correctly affect the generation and change of animation, adjust the generation of animation according to the user's real-time input, and achieve instant feedback; S4. Animation optimization: optimize the frame rate and transition effect of animation to ensure the smoothness of animation, and enhance the realism of animation by adding lighting, shadow and material effects; S5. Animation output: display the animation on the screen or save it as a file for users to download and share.

5. The interactive three-dimensional animation generation method according to claim 1, characterized in that: The interactive control comprises the following steps: S1. User input capture, prepare and configure input devices to capture user input and collect user input data in real time; S2. Input data preprocessing, removing invalid or erroneous input data and converting the input data into a format that the model can process; S3. Interaction logic implementation, defining the rules between user input and animation response, and writing code to handle user input events; S4. Animation response, updating the state of the animation according to the user input, and adjusting the animation content to respond to the user input; S5. Real-time feedback: design an animation feedback mechanism to generate corresponding feedback content based on user input; S6. Interaction optimization: optimize the performance of interactive control, ensure the smoothness and real-time response of animation, optimize the interaction design and improve the user experience based on user feedback and behavior data; S7. Interaction testing: testing whether the interactive functions work as expected, such as whether the input can correctly trigger the animation response. Invite users to participate in the test, collect feedback, and optimize the interactive design and animation content. S8. Interaction integration: Integrate interactive control into the 3D animation system to ensure that user input can affect the animation content in real time, and publish the completed interactive animation to the target platform; S9. Interaction monitoring and maintenance: monitor the interaction data between users and interactive animations, analyze user behaviors and preferences, and continuously update and optimize interactive controls and animation content based on monitoring data and user feedback.

6. The interactive three-dimensional animation generation method according to claim 1, characterized in that: The animation optimization comprises the following steps: S1. Performance analysis: analyze the CPU, GPU and memory usage of animation during rendering, monitor the frame rate of animation, and ensure that the animation runs smoothly without obvious lag; S2. Optimize animation fluency, optimize the transition between key frames, ensure the smoothness of animation movements, adjust the animation timeline, and shorten the loading and execution time of animation; S3. Improve visual effects, adjust lighting and shadows in the scene, enhance the realism and depth of animation, optimize the material and texture of the model, improve visual quality, and pay attention to maintaining reasonable resource consumption; S4. Optimize the interactive experience, reduce the time from user interaction to animation response, improve the interactive experience, enhance visual and auditory feedback, and enable users to clearly perceive the interaction results; S5. Code and resource optimization: optimize animation-related codes, improve code execution efficiency, compress images and audio used in animations, reduce file size, and speed up loading; S6. Compatibility testing: testing animations on different devices and browsers to ensure compatibility and consistency, running animations on devices with different configurations to evaluate their performance; S7. Collect user feedback: invite users to participate in the test, collect their feedback on the animation, analyze user behavior data, and understand users' usage habits and preferences for the animation; S8. Continuously iterate and optimize animations based on performance analysis and user feedback, regularly release updated versions of animations, and introduce new optimizations and features; S9. Documentation and guidelines: Write optimization documents, record the optimization process and results, provide reference for subsequent maintenance and updates, and provide user guides to help users better understand and use animations.

7. The interactive three-dimensional animation generation method according to claim 1, characterized in that: The animation output comprises the following steps: S1. Determine the output format. According to the target platform and user device, select the appropriate animation output format to ensure that the selected format has good compatibility and support on the target platform; S2. Rendering the animation, setting the rendering parameters of the animation according to the requirements of the output format, using the animation software or rendering engine to perform the rendering process, and converting the animation sequence into a video file; S3. Post-processing: editing and adjusting the rendered animation, compressing and optimizing the video file to reduce the file size while maintaining the image quality as much as possible; S4. Audio synchronization: add background music and necessary sound effects according to the animation content to enhance the viewing experience and ensure that the audio is synchronized with the animation action to avoid the situation where the audio and the picture are out of sync; S5. Output and test: export the processed animation to the final video file, and test the animation playback on different devices and browsers to ensure compatibility and smoothness; S6. Publishing and distribution: Select a suitable publishing platform based on the target audience and usage scenarios of the animation, upload the animation to the selected platform, and ensure that the animation can be accessed and downloaded by users; S7. Collect user feedback, encourage users to watch the animation and provide feedback, understand their viewing experience and opinions, analyze user feedback, and use it to improve and optimize future animation content; S8. Animation update and maintenance: Based on user feedback and animation performance, regularly update and maintain animation content, provide necessary technical support, and solve problems encountered by users during viewing and interaction.

Images