Virtual AI modeling method based on CPU picture processing

Abstract

The invention relates to a virtual AI (artificial intelligence) modeling method based on CPU (central processing unit) picture processing, which is characterized by comprising the following steps: S1, training and deploying a deep learning model; s2, generating an adversarial network GANs; s3, a variational auto-encoder VAEs is set; s4, establishing an autoregression model; s5, a Visual GLM model is established; and S6, environment installation and deployment. The method has the advantages that the generation speed is high, the generated model can be flexibly modified and finely adjusted, and the method is more flexible and convenient to use.

Description

Technical Field

[0001] The present invention relates to the field of AI modeling technology, and in particular to a virtual AI modeling method based on CPU image processing. Background Art

[0002] AI modeling refers to the use of artificial intelligence technology to quickly generate and optimize three-dimensional models. Compared with traditional manual modeling, AI modeling is faster, more efficient, more accurate, and can automatically process complex geometries and topological structures. AI modeling is usually divided into the following steps:

[0003] 1. Data collection: Collect and prepare data sets, including cleaning and transforming data into the format required by AI models.

[0004] 2. Feature extraction: Feature extraction refers to extracting useful feature vectors from raw data. Usually some machine learning algorithms such as decision trees, random forests, neural networks, etc. are needed to determine the extracted features.

[0005] 3. Model selection: Select a suitable model algorithm, determine the model type (such as decision tree, neural network, decision boundary, etc.), and choose different algorithms according to specific problems.

[0006] 4. Model training: Use the data set to train the model and adjust the model parameters based on the training results to make it better fit the data set.

[0007] 5. Model evaluation: Use the test set to evaluate the model performance and adjust the model to improve its performance.

[0008] 6. Deploy the model: Deploy the model to the production environment to solve practical problems.

[0009] When the existing AI modeling method model is established, the model image cannot be generated quickly and accurately, and the model is not easy to flexibly modify and fine-tune. Summary of the invention

[0010] 1. Technical Problems Solved

[0011] The technical problem to be solved by the present invention is that the existing model generation method has a poor model generation speed and cannot flexibly modify and fine-tune the generated model.

[0012] 2. Technical Solution

[0013] In order to solve the above technical problems, the present invention provides a virtual AI modeling method based on CPU image processing, which is characterized by comprising the following steps:

[0014] S1. Deep learning model training and deployment. To use deep learning models for image processing and virtual AI modeling, you first need to train the model to recognize and imitate the laws of human visual art. Use a large amount of image data for training so that the computer can learn how to generate new images or perform virtual modeling. The model training and deployment can be performed on the CPU.

[0015] S2. Generative Adversarial Networks (GANs). GANs is a generative model based on game theory. The generator tries to generate realistic images, while the discriminator distinguishes between real images and generated images. Through continuous adversarial training, GANs can generate high-quality images.

[0016] S3. Variational Autoencoders (VAEs). VAEs is a generative model based on probabilistic encoding and decoding. The encoder maps the input image to the distribution of the latent space, and the decoder regenerates the image from the distribution of the latent space. VAEs performs well in image generation and can also be used for tasks such as image reconstruction and image interpolation.

[0017] S4, autoregressive model, the autoregressive model generates images through conditional probability, regards the pixels of the image as a sequence, and generates images by modeling each pixel;

[0018] S5, VisualGLM model. VisualGLM is an open source conversation model that can process Chinese and English conversations and images. It has up to 7.8 billion parameters and provides powerful language and visual processing capabilities, demonstrating its comprehensive application capabilities in vision and language understanding.

[0019] S6. Environment installation and deployment. Performing these calculations on the CPU requires a suitable software environment and hardware support. Using the Ubuntu 22.04 operating system, sufficient video memory is required to deploy these models. For model training and deployment, a dedicated tool library is required to support flexible modification and efficient fine-tuning of the model.

[0020] Furthermore, the GANs in S2 are composed of a generator and a discriminator.

[0021] Furthermore, the VAEs model in S3 consists of an encoder and a decoder.

[0022] Furthermore, the autoregressive model in S4 includes PixelCNN and PixelRNN.

[0023] Furthermore, the VisualGLM model in S5 is used for image description and knowledge question answering tasks.

[0024] Furthermore, the database for model training and deployment in S6 adopts SwissArmyTransformer.

[0025] 3. Beneficial Effects

[0026] The advantages of the present invention compared with the prior art are:

[0027] 1. Ability to autonomously learn how to generate new images or perform virtual modeling;

[0028] 2. Images can be regenerated from the distribution of the latent space and can also be used for tasks such as image reconstruction and image interpolation;

[0029] 3. Supports flexible modification and efficient fine-tuning of the generated model, making it more flexible to use. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present invention are described clearly and completely below. 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.

[0031] It should be noted that all directional indications such as up, down, left, right, front, back, etc. in the embodiments of the present invention are only used to explain the relative position relationship, movement status, etc. between the components in a certain specific posture. If the specific posture changes, the directional indication will also change accordingly.

[0032] In addition, in the present invention, descriptions such as "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0033] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the invention.

[0034] The present invention will now be further described.

[0035] The present invention provides a virtual AI modeling method based on CPU image processing, comprising the following steps:

[0036] S1. Deep learning model training and deployment. To use deep learning models for image processing and virtual AI modeling, you first need to train the model to recognize and imitate the laws of human visual art. Use a large amount of image data for training so that the computer can learn how to generate new images or perform virtual modeling. The model training and deployment can be performed on the CPU.

[0037] S2. Generative Adversarial Networks (GANs). GANs is a generative model based on game theory. GANs consists of a generator and a discriminator. The generator tries to generate realistic images, while the discriminator distinguishes between real images and generated images. Through continuous adversarial training, GANs can generate high-quality images.

[0038] S3. Variational Autoencoders (VAEs). VAEs is a generative model based on probabilistic encoding and decoding. The VAEs model consists of an encoder and a decoder. The encoder maps the input image to the distribution of the latent space, and the decoder regenerates the image from the distribution of the latent space.

[0039] S4. Autoregressive model. The autoregressive model generates images through conditional probability. It regards the pixels of the image as a sequence and generates images by modeling each pixel. The autoregressive model includes PixelCNN and PixelRNN.

[0040] S5, VisualGLM model. VisualGLM is an open source dialogue model that can process Chinese and English dialogues and images. It has up to 7.8 billion parameters, provides powerful language and visual processing capabilities, and demonstrates its comprehensive application capabilities in vision and language understanding. The VisualGLM model is used for image description and knowledge question answering tasks.

[0041] S6. Environment installation and deployment. Performing these calculations on the CPU requires a suitable software environment and hardware support. Using the Ubuntu 22.04 operating system, sufficient video memory is required to deploy these models. For model training and deployment, a special tool library is required. The database for model training and deployment uses SwissArmyTransformer to support flexible modification and efficient fine-tuning of the model.

[0042] The invention and its implementation methods are described, and such description is not restrictive, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and implementation method similar to the technical solution without creativity, which should fall within the protection scope of the present invention.

Claims

1. A virtual AI modeling method based on CPU image processing, characterized in that: The following steps are involved: S1. Deep learning model training and deployment. To use deep learning models for image processing and virtual AI modeling, you first need to train the model to recognize and imitate the laws of human visual art. Use a large amount of image data for training so that the computer can learn how to generate new images or perform virtual modeling. The model training and deployment can be performed on the CPU. S2. Generative Adversarial Networks (GANs). GANs is a generative model based on game theory. The generator tries to generate realistic images, while the discriminator distinguishes between real images and generated images. Through continuous adversarial training, GANs can generate high-quality images. S3. Variational Autoencoders (VAEs). VAEs is a generative model based on probabilistic encoding and decoding. The encoder maps the input image to the distribution of the latent space, and the decoder regenerates the image from the distribution of the latent space. S4, autoregressive model, the autoregressive model generates images through conditional probability, regards the pixels of the image as a sequence, and generates images by modeling each pixel; S5, VisualGLM model. VisualGLM is an open source conversation model that can process Chinese and English conversations and images. It has up to 7.8 billion parameters and provides powerful language and visual processing capabilities, demonstrating its comprehensive application capabilities in vision and language understanding. S6. Environment installation and deployment. Performing these calculations on the CPU requires a suitable software environment and hardware support. Using the Ubuntu 22.04 operating system, sufficient video memory is required to deploy these models. For model training and deployment, a dedicated tool library is required to support flexible modification and efficient fine-tuning of the model.

2. According to claim 1, a virtual AI modeling method based on CPU image processing is characterized in that: The GANs in S2 are composed of a generator and a discriminator.

3. According to claim 1, a virtual AI modeling method based on CPU image processing is characterized in that: The VAEs model in S3 consists of an encoder and a decoder.

4. The method for virtual AI modeling based on CPU image processing according to claim 1, characterized in that: The autoregressive models in S4 include PixelCNN and PixelRNN.

5. The method for virtual AI modeling based on CPU image processing according to claim 1, characterized in that: The VisualGLM model in S5 is used for image description and knowledge question answering tasks.

6. The method for virtual AI modeling based on CPU image processing according to claim 1, characterized in that: The database for model training and deployment in S6 adopts SwissArmyTransformer.