NFT generation method for guaranteeing uniqueness of digital image

Abstract

The invention discloses an NFT generation method for guaranteeing the uniqueness of a digital image. The method comprises the following steps: 1, obtaining user information B of a user on a block chain; step 2, coding the user information B through a Hash algorithm F to generate a Hash value H; step 3, loading a dual-module generation model C1; 4, extracting feature vectors from the NFT data set through an encoder module in the dual-module generation model C1; 5, using the hash value H to match the feature vector to obtain a new feature vector as the input of a decoder module in the dual-module generation model C1; step 6, generating an image I through a decoder module in the dual-module generation model C1; and step 7, storing the image I in a decentralized network, generating a credential CID, and casting an NFT. The method focuses on enhancing protection of uniqueness of digital assets, realizes efficient extraction of image semantic information, effectively reduces the storage cost, and effectively solves the problems of NFT copying and homogenization.

Description

Technical Field

[0001] The present invention relates to the field of NFT generation and storage technology, and in particular to an NFT generation method that ensures the uniqueness of a digital image. Background Art

[0002] The recent boom in blockchain and decentralized network technologies has triggered an unprecedented wave of growth in the digital economy and provided a new platform for the creation, trading, and management of digital assets. Against this backdrop, non-fungible tokens (NFTs) have rapidly become a key tool for representing the ownership and uniqueness of digital assets. However, while NFTs offer numerous advantages in promoting digital assets, increasing artist compensation, and fostering new business models, several key challenges remain. The uniqueness and non-fungibility of users' digital assets cannot be guaranteed; high gas fees and energy consumption make NFT transactions prohibitively expensive, limiting their widespread adoption; and reliance on external storage systems (such as IPFS) carries the risk of instability and data loss. Reducing reliance on external servers or storage systems like IPFS and adopting more efficient chain data management technologies that reduce gas fees will effectively mitigate the risks associated with unstable storage. Summary of the Invention

[0003] The present invention is proposed in view of the problems existing in the existing NFT generation method for ensuring the uniqueness of digital images. Therefore, the problem to be solved by the present invention is how to provide an NFT generation method for ensuring the uniqueness of digital images.

[0004] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides an NFT generation method for ensuring the uniqueness of a digital image, which includes the following steps: Step 1: obtaining user information B of a user on a blockchain; Step 2: Encode user information B using hash algorithm F to generate hash value H; Step 3: Load the dual-module generation model C1; Step 4: Extract feature vectors from the NFT image dataset through the encoder module in the dual-module generative model C1; Step 5: Use the hash value H to match the feature vector to obtain a new feature vector as the input of the decoder module in the dual-module generation model C1; Step 6: Generate image I through the decoder module in the dual-module generative model C1; Step 7: Store image I on the decentralized network, generate a certificate CID, and mint NFT.

[0005] As a preferred solution of the NFT generation method for ensuring the uniqueness of digital images described in the present invention, the user information B is several types of digital user information that can be encoded using a hash algorithm F, and is the unique information of the user, and the method for obtaining the user information B is that the user actively uploads it.

[0006] As a preferred solution of the NFT generation method for ensuring the uniqueness of digital images described in the present invention, the hash algorithm F includes the MD5 algorithm, the SHA-1 algorithm, the SHA-256 algorithm and various hash algorithms whose output can be converted into binary strings.

[0007] As a preferred solution of the NFT generation method for ensuring the uniqueness of digital images described in the present invention, the specific steps of encoding the user information B by the hash algorithm F are: first, preprocessing the collected user information B, specifically including converting the digital information into a string form, using a unified encoding for the text information, and removing redundant spaces or invalid characters, then, splicing and combining the preprocessed user information according to a fixed format, and finally, converting the spliced ​​string into a byte sequence according to the unified encoding as the input of the hash algorithm, calculating the hash value, and obtaining a fixed-length and unique encoding result.

[0008] As a preferred solution of the NFT generation method for ensuring the uniqueness of digital images described in the present invention, the dual-module generation model C1 includes an encoder module and a decoder module. The encoder module extracts deep semantic features from the NFT image based on a convolutional neural network and performs feature vector conversion. The decoder module generates a unique image based on the feature vector and combined with the hash value H, and the hash value H is set to m bits.

[0009] As a preferred solution of the NFT generation method for ensuring the uniqueness of digital images described in the present invention, the specific steps of loading the dual-module generation model C1 are: defining the model structure; instantiating the model object, and loading the weights of the encoder and decoder respectively; completing the loading, loading the dual-module generation model C1 and entering the inference or training mode.

[0010] As a preferred solution of the NFT generation method for ensuring the uniqueness of digital images described in the present invention, the feature vector conversion includes: selecting the NFT image data set to be used; using a convolutional neural network to capture various features of the image and outputting a feature vector.

[0011] As a preferred solution of the NFT generation method for ensuring the uniqueness of digital images described in the present invention, the step of using the hash value H to match the feature vector is: assuming that the obtained feature vector is n bits, if n is less than m, then take the first n bits of the hash value H; if n is greater than m, then repeat the hash value H multiple times until it is greater than n bits, and take the first n bits of the value, and this n-bit encoding value is named D; then add D to the feature vector to obtain a new n-bit feature vector.

[0012] As a preferred solution of the NFT generation method for ensuring the uniqueness of digital images described in the present invention, the steps of generating image I are: first, converting the input feature vector to StyleSpace as the modulation parameter of the subsequent synthesis layer; then the synthesis layer gradually refines the feature representation; finally, the high-dimensional feature vector mapping obtained after multi-layer synthesis conversion is decoded into RGB pixel values ​​that conform to the spatial structure, completing the conversion from abstract features to specific images.

[0013] As a preferred solution of the NFT generation method for ensuring the uniqueness of digital images described in the present invention, the specific steps of casting NFT are: first, hash verification is performed on the personalized generated image, and a globally unique asset fingerprint is generated using a hash algorithm; then, according to the ERC-721 extension standard, a metadata package containing key attributes is constructed, and then the casting function in the smart contract is called to submit the metadata package to the blockchain network; finally, the smart contract initializes the ownership of the NFT to the creator's wallet address.

[0014] The beneficial effects of the present invention are: the present invention proposes a new decentralized NFT storage and verification framework, focusing on strengthening the protection of the uniqueness of digital assets; through image inversion and reconstruction technology, efficient extraction of image semantic information is achieved, effectively reducing storage costs; the present invention designs a personalized image editing algorithm, combined with the hash value generated by blockchain user information, effectively solving the problems of NFT duplication and homogenization. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 is a flow chart of the method of the present invention; Figure 2 This is a simplified diagram of the generative neural network algorithm. DETAILED DESCRIPTION

[0017] To make the above-mentioned objects, features, and advantages of the present invention more easily understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0018] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0019] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it constitute a separate or selective embodiment that is mutually exclusive with other embodiments.

[0020] Reference Figure 1 This embodiment provides a method for generating an NFT that ensures the uniqueness of a digital image, including the following steps: Step 1: Obtain the user's user information B on the blockchain. User information B is a number of digital user information that can be encoded using the hash algorithm F and is unique to the user. Step 2: Encode the user information B using a hash algorithm F to generate a hash value H. The hash algorithm F includes various hash algorithms whose output can be converted into a binary string, such as the MD5 algorithm, the SHA-1 algorithm, and the SHA-256 algorithm. Step 3: Load the dual-module generation model C1; Step 4: Extract compact feature vectors from the NFT dataset through the encoder module in the dual-module generative model C1; Step 5: Use the hash value H to match the feature vector to obtain a new feature vector as the input of the decoder module in the dual-module generation model C1; Step 6: Generate image I through the decoder module in the dual-module generative model C1; Step 7: Store image I in the decentralized network, generate a certificate CID, and mint NFT. After minting NFT, the authenticity of the NFT asset needs to be verified.

[0021] User information B is several types of digital user information that can be encoded using the hash algorithm F, and is the unique information of the user. The method for obtaining user information B is that the user actively uploads it.

[0022] The dual-module generation model C1 includes an encoder module and a decoder module. The encoder module extracts deep semantic features from NFT images based on a convolutional neural network and converts them into compact feature vectors. The decoder module takes the feature vector as input and combines it with the hash value H to generate a unique image based on the StyleGan series model.

[0023] The specific steps for loading the dual-module generative model C1 are as follows: First, define the model structure (including the encoder and decoder modules). Then, instantiate the model object and load the encoder and decoder weights separately. Finally, after loading, the model enters inference or training mode.

[0024] The steps for the encoder to extract feature vectors are: first, select the NFT image dataset to be used; then use a convolutional neural network (such as ResNet, DenseNet, etc.) to capture various image features and output feature vectors.

[0025] The specific steps for encoding user information using a hash algorithm are as follows: First, preprocess the collected user information B to ensure that the formats of various information are unified and standardized. This includes converting digital information into string form, using a unified encoding (such as UTF-8) for text information, and removing redundant spaces or invalid characters to ensure input consistency and standardization. Next, the preprocessed user information is concatenated and combined according to a fixed format, for example, "Username: <user>","Mail: <user>@example.com", "Mobile number:138****0000" are concatenated into the string "Username: <user>Email: <user>@example.com|Phone Number:138****0000", where "|" serves as a delimiter to ensure clarity and uniqueness of the information structure. Finally, the concatenated string is converted into a byte sequence using a unified encoding (such as UTF-8). This is used as input for the hash algorithm, which calculates the hash value to produce a fixed-length, unique encoding result.

[0026] The hash value H is set to m bits. The steps for matching the hash value H with the feature vector are as follows: assuming that the obtained feature vector is n bits, if n is less than m, take the first n bits of the hash value H; if n is greater than m, repeat the hash value H multiple times until it is greater than n bits, and take the first n bits of the value. This n-bit encoding value is named D; then the encoding value D is added to the feature vector bit by bit to obtain a new n-bit feature vector.

[0027] The steps for generating an image from a new feature vector are as follows: First, the resulting new feature vector is mapped to StyleSpace via an affine transformation (implemented by an MLP), converting it into a style vector that serves as a modulation parameter for the subsequent synthesis network. StyleSpace is the latent space used in the StyleGAN family of models to control the style of image generation. Its vectors modulate the convolutional weights of the synthesis network, achieving multi-scale style control of the generated image. The multi-layer convolutional neural network in the StyleGAN synthesis network then generates image features step by step from low to high resolution. At each layer, the style vector modulates the convolution kernel weights using adaptive instance normalization or weight demodulation techniques to control the feature style at the current resolution. Finally, the high-dimensional feature vectors obtained after these multi-layer synthesis transformations are mapped and decoded into RGB pixel values ​​that conform to the spatial structure, completing the transformation from abstract features to concrete images.

[0028] When verifying the authenticity of NFT assets, the smart contract first reads the semantic vector and model hash stored on the chain. Then the oracle network passes the data to the trusted execution environment. The oracle node loads the corresponding version of the StyleGAN decoder, reconstructs the image and calculates its hash value. The reconstructed hash is compared with the hash value recorded during minting. If they match, the verification is passed.

[0029] The specific steps for minting an NFT are as follows: First, a hash verification is performed on the personalized image, using a hash algorithm to generate a globally unique asset fingerprint, ensuring the uniqueness of the NFT asset. Next, a metadata package containing key attributes is constructed according to the ERC-721 extension standard (or other smart contract standards such as ERC-20 and ERC-1155). These include information such as the decentralized storage credential CID, the model version fingerprint, and the creation timestamp. The minting function (such as MintWithProof) in the smart contract is then called to submit the metadata package to the blockchain network. Finally, the smart contract initializes the NFT's ownership to the creator's wallet address, officially minting the NFT on the blockchain and ensuring uniqueness, immutability, and tradability.

[0030] like Figure 2 As shown in the figure, this is a simplified version of the generative neural network algorithm diagram, which is used to further explain the technical solution described in the summary of the invention. Assume that the feature vector W output by the encoder is 8-dimensional, recorded as (w7, w6, w5, w4, w3, w2, w1, w0). In actual applications, the dimension of the output feature vector is much larger than 8 dimensions. This is simplified here for detailed explanation. Therefore, the number of bits required to obtain the encoding value D from the hash value H should also be 8 bits; Figure 2 In the example, the value of D is recorded bit by bit as D7D6D5D4D3D2D1D0. Therefore, according to the technical solution described in the Summary of the Invention, the n-bit code value D obtained by hashing the value H is added one-to-one to each dimension of the feature vector to generate a new feature vector. The new feature vector is (w7+D7,w6+D6,w5+D5,w4+D4,w3+D3,w2+D2,w1+D1,w0+D0). This new feature vector is used as the input to the decoder module in the dual-module generative model C1, thereby obtaining the user's unique image output.

[0031] Since the encoding value D depends on the hash value H of the user information, and the hash value H is unique for different users, the encoding value D of different users is also unique, so the generated new feature vector is also unique, thereby ensuring the uniqueness of the image I generated by the user through the dual-module generation model C1.

[0032] In summary, the present invention proposes a new decentralized NFT storage and verification framework, focusing on strengthening the protection of the uniqueness of digital assets; through image inversion and reconstruction technology, efficient extraction of image semantic information is achieved, effectively reducing storage costs; the present invention designs a personalized image editing algorithm, combined with the hash value generated by blockchain user information, effectively solving the problems of NFT duplication and homogenization. The present invention can ensure the uniqueness of user digital assets in a decentralized network; the present invention is highly applicable and can be applied to all convolutional neural network architectures, and all images generated by all users are unique and confirmed.

[0033] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.< / user> < / user> < / user> < / user>

Claims

1. A method for generating an NFT that ensures the uniqueness of a digital image, characterized by: include, Step 1: Obtain the user's user information B on the blockchain; Step 2: Encode user information B using hash algorithm F to generate hash value H; Step 3: Load the dual-module generation model C1; Step 4: Extract feature vectors from the NFT image dataset through the encoder module in the dual-module generative model C1; Step 5: Use the hash value H to match the feature vector to obtain a new feature vector as the input of the decoder module in the dual-module generation model C1; Step 6: Generate image I through the decoder module in the dual-module generative model C1; Step 7: Store image I on the decentralized network, generate a certificate CID, and mint NFT.

2. The NFT generation method for ensuring the uniqueness of a digital image according to claim 1, characterized in that: The user information B is several types of digital user information that can be encoded using the hash algorithm F, and is the unique information of the user. The method for obtaining the user information B is that the user actively uploads it.

3. The NFT generation method for ensuring the uniqueness of a digital image according to claim 1, characterized in that: The hash algorithm F includes the MD5 algorithm, the SHA-1 algorithm, the SHA-256 algorithm, and various hash algorithms whose output can be converted into a binary string.

4. The NFT generation method for ensuring the uniqueness of a digital image according to claim 1, characterized in that: The specific steps of encoding the user information B by the hash algorithm F are as follows: first, the collected user information B is preprocessed, specifically including converting the digital information into a string form, using a unified encoding for the text information, and removing redundant spaces or invalid characters; then, the preprocessed user information is spliced ​​and combined according to a fixed format; finally, the spliced ​​string is converted into a byte sequence according to the unified encoding as the input of the hash algorithm, and the hash value is calculated to obtain a fixed-length and unique encoding result.

5. The NFT generation method for ensuring the uniqueness of a digital image according to claim 1, characterized in that: The dual-module generation model C1 includes an encoder module and a decoder module. The encoder module extracts deep semantic features from the NFT image based on a convolutional neural network and performs feature vector conversion. The decoder module generates a unique image based on the feature vector and combined with a hash value H, and the hash value H is set to m bits.

6. The NFT generation method for ensuring the uniqueness of a digital image according to claim 1, characterized in that: The specific steps of loading the dual-module generation model C1 are: defining the model structure; instantiating the model object, and loading the weights of the encoder and decoder respectively; completing the loading, and loading the dual-module generation model C1 into the inference or training mode.

7. The NFT generation method for ensuring uniqueness of a digital image according to claim 5, characterized in that: The feature vector conversion includes: selecting an NFT image dataset to be used; using a convolutional neural network to capture various features of the image and outputting a feature vector.

8. The NFT generation method for ensuring the uniqueness of a digital image according to claim 1, characterized in that: The steps of using the hash value H to match the feature vector are as follows: assuming that the obtained feature vector is n bits, if n is less than m, then take the first n bits of the hash value H; if n is greater than m, then repeat the hash value H multiple times until it is greater than n bits, and take the first n bits of the value. This n-bit code value is named D; then add D to the feature vector to obtain a new n-bit feature vector.

9. The NFT generation method for ensuring the uniqueness of a digital image according to claim 1, characterized in that: The steps for generating image I are as follows: first, convert the input feature vector to StyleSpace as the modulation parameter of the subsequent synthesis layer; then the synthesis layer gradually refines the feature representation; finally, the high-dimensional feature vector mapping obtained after the multi-layer synthesis conversion is decoded into RGB pixel values ​​that conform to the spatial structure, completing the conversion from abstract features to concrete images.

10. The NFT generation method for ensuring the uniqueness of a digital image according to claim 1, characterized in that: The specific steps of minting NFT are: first, hash verification is performed on the personalized generated image, and a globally unique asset fingerprint is generated using a hash algorithm; then, according to the ERC-721 extension standard, a metadata package containing key attributes is constructed, and then the minting function in the smart contract is called to submit the metadata package to the blockchain network; finally, the smart contract initializes the ownership of the NFT to the creator's wallet address.

Images