Content fingerprint generation method and system based on deep learning model and related method

Abstract

The invention discloses a content fingerprint generation method and system based on a deep learning model and a related method, and belongs to the technical field of digital copyright protection. According to the content fingerprint generation method based on the deep learning model provided by the invention, the size of the digital work data is adjusted to adapt to the input requirement of the deep learning model, and the deep learning model is utilized to perform local feature extraction on the input image, so that subtle differences and key information of the work content can be captured; through the application of a self-attention mechanism, a global feature map is generated, and the global feature map is mapped to a low-dimensional space to obtain a content fingerprint in a high-dimensional vector form, so that the content fingerprint has high uniqueness, even if a file is subjected to operations such as clipping, compression and format conversion, the content fingerprint can still remain unchanged or only has small change, and the content fingerprint can be stored in a high-dimensional space. Therefore, the validity of copyright verification is ensured.

Description

Technical Field

[0001] The present invention relates to the field of digital copyright protection technology, and in particular to a content fingerprint generation method, system and related methods based on a deep learning model. Background Art

[0002] Traditional copyright protection methods have long relied on a variety of technical means to ensure the originality and ownership of creative works (such as images and videos). File hashing and digital watermarking are two of the more common methods. File hashing verifies the integrity and tamper-proof status of a file by generating a unique identifier (i.e., a hash value). However, a significant drawback of this method is that once the file content is modified in any way, even the slightest change, its hash value will undergo a fundamental change, rendering the original copyright verification mechanism ineffective. This is especially true when the file undergoes common processing steps such as compression, cropping, or format conversion. Changes in the hash value make copyright verification particularly difficult.

[0003] Digital watermarking, on the other hand, embeds subtle identifying information (such as a watermark code or copyright logo) into the original work to implicitly mark and track copyright. While this method improves the confidentiality and tamper-resistance of copyright information to a certain extent, it also faces significant challenges. The watermark embedding process often requires ensuring that the watermark information is both imperceptible and difficult to remove while preserving the quality of the work. However, with the continuous advancement of image processing and digital signal processing technologies, watermarking technology has become increasingly vulnerable to specialized removal or tampering tools. This not only makes it easy for infringers to circumvent copyright detection, but also brings significant inconvenience and limitations to copyright owners in enforcing and tracing their rights.

[0004] Therefore, how to provide a method that can effectively verify the copyright of digital works has become a technical problem that needs to be overcome urgently by those skilled in the art. Summary of the Invention

[0005] The purpose of the present invention is to provide a content fingerprint generation method, system and related methods based on a deep learning model to overcome the problem of copyright verification failure caused by file modification, watermark removal or tampering in the prior art.

[0006] The present invention solves the above technical problems through the following technical solutions: A method for generating a content fingerprint based on a deep learning model, wherein the deep learning model includes a convolutional neural network and an encoder, and the method specifically includes the following steps: S1. Obtain digital work data, adjust the size of the digital work data to the fixed size required by the deep learning model input, and obtain an input image; S2. Extract local features of the input image through the convolutional neural network of the deep learning model to obtain a local feature map; S3, expand the local feature map into a long vector through the encoder of the deep learning model, and generate a global feature map through the self-attention mechanism; S4. Map the global feature map to a low-dimensional space to obtain a content fingerprint in the form of a high-dimensional vector.

[0007] A further improvement of the present invention is that: the digital work data includes image data and video data; When the digital work data is image data, S1 specifically includes the following steps: Obtain image data and directly adjust the size of the image data to obtain an input image that is consistent with the fixed size required by the deep learning model input; When the digital work data is video data, S1 specifically includes the following steps: S11, extracting frames of the video data in chronological order, and selecting a number of key frames using a key frame extraction algorithm; S12, adjusting each key frame to have the same width and height, to obtain several key frames of uniform size; S13. Splice several key frames of uniform size into a two-dimensional image with temporal information to obtain an input image of a fixed size consistent with the fixed size required by the deep learning model input.

[0008] The present invention also provides a content fingerprint generation system based on a deep learning model, the deep learning model includes a convolutional neural network and an encoder, and the content fingerprint generation system includes an input layer module, a feature extraction module, an encoding module and a fully connected layer module; The input layer module is used to obtain digital work data and adjust the size of the digital work data to the fixed size required by the deep learning model input to obtain the input image; The feature extraction module is used to extract local features of the input image through the convolutional neural network of the deep learning model to obtain a local feature map; The encoding module is used to expand the local feature map into a long vector through the encoder of the deep learning model and generate the global feature map through the self-attention mechanism; The fully connected layer module is used to map the global feature map to a low-dimensional space to obtain a content fingerprint in the form of a high-dimensional vector.

[0009] The present invention also provides a protection method based on a deep learning model, comprising the following steps: Step 1: Generate a content fingerprint A of the digital work to be protected using the above-mentioned content fingerprint generation system based on the deep learning model; store the digital work to be protected and content fingerprint A in the blockchain; Step 2: Obtain the digital work to be detected; generate the content fingerprint B of the digital work to be detected through the above-mentioned content fingerprint generation system based on the deep learning model; Step 3: Based on the content fingerprint B and the content fingerprint A stored in the blockchain, the cosine similarity measurement method is used to determine whether the digital work to be detected is infringing. If the judgment is yes, it is prompted that the digital work to be detected has infringed; if the judgment is no, it is prompted that the digital work to be detected has not infringed, thereby achieving protection for the digital work to be protected.

[0010] A further improvement of the present invention is that step one also includes: generating a user public key, obtaining the time of confirmation of the ownership of the digital work to be protected, and storing the user public key and the time of confirmation of the ownership of the digital work to be protected in the blockchain to realize copyright tracing of the digital work to be protected.

[0011] A further improvement of the present invention is that the cosine similarity measurement method is used to determine whether the digital work to be detected is infringing. Specifically: Preset cosine similarity threshold; The similarity between the digital work to be detected and the digital work to be protected is calculated using the cosine similarity calculation formula; It is determined whether the similarity is greater than or equal to the cosine similarity threshold. If it is determined to be yes, it is prompted that the digital work to be detected has infringed; if it is determined to be no, it is prompted that the digital work to be detected has not infringed.

[0012] A further improvement of the present invention is that the cosine similarity calculation formula is specifically:

[0013] in, is the cosine similarity; is the content fingerprint B; Content fingerprint A stored in the blockchain; is the modulus of content fingerprint B; The modulus of the content fingerprint A stored in the blockchain.

[0014] A further improvement of the present invention is that the value range of the cosine similarity threshold is 0.9~0.95.

[0015] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned protection method based on the deep learning model.

[0016] The present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the above-mentioned protection method based on the deep learning model.

[0017] Compared with the prior art, the present invention has the following positive effects: The content fingerprint generation method based on a deep learning model provided by the present invention adjusts the size of digital work data to adapt to the input requirements of the deep learning model, and uses the deep learning model to extract local features of the input image, so as to capture subtle differences and key information in the content of the work; through the application of the self-attention mechanism, a global feature map is generated, and the global feature map is mapped to a low-dimensional space to obtain a content fingerprint in the form of a high-dimensional vector, so that the content fingerprint has a high degree of uniqueness. Even if the file undergoes operations such as cropping, compression, and format conversion, its content fingerprint can remain unchanged or only change slightly, thereby ensuring the effectiveness of copyright verification.

[0018] Furthermore, this method is not only applicable to image data, but also to digital works of video data type, showing strong adaptability and flexibility.

[0019] The protection method proposed in this paper, based on a deep learning model, utilizes blockchain to store digital works and their content fingerprints A, ensuring data integrity and security. Even if the work is tampered with or copied, the content fingerprint on the blockchain remains traceable. When suspected infringement is detected, a content fingerprint B is generated and its cosine similarity with A is measured to accurately identify the infringing behavior. If the similarity exceeds a threshold, an infringement is flagged, effectively protecting the rights of copyright owners. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings in the specification are used to provide further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0021] Figure 1 This is a flowchart of a method for generating content fingerprints based on a deep learning model according to the present invention; Figure 2 This is a flowchart of a protection method based on a deep learning model of the present invention; DETAILED DESCRIPTION To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of 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. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0022] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0023] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, which are intended to explain the present invention rather than to limit it.

[0024] See also Figure 1 A method for generating a content fingerprint based on a deep learning model includes the following steps: the deep learning model includes a convolutional neural network and an encoder, and the method for generating a content fingerprint includes the following steps: S1. Obtain digital work data, adjust the size of the digital work data to the fixed size required by the deep learning model input, and obtain an input image; S2. Extract local features of the input image through the convolutional neural network (CNN) of the deep learning model to obtain a local feature map; S3, expand the local feature map into a long vector through the encoder of the deep learning model, and generate a global feature map through the self-attention mechanism; S4. Map the global feature map to a low-dimensional space to obtain a content fingerprint in the form of a high-dimensional vector.

[0025] The content fingerprint generation method based on a deep learning model provided by the present invention ensures the consistency and efficiency of model processing by adjusting the size of digital work data to adapt to the input requirements of the deep learning model; uses the deep learning model to extract local features of the input image, which can capture subtle differences and key information in the content of the work, providing a solid foundation for subsequent copyright verification; through the application of the self-attention mechanism, the local feature map is expanded into a long vector and a global feature map is generated, which not only enhances the correlation between features, but also makes the generated content fingerprint more comprehensive and accurate, and can focus on important areas in the image and ignore redundant information, thereby improving the robustness and uniqueness of the content fingerprint; the global feature map is mapped to a low-dimensional space to obtain a content fingerprint in the form of a high-dimensional vector, so that the content fingerprint has a high degree of uniqueness. Even if the file undergoes operations such as cropping, compression, and format conversion, its content fingerprint can remain unchanged or only change slightly, thereby ensuring the effectiveness of copyright verification and the accuracy and reliability of copyright tracing.

[0026] Specifically, the digital work data includes image data and video data; When the digital work data is image data, S1 specifically includes the following steps: Obtain image data and directly adjust the size of the image data to obtain an input image that is consistent with the fixed size required by the deep learning model input; When the digital work data is video data, S1 specifically includes the following steps: S11, extracting frames of the video data in chronological order, and selecting a number of key frames using a key frame extraction algorithm; S12, adjusting each key frame to have the same width and height, to obtain several key frames of uniform size; S13. Splice several key frames of uniform size into a two-dimensional image with temporal information to obtain an input image of a fixed size consistent with the fixed size required by the deep learning model input.

[0027] This method is applicable to image data and video data. By directly adjusting the size of the image data to make it consistent with the receiving size of the deep learning model, it not only simplifies the processing flow, but also ensures the consistency and efficiency of the model processing, avoids complex image preprocessing steps, and improves the overall processing speed. When processing video data, by extracting video frames in chronological order and selecting several key frames for subsequent processing through the key frame extraction algorithm, the amount of data processed is effectively reduced, while the main information of the video content is retained, thereby improving the processing efficiency. The key frames of uniform size are spliced into a two-dimensional image with temporal information, which not only retains the temporal characteristics of the video data, but also enables the input image to fully reflect the video content, providing a richer source of information for the deep learning model to extract global features, and improving the accuracy and robustness of the content fingerprint.

[0028] Based on the same inventive concept, the present invention also provides a content fingerprint generation system based on a deep learning model, the deep learning model includes a convolutional neural network and an encoder, and the content fingerprint generation system includes an input layer module, a feature extraction module, an encoding module and a fully connected layer module; The input layer module is used to obtain digital work data and adjust the size of the digital work data to the fixed size required by the deep learning model input to obtain the input image; The feature extraction module is used to extract local features of the input image through the convolutional neural network of the deep learning model to obtain a local feature map; The encoding module is used to expand the local feature map into a long vector through the encoder of the deep learning model and generate the global feature map through the self-attention mechanism; The fully connected layer module is used to map the global feature map to a low-dimensional space to obtain a content fingerprint in the form of a high-dimensional vector.

[0029] See also Figure 2Based on the same inventive concept, the present invention also provides a protection method based on a deep learning model, comprising the following steps: Step 1: Generate a content fingerprint A of the digital work to be protected using the above-mentioned content fingerprint generation system based on the deep learning model; store the digital work to be protected and content fingerprint A in the blockchain; Step 2: Obtain the digital work to be detected; generate the content fingerprint B of the digital work to be detected through the above-mentioned content fingerprint generation system based on the deep learning model; Step 3: Based on the content fingerprint B and the content fingerprint A stored in the blockchain, the cosine similarity measurement method is used to determine whether the digital work to be detected is infringing. If the judgment is yes, it is prompted that the digital work to be detected has infringed; if the judgment is no, it is prompted that the digital work to be detected has not infringed, thereby achieving protection for the digital work to be protected.

[0030] The deep learning model-based protection method provided by this invention stores the digital work to be protected and its content fingerprint A on the blockchain, leveraging the blockchain's immutability and decentralization to ensure data integrity and security. Even if the digital work is tampered with or copied during dissemination, its content fingerprint can still be found in the original record on the blockchain, providing a reliable basis for copyright tracing. When a suspected infringing work is detected, the deep learning model is again used to generate the content fingerprint B of the digital work to be tested. This is then compared to the content fingerprint A stored on the blockchain to accurately determine whether the digital work to be tested is highly similar to the original work, thereby accurately identifying infringement. If the similarity exceeds a preset threshold, the system will indicate the existence of infringement, effectively safeguarding the legitimate rights and interests of the copyright owner.

[0031] Specifically, step one also includes: generating a user public key, obtaining the time of confirmation of ownership of the digital work to be protected, and storing the user public key and the time of confirmation of ownership of the digital work to be protected in the blockchain to realize copyright tracing of the digital work to be protected.

[0032] Specifically, the cosine similarity measurement method is used to determine whether the digital work to be detected is infringing: Preset cosine similarity threshold; The similarity between the digital work to be detected and the digital work to be protected is calculated using the cosine similarity calculation formula; It is determined whether the similarity is greater than or equal to the cosine similarity threshold. If it is determined to be yes, it is prompted that the digital work to be detected has infringed; if it is determined to be no, it is prompted that the digital work to be detected has not infringed.

[0033] Specifically, the cosine similarity calculation formula is:

[0034] in, is the cosine similarity; is the content fingerprint B; Content fingerprint A stored in the blockchain; is the modulus of content fingerprint B; The modulus of the content fingerprint A stored in the blockchain.

[0035] Specifically, the cosine similarity threshold value ranges from 0.9 to 0.95.

[0036] Example 1 Obtain identity information to register users and generate a pair of keys for each user: a public key (pubKey) and a private key (privateKey). The pubKey is used to identify the user and is publicly stored on the blockchain. The privateKey is held by the user and used for data signing and authorization operations to ensure operational security.

[0037] Encryption processing is used to ensure the uniqueness and authenticity of user identity. The user's public key and identity information are encrypted and stored in the blockchain to form an unalterable user identity record, saving on-chain storage space. The user's basic information is stored in the database to support subsequent user identity authentication.

[0038] The digital work to be protected is stored in the IPFS distributed storage network to generate a content identifier (CID). The CID is used to identify the location of the file in the IPFS network. The CID is encrypted using the user's public key to generate an encrypted content identifier (Encrypted CID). The encrypted Encrypted CID will be recorded in the blockchain. At the same time, the original CID will not be directly exposed to prevent unauthorized access.

[0039] A deep learning model combining Transformer and convolutional neural network (CNN) is used to generate a unique content fingerprint for the digital work to be protected. This deep learning model can efficiently extract the characteristics of digital works. Even if the file has undergone common modifications such as cropping, compression, and format conversion, it can still be effectively matched with the original copyright record. Compared with the traditional copyright registration method based on file hashing, this method can better improve the accuracy and reliability of copyright protection and provide more solid technical support for the copyright management of digital works.

[0040] The content fingerprint of the digital work to be protected, the user's public key, the time the copyright was confirmed, and the corresponding content identifier are stored on the blockchain. The content fingerprint of the digital work to be protected is a unique identifier; the user's public key is used to bind the copyright to the owner's identity; the time the copyright was confirmed is used to record the specific time of copyright confirmation, providing a basis for copyright history tracing; and the content identifier is used to retrieve digital work files from distributed storage. When a copyright tracing request occurs, the digital work to be traced is selected and the digital work information on the blockchain is called for tracing.

[0041] When infringement detection occurs, content fingerprints are used. First, the infringing work to be detected is obtained. By comparing the content fingerprint of the digital work to be detected with the content fingerprint of the digital work to be protected that has been registered on the blockchain, it is possible to detect whether there is copyright infringement. Specifically: if the content fingerprint of the digital work to be detected is highly similar to the content fingerprint of the digital work to be protected that has been registered on the blockchain, it is indicated that the work may be infringing, and the infringement record is stored in the database. Among them, cosine similarity is used as a similarity measurement method. Cosine similarity measures the similarity between two vectors by calculating the angle between them. The closer the value is to 1, the more similar the two content fingerprints are. If the calculated similarity S is greater than a preset threshold (for example, 0.9), it is determined that the digital work to be detected is highly similar to the digital work to be protected stored on the blockchain, indicating that infringement may have occurred.

[0042] This method achieves efficient storage and trusted management of copyright data by combining blockchain and IPFS distributed storage technologies, solving the problems of low storage efficiency and easy data tampering in traditional technologies; it adopts copyright registration technology based on content fingerprints to effectively solve the copyright confirmation and tracing problems of traditional copyright protection methods after file modification; even if the digital protected work file is further cropped, compressed, format converted and other operations, the generated content fingerprint can still maintain a high degree of uniqueness and robustness, ensuring the accuracy and reliability of copyright tracing.

[0043] Based on the same inventive concept, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the protection method based on the deep learning model are implemented. Specifically, the computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may include RAM (Random Access Memory) and / or cache memory, etc. The non-volatile memory may include ROM (Read-Only Memory), hard disk, flash memory, optical disk, magnetic disk, etc.

[0044] Based on the same inventive concept, an embodiment of the present application provides a computer program product, which includes a computer program stored on a computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer device, the computer device executes the steps of the above-mentioned deep learning model-based protection method.

[0045] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROMs (Compact Disc Read-Only Memory), optical storage, etc.) containing computer-usable program code.

[0046] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer device or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0047] These computer program instructions may also be stored in a computer readable memory that can direct a computer device or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0048] These computer program instructions can also be loaded onto a computer device or other programmable data processing device so that a series of operating steps are executed on the computer device or other programmable device to produce a process implemented by the computer device, thereby providing instructions for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0049] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0050] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A content fingerprint generation method based on a deep learning model, characterized in that: The deep learning model includes a convolutional neural network and an encoder. The content fingerprint generation method specifically includes the following steps: S1. Obtain digital work data, adjust the size of the digital work data to the fixed size required by the deep learning model input, and obtain an input image; S2. Extract local features of the input image through the convolutional neural network of the deep learning model to obtain a local feature map; S3, expand the local feature map into a long vector through the encoder of the deep learning model, and generate a global feature map through the self-attention mechanism; S4. Map the global feature map to a low-dimensional space to obtain a content fingerprint in the form of a high-dimensional vector.

2. The method for generating content fingerprints based on a deep learning model according to claim 1, characterized in that: The digital work data includes image data and video data; When the digital work data is image data, S1 specifically includes the following steps: Obtain image data and directly adjust the size of the image data to obtain an input image that is consistent with the fixed size required by the deep learning model input; When the digital work data is video data, S1 specifically includes the following steps: S11, extracting frames of the video data in chronological order, and selecting a number of key frames using a key frame extraction algorithm; S12, adjusting each key frame to have the same width and height, to obtain several key frames of uniform size; S13. Splice several key frames of uniform size into a two-dimensional image with temporal information to obtain an input image of a fixed size consistent with the fixed size required by the deep learning model input.

3. A content fingerprint generation system based on a deep learning model, characterized in that: The deep learning model includes a convolutional neural network and an encoder, and the content fingerprint generation system includes an input layer module, a feature extraction module, an encoding module, and a fully connected layer module; The input layer module is used to obtain digital work data and adjust the size of the digital work data to the fixed size required by the deep learning model input to obtain the input image; The feature extraction module is used to extract local features of the input image through the convolutional neural network of the deep learning model to obtain a local feature map; The encoding module is used to expand the local feature map into a long vector through the encoder of the deep learning model and generate the global feature map through the self-attention mechanism; The fully connected layer module is used to map the global feature map to a low-dimensional space to obtain a content fingerprint in the form of a high-dimensional vector.

4. A protection method based on a deep learning model, characterized in that: The following steps are involved: Step 1: Generate a content fingerprint A of the digital work to be protected by using the content fingerprint generation system based on the deep learning model as described in claim 3; The digital work to be protected and the content fingerprint A are stored in the blockchain; Step 2: Obtain the digital work to be detected; generate a content fingerprint B of the digital work to be detected by the content fingerprint generation system based on the deep learning model as described in claim 3; Step 3: Based on the content fingerprint B and the content fingerprint A stored in the blockchain, the cosine similarity measurement method is used to determine whether the digital work to be detected is infringing. If the judgment is yes, it is prompted that the digital work to be detected has infringed; if the judgment is no, it is prompted that the digital work to be detected has not infringed, thereby achieving protection for the digital work to be protected.

5. A protection method based on a deep learning model according to claim 4, characterized in that: Step one also includes: generating a user public key, obtaining the time of confirmation of ownership of the digital work to be protected, and storing the user public key and the time of confirmation of ownership of the digital work to be protected in the blockchain to realize copyright tracing of the digital work to be protected.

6. A protection method based on a deep learning model according to claim 4, characterized in that: The cosine similarity measurement method is used to determine whether the digital work to be detected is infringing: Preset cosine similarity threshold; The similarity between the digital work to be detected and the digital work to be protected is calculated using the cosine similarity calculation formula; It is determined whether the similarity is greater than or equal to the cosine similarity threshold. If it is determined to be yes, it is prompted that the digital work to be detected has infringed; if it is determined to be no, it is prompted that the digital work to be detected has not infringed.

7. A protection method based on a deep learning model according to claim 6, characterized in that: The cosine similarity calculation formula is as follows: in, is the cosine similarity; is the content fingerprint B; Content fingerprint A stored in the blockchain; is the modulus of content fingerprint B; The modulus of the content fingerprint A stored in the blockchain.

8. The protection method based on deep learning model according to claim 6, characterized in that: The cosine similarity threshold value ranges from 0.9 to 0.

95.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the protection method based on the deep learning model described in claims 4 to 8 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the protection method based on the deep learning model as described in claims 4 to 8 are implemented.

Images