Jewelry rendering display method and system for treasure shopping mall

Abstract

The invention discloses a jewelry rendering display method and system for a treasure shopping mall, and relates to the field of intelligent rendering, and the method comprises the steps: obtaining a jewelry rendering style natural language description inputted by a user, and extracting a to-be-displayed jewelry original image from a jewelry image library; carrying out semantic feature extraction of the original image of the jewelry to be displayed and semantic coding of natural language description of the jewelry rendering style by adopting an image analysis and natural language processing technology based on machine vision; thereby, a jewelry rendered image is automatically generated according to a hint gate guidance joint feature between the jewelry image semantic coding feature and the jewelry rendered style semantic coding representation. Thus, the style preference of the user can be accurately understood, the jewelry rendering image meeting the expectation of the user is automatically generated, and highly personalized display experience is provided.

Description

Technical Field

[0001] The present application relates to the field of intelligent rendering, and more specifically, to a jewelry rendering and display method and system for use in a Baowuzhi mall. Background Art

[0002] With the rapid development of e-commerce, online shopping has become one of the important ways for consumers to purchase goods. For products like jewelry that are highly personalized and aesthetically pleasing, how to accurately convey their unique charm in an online environment has become a major challenge for merchants. Traditional static pictures or video displays often fail to fully display the beauty of jewelry details, which limits customers' comprehensive understanding of the product and thus affects their purchasing decisions. Therefore, jewelry rendering technology has emerged and has gradually become one of the key means to enhance the online jewelry sales experience. Through advanced 3D modeling and rendering technology, a more realistic visual presentation can be achieved, allowing customers to feel as if they were in the store even when they are in front of the screen.

[0003] However, traditional rendering methods rely on preset templates or fixed designs. Although this method simplifies the design process to a certain extent, it also significantly limits its flexibility and ability to express individuality. When users make specific style requirements, such as wanting jewelry to combine retro and modern elements, or wanting a design that incorporates an ocean theme, such complex ideas are often difficult to implement directly through existing templates. At this time, designers need to manually adjust the design based on personal experience and creativity. This process is time-consuming and prone to misunderstandings, resulting in the final result may deviate from user expectations. In addition, the ambiguity and diversity of language descriptions make it difficult for designers to accurately capture user intentions. Different designers may give very different solutions, increasing communication costs and the risk of multiple modifications.

[0004] Therefore, an optimized jewelry rendering and display solution for the Baowuzhi Mall is desired. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed.

[0006] According to one aspect of the present application, a jewelry rendering display method for a treasure record mall is provided, which includes: obtaining a natural language description of a jewelry rendering style input by a user; extracting an original image of jewelry to be displayed from a jewelry image library; performing semantic extraction on the original image of jewelry to be displayed to obtain a semantic coding feature of the jewelry image; performing semantic coding on the natural language description of the jewelry rendering style to obtain a semantic coding representation of the jewelry rendering style; performing a cross-modal union guided by a jewelry image-style hint gate on the semantic coding feature of the jewelry image and the semantic coding representation of the jewelry rendering style to obtain a rendering style-guided jewelry image semantic coding modulation feature, including: calculating the autocorrelation feature of the jewelry rendering style semantic coding representation to obtain a jewelry rendering style semantic autocorrelation coding feature; performing feature decoupling on the jewelry image semantic coding feature to obtain a set of jewelry image semantic local features; performing hint information joint modulation on the jewelry rendering style semantic autocorrelation coding feature and the set of jewelry image semantic local features to obtain the rendering style-guided jewelry image semantic coding modulation feature; obtaining a jewelry rendering image based on the rendering style-guided jewelry image semantic coding modulation feature, and displaying the jewelry rendering image.

[0007] According to another aspect of the present application, a jewelry rendering display system for a treasure record mall is provided, which includes: a rendering style language description acquisition module, which is used to acquire a natural language description of the jewelry rendering style input by a user; a jewelry original image acquisition module, which is used to extract the original image of the jewelry to be displayed from the jewelry image library; an original image semantic extraction module, which is used to perform semantic extraction on the original image of the jewelry to be displayed to obtain a jewelry image semantic coding feature; a style description semantic coding module, which is used to perform semantic coding on the natural language description of the jewelry rendering style to obtain a jewelry rendering style semantic coding representation; a rendering style jewelry image combination module, which is used to perform a jewelry image-style hint gate guided cross-modal combination on the jewelry image semantic coding feature and the jewelry rendering style semantic coding representation to obtain a rendering style guided jewelry image Like semantic coding modulation features, the rendering style jewelry image joint module includes: a rendering style autocorrelation calculation unit, which is used to calculate the autocorrelation features of the jewelry rendering style semantic coding representation to obtain the jewelry rendering style semantic autocorrelation coding features; a rendering jewelry image feature decoupling unit, which is used to perform feature decoupling on the jewelry image semantic coding features to obtain a set of jewelry image semantic local features; a rendering style guided modulation unit, which is used to perform prompt information joint modulation on the jewelry rendering style semantic autocorrelation coding features and the set of jewelry image semantic local features to obtain the rendering style guided jewelry image semantic coding modulation features; a jewelry rendering module, which is used to obtain a jewelry rendering image based on the rendering style guided jewelry image semantic coding modulation features, and display the jewelry rendering image.

[0008] Compared with the prior art, the present application provides a jewelry rendering display method and system for the Baowuzhi Mall, which obtains the natural language description of the jewelry rendering style input by the user, extracts the original image of the jewelry to be displayed from the jewelry image library, and uses image analysis and natural language processing technology based on machine vision to extract the semantic features of the original image of the jewelry to be displayed and the semantic encoding of the natural language description of the jewelry rendering style, so as to automatically generate the jewelry rendering image according to the prompt gate-guided joint features between the semantic encoding features of the jewelry image and the semantic encoding representation of the jewelry rendering style. In this way, the user's style preference can be accurately understood according to the specific style description provided by the user, and the design deviation caused by subjective understanding differences can be reduced, so as to automatically generate the jewelry rendering image that meets the user's expectations and provide a highly personalized display experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 The present invention is a flowchart of a jewelry rendering and display method for the Baowuzhi Mall according to an embodiment of the present application.

[0011] Figure 2 The present invention is a flowchart of performing a cross-modal union guided by a jewelry image-style hint gate on the jewelry image semantic coding features and the jewelry rendering style semantic coding representation in a jewelry rendering display method for a treasure store according to an embodiment of the present application to obtain a rendering style guided jewelry image semantic coding modulation feature.

[0012] Figure 3 The present invention is a flowchart of jointly modulating the jewelry rendering style semantic autocorrelation coding features and the jewelry image semantic local features in the jewelry rendering display method for the Baowuzhi Mall according to an embodiment of the present application to obtain the rendering style-guided jewelry image semantic coding modulation features.

[0013] Figure 4 The block diagram is a jewelry rendering and display system for the Baowuzhi Mall according to an embodiment of the present application. DETAILED DESCRIPTION

[0014] The embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, but rather these embodiments are provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0015] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0016] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0017] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0018] It is worth noting that in this application, all actions to obtain signals, information or data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0019] With the rapid development of e-commerce, online shopping has become one of the important ways for consumers to purchase goods. For products like jewelry that are highly personalized and aesthetically pleasing, how to accurately convey their unique charm in an online environment has become a major challenge for merchants. Traditional static pictures or video displays often fail to fully display the beauty of jewelry details, which limits customers' comprehensive understanding of the product and thus affects their purchasing decisions. Therefore, jewelry rendering technology has emerged and has gradually become one of the key means to enhance the online jewelry sales experience. Through advanced 3D modeling and rendering technology, a more realistic visual presentation can be achieved, allowing customers to feel as if they were in the store even when they are in front of the screen.

[0020] However, traditional rendering methods rely on preset templates or fixed designs. Although this method simplifies the design process to a certain extent, it also significantly limits its flexibility and ability to express individuality. When users make specific style requirements, such as wanting jewelry to combine retro and modern elements, or wanting a design that incorporates an ocean theme, such complex ideas are often difficult to implement directly through existing templates. At this time, designers need to manually adjust the design based on personal experience and creativity. This process is time-consuming and prone to misunderstandings, resulting in the final result may deviate from user expectations. In addition, the ambiguity and diversity of language descriptions make it difficult for designers to accurately capture user intentions. Different designers may give very different solutions, increasing communication costs and the risk of multiple modifications.

[0021] Therefore, in response to the above technical problems, the technical concept of the present application is to obtain the natural language description of the jewelry rendering style input by the user, extract the original image of the jewelry to be displayed from the jewelry image library, and use image analysis and natural language processing technology based on machine vision to extract the semantic features of the original image of the jewelry to be displayed and the semantic encoding of the natural language description of the jewelry rendering style, so as to automatically generate the jewelry rendering image according to the prompt gate-guided joint features between the semantic encoding features of the jewelry image and the semantic encoding representation of the jewelry rendering style. In this way, the user's style preference can be accurately understood according to the specific style description provided by the user, and the design deviation caused by subjective understanding differences can be reduced, so as to automatically generate the jewelry rendering image that meets the user's expectations and provide a highly personalized display experience.

[0022] Figure 1 FIG. 1 is a flowchart of a jewelry rendering and display method for a treasure store according to an embodiment of the present application. Figure 1 As shown, according to the embodiment of the present application, the jewelry rendering display method for the Baowuzhi mall includes: S110, obtaining a natural language description of the jewelry rendering style input by a user; S120, extracting the original image of the jewelry to be displayed from the jewelry image library; S130, performing semantic extraction on the original image of the jewelry to be displayed to obtain the semantic coding features of the jewelry image; S140, performing semantic coding on the natural language description of the jewelry rendering style to obtain the semantic coding representation of the jewelry rendering style; S150, performing a cross-modal union guided by a jewelry image-style hint gate on the semantic coding features of the jewelry image and the semantic coding representation of the jewelry rendering style to obtain the rendering style guided jewelry image semantic coding modulation features; S160, obtaining a jewelry rendering image based on the rendering style guided jewelry image semantic coding modulation features, and displaying the jewelry rendering image.

[0023] In step S110, the natural language description of the jewelry rendering style input by the user is obtained. It should be understood that the natural language description of the jewelry rendering style refers to the user using natural language (i.e., the language of daily communication) to describe the jewelry rendering style they want to see. For example, the user may say: "I want a vintage-style necklace rendering with complex patterns and dark tones." Or, "Please design a simple and modern earring, mainly silver, with smooth lines." That is, each user may have very different preferences for jewelry, some like vintage style, and some prefer modern simplicity. Therefore, by obtaining the user's natural language description, the merchant can more accurately understand the user's personal preferences, thereby providing a more personalized product display and improving user satisfaction and shopping experience.

[0024] In order to provide an intuitive and efficient user experience, the system needs to design a user-friendly interface or platform that is highly interactive and flexible so that users can describe their ideal jewelry style in detail in natural language. Such an interface must not only be concise and easy to navigate, but must also be able to accept and process various types of descriptive information, including but not limited to specific attributes such as color, material, shape, as well as more abstract design concepts and style preferences.

[0025] During this process, users can be very specific about their needs. For example, a user might say, "I hope this necklace can combine vintage and modern elements, with some marine-themed designs." Here, the user not only points out the combination of two styles (vintage and modern) that they want to see, but also specifically mentions a specific theme (ocean). This type of description is invaluable to the system because it provides clear design direction and helps the algorithm better understand the user's intentions. In addition, the user may also say, "Please design a simple yet elegant bracelet, preferably silver with smooth lines." In this sentence, the user makes it clear that the bracelet should be simple and elegant, while emphasizing the material (silver) and the details of the design (smooth lines), which are all indispensable information points when generating jewelry renderings that meet the user's expectations. Through the above measures, when users can freely and accurately express their unique insights into jewelry design, the system can better understand these needs.

[0026] In step S120, the original image of the jewelry to be displayed is extracted from the jewelry image library. Accordingly, it is considered that the original image of the jewelry to be displayed refers to the jewelry picture stored in the jewelry image library without any post-processing or modification. These pictures are usually taken by professional photographers or generated by high-precision 3D modeling technology, aiming to truly reflect the design, material, color and other characteristics of the jewelry. In other words, the original image usually has a high resolution and clarity, and can truly show the details of the jewelry. Using high-quality original images as a basis can ensure the quality of the pictures finally displayed to the user, and avoid affecting the user's judgment and purchase decision due to blurred or distorted pictures.

[0027] As the cornerstone of the entire jewelry rendering and display process, the importance of the jewelry image library is self-evident. This library not only contains a large number of high-quality pictures carefully taken by professional photographers, but also incorporates virtual images generated by high-precision 3D modeling technology. These pictures not only have extremely high resolution and clarity visually, but also can truly reproduce the design details, material texture and color characteristics of jewelry, providing users with a visual experience as if the real object is right in front of them.

[0028] In order to ensure the best visual effect, each picture that enters the jewelry image library has undergone a rigorous screening and post-processing process. This includes but is not limited to color correction, light and shadow adjustment, and detail optimization to ensure that each work can meet commercial-grade quality standards. In addition, the professional image editing team will carefully classify and organize each picture, and establish indexes according to different styles, materials, colors and other dimensions to facilitate subsequent rapid retrieval and matching.

[0029] As fashion trends continue to evolve and technology advances, it is particularly important to keep the content of the jewelry image library fresh and diverse. To this end, the relevant team will regularly update the image library to introduce the latest design trends and popular elements. This not only means adding more diverse jewelry styles, but also involves reviewing and upgrading existing image collections to eliminate outdated or non-conforming works. At the same time, we keep up with the forefront of technology and use more advanced 3D modeling software and rendering engines to create more realistic virtual models to further enhance the user experience.

[0030] It is worth noting that in the process of building such a rich and diverse jewelry image library, data management also plays a vital role. An efficient database management system can support large-scale data storage and provide powerful search functions, so that even in the face of massive information, the required information can be found quickly. In addition, considering the issue of copyright protection, all images in the library must be legally authorized and follow the corresponding regulations when used to ensure that the platform operates in compliance with the law. Therefore, a high-quality jewelry image library is not only one of the basic conditions for realizing personalized jewelry rendering, but also a key component of the bridge between merchants and consumers.

[0031] When the user submits a specific description of the jewelry rendering style, the system will start the process of selecting the most appropriate original image from the jewelry image library based on this information. Specifically, after receiving the user's customization request, the system needs to convert the natural language description provided by the user into specific query parameters. For example, when a user wants to see a design that combines retro and modern elements, the system not only needs to identify keywords such as "retro" and "modern", but also needs to further understand the style characteristics behind them. For example, retro may mean the use of classic patterns or craftsmanship, while modern may refer to simple lines or innovative materials. Through semantic analysis technology, the system can map these abstract concepts to corresponding tags or attributes in the database, so as to accurately retrieve jewelry images that contain both traditional aesthetics and fashion sense, ensuring that the search results meet the user's personalized needs.

[0032] In step S130, semantic extraction is performed on the original image of the jewelry to be displayed to obtain semantic coding features of the jewelry image. Specifically, in the embodiment of the present application, semantic extraction is performed on the original image of the jewelry to be displayed to obtain semantic coding features of the jewelry image, including: inputting the original image of the jewelry to be displayed into a dynamic convolutional labeling model to obtain a jewelry image semantic coding feature map as the jewelry image semantic coding features.

[0033] More specifically, in an embodiment of the present application, the original image of the jewelry to be displayed is input into a dynamic convolutional labeling model to obtain a semantic coding feature map of the jewelry image, including: segmenting the original image of the jewelry to be displayed into local image blocks to obtain multiple local image blocks of the jewelry to be displayed; inputting each original local image block of the jewelry to be displayed in the multiple local image blocks of the jewelry to be displayed into a convolutional neural network model to obtain multiple local feature maps of the jewelry to be displayed; after splicing the multiple local feature maps of the jewelry to be displayed to obtain a spliced ​​feature map of the jewelry to be displayed, the spliced ​​feature map of the jewelry to be displayed is input into a Transformer model to obtain the semantic coding feature map of the jewelry image. It should be understood that considering that the original image of the jewelry to be displayed contains low-level features such as color and texture, it also includes higher-level information such as shape, structure, style, etc. Therefore, in order to better understand and describe the uniqueness of jewelry, such as design style, material texture, etc., in the technical solution of the present application, the original image of the jewelry to be displayed is input into a dynamic convolutional labeling model to obtain a semantic coding feature map of the jewelry image. It should be understood that dynamic convolution can capture features at different scales, from local details to overall structure, so as to more comprehensively describe the characteristics of jewelry. Specifically, the dynamic convolution labeling model can capture relatively fine local features in the image, such as texture, edge and other information, by dividing the input image into multiple local blocks and using convolutional neural networks to process each local block. This is where CNN excels. After that, the feature maps of these local blocks are spliced ​​together and further processed globally through the Transformer model. In other words, the Transformer has a powerful self-attention mechanism, which can well capture the long-distance dependencies between the parts of the local image of the jewelry to be displayed, so as to learn richer global semantic information. Compared with the traditional static model, the dynamic convolution labeling model emphasizes the word "dynamic", which means that it can use the self-attention mechanism to automatically adjust its internal operations according to the characteristics of the input data, such as changing the attention weight distribution or adaptively selecting which local blocks need to be analyzed more carefully. This flexibility helps to improve the adaptability and efficiency of the model for scenarios of different complexity.

[0034] In step S140, the natural language description of the jewelry rendering style is semantically encoded to obtain a semantically encoded representation of the jewelry rendering style. Specifically, in an embodiment of the present application, the natural language description of the jewelry rendering style is semantically encoded to obtain a semantically encoded representation of the jewelry rendering style, including: inputting the natural language description of the jewelry rendering style into a semantic encoder including a Bert model to obtain a semantically encoded representation vector of the jewelry rendering style as the semantically encoded representation of the jewelry rendering style. Accordingly, considering that the natural language description of the jewelry rendering style contains some requirements such as the design style, material, color, etc. expected by the user, therefore, in order to better understand the meaning of each word in the sentence in a specific context, so as to better capture the user's specific requirements and preferences for the jewelry rendering style, in the technical solution of the present application, the natural language description of the jewelry rendering style is input into a semantic encoder including a Bert model to learn the rich contextual information of the rendering style from the natural language description of the jewelry rendering style through the bidirectional Transformer architecture in the Bert model, and obtain the semantically encoded representation vector of the jewelry rendering style.

[0035] The following is the specific processing of the semantic encoder containing the Bert model: First, the core of the semantic encoder is the Bert model, which is a pre-trained model based on the Transformer architecture that can learn rich contextual information from a large amount of unlabeled text data. Specifically, the Bert model uses a bidirectional Transformer to process the input sequence, which means that it not only considers the context on the left side of each word (i.e., forward), but also the context on the right side (i.e., backward). This bidirectionality enables the Bert model to more comprehensively understand each word in the sentence and its meaning in a specific context.

[0036] In this specific network structure, when users submit natural language descriptions of jewelry rendering styles, the texts are first preprocessed. The preprocessing steps include word segmentation, adding special tags (such as [CLS] and [SEP]), and converting words into corresponding vocabulary indices. Among them, the [CLS] tag is usually located at the beginning of each sentence, and its final hidden state can be used as a representation of the entire sentence; while [SEP] is used to distinguish the boundaries between different sentences or paragraphs. The preprocessed text is fed into the first layer of the Bert model - the embedding layer. The embedding layer maps each vocabulary index to a vector space of fixed dimension to form a word embedding vector. In addition, the embedding layer also adds position encoding to retain the relative position information of words in the sequence.

[0037] Next, the word embedding vector passes through a series of Transformer encoder layers. Each encoder layer consists of a multi-head self-attention mechanism module and a fully connected feedforward neural network. The multi-head self-attention mechanism allows the model to focus on multiple different information sources at the same time, thereby capturing the complex dependencies within the sentence. The fully connected feedforward neural network further enhances the learning ability of the model and improves the richness of feature expression through nonlinear transformations. After passing through multiple such encoder layers, the original word embedding vector will be gradually updated, ultimately forming a deeper understanding of the input text.

[0038] In particular, in the embodiment of the present application, in order to obtain the semantic coding representation vector of jewelry rendering style, the focus is on the hidden state corresponding to the [CLS] tag in the output of the Bert model. This vector integrates the information of the entire sentence and is very suitable for representing the semantics of the entire sentence. Therefore, this vector is used as the semantic coding representation vector of jewelry rendering style for subsequent joint processing with the semantic coding feature map of jewelry image to guide the final jewelry rendering image generation. It is worth noting that although the above description provides a basic framework of a semantic encoder based on the Bert model, in practical applications, it may be necessary to fine-tune the model according to specific circumstances. For example, the Bert model can be better adapted to the terms and expressions of a specific field by additional training on large-scale jewelry-related text data. In addition, other technical means can be introduced, such as using a domain adaptation strategy or combining other types of pre-trained models to further improve the quality and accuracy of semantic coding. Through such steps, the natural language description provided by the user can be deeply parsed, and the key information can be effectively extracted, providing a solid foundation for personalized jewelry rendering. In this way, the system can not only accurately understand the user's style preferences, but also generate high-quality and expected jewelry rendering images on this basis, greatly improving the user experience.

[0039] In step S150, the jewelry image semantic coding features and the jewelry rendering style semantic coding representation are cross-modally combined under the guidance of the jewelry image-style hint gate to obtain the rendering style guided jewelry image semantic coding modulation features. Furthermore, considering that the jewelry image semantic coding features capture the key spatial relationships and local patterns in the image, and contain important visual information in the image. The jewelry rendering style semantic coding representation vector is intended to capture the semantic information of a specific style or design element, and is obtained based on the text description. That is, the two are features belonging to different modalities. Therefore, in order to effectively fuse information from different modalities (i.e., image features and text descriptions), the style information in the text description is used to modulate the image features, so that the generated image is more in line with the user's expectations. In the technical solution of the present application, the jewelry image semantic coding features and the jewelry rendering style semantic coding representation are cross-modally combined under the guidance of the jewelry image-style hint gate to obtain the rendering style guided jewelry image semantic coding modulation features. In this way, through the cross-modal union guided by the hint gate, the generated jewelry rendering image can not only be more visually realistic, but also more accurately reflect the style described by the user, which helps to improve the quality of the final rendered image, making it more attractive and in line with the specific needs of the user.

[0040] Figure 2 The present invention is a flow chart of performing a cross-modal combination guided by a jewelry image-style hint gate on the jewelry image semantic coding feature and the jewelry rendering style semantic coding representation in a jewelry rendering display method for a treasure store according to an embodiment of the present application to obtain a rendering style guided jewelry image semantic coding modulation feature. Specifically, in the embodiment of the present application, Figure 2 As shown, the jewelry image semantic coding feature and the jewelry rendering style semantic coding representation are cross-modally combined under the guidance of a jewelry image-style hint gate to obtain a rendering style-guided jewelry image semantic coding modulation feature, including: S210, calculating the autocorrelation feature of the jewelry rendering style semantic coding representation to obtain a jewelry rendering style semantic autocorrelation coding feature; S220, performing feature decoupling on the jewelry image semantic coding feature to obtain a set of jewelry image semantic local features; S230, performing hint information joint modulation on the jewelry rendering style semantic autocorrelation coding feature and the set of jewelry image semantic local features to obtain the rendering style-guided jewelry image semantic coding modulation feature.

[0041] Specifically, in step S210, the autocorrelation features of the jewelry rendering style semantic coding representation are calculated to obtain the jewelry rendering style semantic autocorrelation coding features, including: calculating the product between the jewelry rendering style semantic coding representation vector and the transposed vector of the jewelry rendering style semantic coding representation vector to obtain the jewelry rendering style semantic autocorrelation coding matrix as the jewelry rendering style semantic autocorrelation coding features. That is, by calculating the autocorrelation matrix, it is possible to analyze how different dimensions (i.e., different style features) are related to each other, which helps to understand which style features are closely related and how this relationship affects the overall design, thereby revealing the original feature information and the inherent patterns and structures in the vector.

[0042] The processing process of step S210 can be expressed by the following formula: ;in, A semantic encoding representation vector for the jewelry rendering style, is matrix multiplication, for The transposed vector of is the jewelry rendering style semantic autocorrelation encoding matrix.

[0043] Specifically, in step S220, feature decoupling is performed on the semantic coding features of the jewelry image to obtain a set of semantic local features of the jewelry image, including: feature decoupling is performed on the semantic coding feature map of the jewelry image along the channel dimension to obtain a set of semantic local feature matrices of the jewelry image as the set of semantic local features of the jewelry image. By decomposing the feature map into multiple local feature matrices along the channel dimension, different jewelry visual features can be separated more clearly. For example, one feature matrix may mainly represent color information, another may represent texture information, and a third may represent shape or edge information, which can enable the model to more flexibly adjust the importance of each local feature, laying the foundation for the subsequent realization of more accurate cross-modal fusion.

[0044] The processing process of step S220 can be expressed by the following formula: ;in, is the semantic coding feature map of the jewelry image, For Perform feature decoupling operations. and They are the first and the second in the set of semantic local feature matrices of jewelry images. and Semantic local feature matrix of jewelry images.

[0045] Figure 3The present invention is a flowchart of performing prompt information joint modulation on the jewelry rendering style semantic autocorrelation coding feature and the jewelry image semantic local feature set in the jewelry rendering display method for the treasure store according to the embodiment of the present application to obtain the rendering style guided jewelry image semantic coding modulation feature. Specifically, in the embodiment of the present application, Figure 3 As shown, step S230, the set of the jewelry rendering style semantic autocorrelation coding features and the jewelry image semantic local features are jointly modulated by prompt information to obtain the rendering style guided jewelry image semantic coding modulation features, including: S310, linearly transforming the jewelry rendering style semantic autocorrelation coding matrix to obtain a jewelry rendering style semantic query coding matrix and a jewelry rendering style semantic value coding matrix; S320, inputting each jewelry image semantic local feature matrix in the set of the jewelry rendering style semantic query coding matrix, the jewelry rendering style semantic value coding matrix and the jewelry image semantic local feature matrix into a cross-modal prompt information encoder based on a converter structure to obtain a set of jewelry rendering style-image semantic cross-modal prompt information semantic coding matrices. ; S330, input the set of jewelry rendering style-image semantic cross-modal prompt information semantic coding matrices into the decoder-based information gating unit to obtain a set of jewelry rendering style-image cross-modal semantic interaction attention weights; S340, input the set of jewelry rendering style semantic autocorrelation coding matrices and the set of jewelry image semantic local feature matrices into the cross-modal interaction unit to obtain a set of jewelry rendering style-image cross-modal interaction local feature matrices; S350, input the set of jewelry rendering style-image cross-modal semantic interaction attention weights and the set of jewelry rendering style-image cross-modal interaction local feature matrices into the cross-modal interaction optimization unit to obtain the rendering style guided jewelry image semantic coding modulation feature map as the rendering style guided jewelry image semantic coding modulation feature.

[0046] Then, the jewelry rendering style semantic autocorrelation encoding matrix is ​​linearly transformed to obtain the jewelry rendering style semantic query encoding matrix and the jewelry rendering style semantic value encoding matrix. In this way, the query encoding matrix and the value encoding matrix are obtained by linear transformation, and the jewelry rendering style semantic autocorrelation encoding matrix is ​​converted into a form suitable for attention mechanism calculation. In particular, the query encoding matrix is ​​used to represent the style features to be found, while the value encoding matrix stores specific style information. This transformation is the core of cross-modal interaction, allowing unified comparison and integration of different modal information. It is worth mentioning that linear transformation not only performs basic mathematical operations, but also optimizes the spatial distribution of features through trainable parameters to adapt to cross-modal matching and fusion.

[0047] Next, each jewelry image semantic local feature matrix in the set of the jewelry rendering style semantic query encoding matrix, the jewelry rendering style semantic value encoding matrix, and the jewelry image semantic local feature matrix is ​​respectively input into the cross-modal prompt information encoder based on the converter structure to obtain a set of jewelry rendering style-image semantic cross-modal prompt information semantic encoding matrices. It should be understood that the converter structure uses the self-attention mechanism to handle long-distance dependencies and capture the complex interaction of cross-modal features, that is, the query encoding matrix can be matched with the style keywords in the text description, and the value encoding matrix provides specific image style information, so that the generated matrix not only retains the original features, but also integrates the knowledge of other modalities, so as to better understand the relationship between complex style features and image features.

[0048] Afterwards, the set of jewelry rendering style-image semantic cross-modal hint information semantic encoding matrices is input into the decoder-based information gating unit to obtain a set of jewelry rendering style-image cross-modal semantic interaction attention weights. It can be understood that the information gating unit can dynamically adjust the importance of information from different sources according to the context, so that the model can better integrate data from multiple modalities. This can make the final generated jewelry rendering image more natural and coherent while maintaining the consistency of style. The obtained attention weights reflect the importance of different areas or elements to the final output. Using these weights, more attention can be given to certain specific parts during the generation process, such as emphasizing a certain design feature or material texture, thereby producing more detailed and expected results.

[0049] Subsequently, the set of the jewelry rendering style semantic autocorrelation encoding matrix and the jewelry image semantic local feature matrix is ​​input into the cross-modal interaction unit to obtain the set of jewelry rendering style-image cross-modal interaction local feature matrices. At this stage, the model directly interacts with the features of different modalities between the style and the image by positional dot multiplication operation, captures the complementary information at the same position, and enhances the spatial consistency of the features. This not only retains the uniqueness of each local feature, but also strengthens the expression of the common pattern, ensuring that the generated jewelry rendering is both detailed and harmonious as a whole.

[0050] Finally, the set of jewelry rendering style-image cross-modal semantic interaction attention weights and the set of jewelry rendering style-image cross-modal interaction local feature matrices are input into the cross-modal interaction optimization unit to obtain the rendering style guided jewelry image semantic coding modulation feature map. In this way, the unit generates a comprehensive representation with strong expressiveness and less redundancy by weighted fusion of multimodal features through attention.

[0051] More specifically, the processing of step S230 can be expressed by the following formula: ;in, is the query embedding matrix, is the jewelry rendering style semantic query encoding matrix, is the value embedding matrix, is the jewelry rendering style semantic value encoding matrix, yes The transposed matrix of yes The scale is the matrix width multiplied by the matrix height, yes function, yes and Jewelry rendering style-image semantic cross-modal hint information semantic encoding matrix, is the weight matrix, It is a decoder. yes Corresponding jewelry rendering style-image cross-modal semantic interaction attention weights, It is the point multiplication by position. is the characteristic coupling along the channel dimension, It is the rendering style guided jewelry image semantic coding modulation feature map.

[0052] In step S160, based on the rendering style guiding the semantic coding modulation features of the jewelry image, a jewelry rendering image is obtained, and the jewelry rendering image is displayed. Specifically, in the embodiment of the present application, based on the rendering style guiding the semantic coding modulation features of the jewelry image, a jewelry rendering image is obtained, and the jewelry rendering image is displayed, including: inputting the rendering style guiding the semantic coding modulation feature map of the jewelry image into a rendering module based on a diffusion model to obtain the jewelry rendering image; and displaying the jewelry rendering image on the screen. That is, the rendering style guiding the semantic coding modulation features of the jewelry image obtained by the prompt gate guidance of the jewelry rendering style semantic coding representation vector is processed to automatically generate the jewelry rendering image. In this way, the user's style preference can be accurately understood according to the specific style description provided by the user, and the design deviation caused by subjective understanding differences can be reduced, so as to automatically generate a jewelry rendering image that meets the user's expectations and provide a highly personalized display experience.

[0053] Specifically, the process of inputting the rendering style-guided jewelry image semantic coding modulation feature map into the rendering module based on the diffusion model to obtain the jewelry rendering image and displaying the jewelry rendering image on the screen is as follows: It should be understood that the rendering style-guided jewelry image semantic coding modulation feature map already contains the core information about jewelry design, such as color, texture, shape, etc., and also incorporates the style elements expected by the user. The next task is to use this information to generate a specific image.

[0054] To achieve this, a method based on diffusion model is a very effective way. Diffusion model is a generative model inspired by the diffusion process in physics. This type of model generates high-quality images by gradually adding structured information to random noise.

[0055] Specifically, the process of inputting the rendering style-guided jewelry image semantic coding modulation feature map into the diffusion model-based rendering module is as follows: First, the system creates an initial state, which is usually random noise or a simple template image. This starting point represents the starting point of the rendering process.

[0056] The model then begins to iteratively modify this initial state. In each iteration, the model adjusts the state of the current image based on the information in the semantic coding modulation feature map of the jewelry image guided by the rendering style mentioned above, so that it gradually approaches the expected design. Specifically, the model analyzes the difference between the existing image and the target style, and calculates how to adjust the pixel values ​​in the next step to make the image more in line with the user's requirements.

[0057] It is worth noting that in this process, the diffusion model not only relies on the rendering style to guide the information provided by the semantic coding modulation feature map of the jewelry image, but can also add additional conditional constraints, such as keeping the color or shape of certain specific areas unchanged. Such a mechanism helps to ensure that the final generated image is both creative and faithful to the characteristics of the original jewelry.

[0058] As the number of iterations increases, the generated image becomes closer and closer to the ideal form. In order to ensure the quality of the output, some evaluation indicators can be introduced in the training stage, such as perceptual loss or adversarial loss, which can help the model learn how to generate more realistic and beautiful images.

[0059] When the diffusion model completes all iterations, a high-quality jewelry rendering image is generated, which not only reflects the user's personalized needs but also has realistic visual effects.

[0060] Next, it is the process of displaying this image to the user. Considering that different types of screens have different resolutions, color spaces, and display characteristics. Therefore, before the final display, the image may need to be properly converted to adapt to the specific requirements of the target device. For example, if it is viewed on a mobile phone, the image size should be ensured to be suitable for viewing on a small screen; if it is displayed on a professional monitor, as many details as possible should be retained. At the same time, in order to allow users to better appreciate their customized jewelry designs, multiple viewing angles and zoom options can be provided so that users can observe every detail of the jewelry from different angles. In addition, users can also be supported to rotate the image freely to simulate the feeling of the real object in their hands and enhance the sense of immersion. Although after a series of complex processing, it is theoretically possible to generate an image that is very close to the user's expectations, there may still be slight deviations. For this reason, it is recommended to set up a feedback channel to allow users to make modification suggestions so that they can respond quickly and make corresponding adjustments. This interaction not only improves customer satisfaction, but is also one of the important ways to continuously improve the performance of the algorithm. In this way, not only the attractiveness of online jewelry sales is greatly enhanced, but also the user experience is enhanced, which promotes the advancement of the display method of jewelry products in the field of e-commerce.

[0061] In particular, in one example, since the jewelry image semantic coding feature map and the jewelry rendering style semantic coding representation vector respectively represent the image semantic features of the original jewelry image to be displayed and the coded semantic features of the natural language description of the jewelry rendering style, when performing cross-modal interaction optimization coding based on a cross-modal prompt gate, due to the difference in cross-modal gating prompting caused by the difference in semantic feature distribution of cross-modal data, it is expected to improve the feature space of the rendering style-guided jewelry image semantic coding modulation feature map obtained by cross-modal interaction optimization coding - generating the target semantic response effect.

[0062] Based on this, the applicant of the present application optimizes the rendering style-guided jewelry image semantic coding modulation feature map when inputting the rendering style-guided jewelry image semantic coding modulation feature map into a rendering module based on a diffusion model to obtain a jewelry rendering image, including: calculating the absolute value sum of all eigenvalues ​​of the rendering style-guided jewelry image semantic coding modulation feature map, and the square root of the square sum of all eigenvalues ​​of the rendering style-guided jewelry image semantic coding modulation feature map, and dividing the square root of the square sum by the absolute value sum to obtain the rendering style-guided jewelry image semantic coding modulation probability value, that is: ;in, The rendering style guides the jewelry image semantic coding modulation feature map. eigenvalues, represents the absolute value sum of all eigenvalues ​​of the semantic coding modulation feature map guided by the rendering style, represents the square root of the sum of squares of all eigenvalues ​​of the semantic coding modulation feature map guided by the rendering style, Representation rendering style guides the semantic encoding modulation probability value of jewelry images.

[0063] Each feature value in the rendering style-guided jewelry image semantic coding modulation feature map is normalized to the maximum value to obtain the rendering style-guided jewelry image semantic coding modulation probability feature map: ;in, represents the maximum value in the semantic coding modulation feature map of the jewelry image guided by the rendering style, The rendering style guides the jewelry image semantic coding modulation probability feature map. eigenvalues, The rendering style guides the jewelry image semantic coding modulation probability feature map.

[0064] Multiply each feature value of the rendering style-guided jewelry image semantic coding modulation probability feature map by the rendering style-guided jewelry image semantic coding modulation probability value to obtain a rendering style-guided jewelry image semantic coding modulation field feature map ,in, Represents dot product.

[0065] After subtracting the rendering style guided jewelry image semantic coding modulation probability feature map from the unit feature map, the rendering style guided jewelry image semantic coding modulation field feature map is divided by the rendering style guided jewelry image semantic coding modulation differential feature map, that is: ;in, represents the unit eigenvector, Indicates point reduction, The bit-by-bit reciprocal of the semantic coding modulation field feature map guided by the rendering style of the jewelry image is represented by calculating the reciprocal of each eigenvalue of the semantic coding modulation field feature map guided by the rendering style of the jewelry image. Representing rendering style guided by semantic encoding modulation differential feature map of jewelry images.

[0066] Calculate the rendering style guided jewelry image semantic coding modulation field feature map with the rendering style guided jewelry image semantic coding modulation probability value as the exponent of the power function , and the rendering style guides the jewelry image semantic encoding to modulate the differential feature map of the exponential function with the natural constant as the base The weighted summation is performed to obtain the optimized rendering style-guided jewelry image semantic coding modulation feature map, namely: ;in, and represents the weighted hyperparameter, Indicates point addition, Representation-optimized rendering style guided by semantically encoded modulation feature maps for jewelry images.

[0067] That is, the mutually exclusive generalized representation behavior units of the rendering style-guided semantic coding modulation feature map of the jewelry image are constructed through the discrete differential of the probability density domain gradient corresponding to the rendering style-guided semantic coding modulation feature map of the jewelry image, so as to enhance the static response characteristics of the short-range correlation mesoscopic information configuration of the rendering style-guided semantic coding modulation feature map to the generative probabilistic generalized representation behavior based on the different behavior unit organization space under the non-uniform topological architecture of the rendering style-guided semantic coding modulation feature map of the jewelry image, thereby ensuring the steady-state convergence of the optimization process between the generated target and the extracted feature in the feature space-generated probability mapping, so as to improve the image quality of the jewelry rendered image obtained by inputting the rendering style-guided semantic coding modulation feature map of the jewelry image into the rendering module based on the diffusion model.

[0068] In summary, the jewelry rendering display method for the Baowuzhi Mall based on the embodiment of the present application is explained, which obtains the natural language description of the jewelry rendering style input by the user, extracts the original image of the jewelry to be displayed from the jewelry image library, and uses image analysis and natural language processing technology based on machine vision to extract the semantic features of the original image of the jewelry to be displayed and the semantic encoding of the natural language description of the jewelry rendering style, so as to automatically generate the jewelry rendering image according to the prompt gate guided joint features between the semantic encoding features of the jewelry image and the semantic encoding representation of the jewelry rendering style. In this way, the user's style preference can be accurately understood according to the specific style description provided by the user, and the design deviation caused by subjective understanding differences can be reduced, so as to automatically generate a jewelry rendering image that meets the user's expectations and provide a highly personalized display experience.

[0069] Figure 4 FIG. 1 is a block diagram of a jewelry rendering and display system for a treasure store according to an embodiment of the present application. Figure 4As shown, according to an embodiment of the present application, a jewelry rendering display system 100 for a treasure record mall includes: a rendering style language description acquisition module 110, which is used to acquire a natural language description of a jewelry rendering style input by a user; a jewelry original image acquisition module 120, which is used to extract the original image of the jewelry to be displayed from the jewelry image library; an original image semantic extraction module 130, which is used to perform semantic extraction on the original image of the jewelry to be displayed to obtain a semantic coding feature of the jewelry image; a style description semantic coding module 140, which is used to perform semantic coding on the natural language description of the jewelry rendering style to obtain a semantic coding representation of the jewelry rendering style; a rendering style jewelry image combination module 150, which is used to perform a cross-modal combination guided by a jewelry image-style hint gate on the semantic coding feature of the jewelry image and the semantic coding representation of the jewelry rendering style to obtain a rendering style Guide the jewelry image semantic coding modulation features, wherein the rendering style jewelry image joint module includes: a rendering style autocorrelation calculation unit, used to calculate the autocorrelation features of the jewelry rendering style semantic coding representation to obtain the jewelry rendering style semantic autocorrelation coding features; a rendering jewelry image feature decoupling unit, used to perform feature decoupling on the jewelry image semantic coding features to obtain a set of jewelry image semantic local features; a rendering style guided modulation unit, used to perform prompt information joint modulation on the jewelry rendering style semantic autocorrelation coding features and the set of jewelry image semantic local features to obtain the rendering style guided jewelry image semantic coding modulation features; a jewelry rendering module 160, used to obtain a jewelry rendered image based on the rendering style guided jewelry image semantic coding modulation features, and display the jewelry rendered image.

[0070] Here, those skilled in the art can understand that the specific operations of each step in the jewelry rendering display system for the Treasure Book Mall have been described in the above reference. Figures 1 to 3 The description of the jewelry rendering display method for the Baowuzhi Mall has been introduced in detail, and therefore, its repeated description will be omitted.

[0071] As described above, the jewelry rendering and display system 100 for the Baowuzhi Mall according to the embodiment of the present disclosure can be implemented in various wireless terminals, such as a server with a jewelry rendering and display algorithm for the Baowuzhi Mall. In a possible implementation, the jewelry rendering and display system 100 for the Baowuzhi Mall according to the embodiment of the present disclosure can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the jewelry rendering and display system 100 for the Baowuzhi Mall can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the jewelry rendering and display system 100 for the Baowuzhi Mall can also be one of the many hardware modules of the wireless terminal.

[0072] Alternatively, in another example, the jewelry rendering and display system 100 for the Baowuzhi mall and the wireless terminal may also be separate devices, and the jewelry rendering and display system 100 for the Baowuzhi mall may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0073] The above is merely an example of the principles of the present disclosure, and various modifications may be made by those skilled in the art without departing from the scope of the present disclosure. The above embodiments are presented for the purpose of illustration rather than limitation. The present disclosure may also take many forms other than those explicitly described herein.

Claims

1. A jewelry rendering and display method for a treasure store, characterized in that: include: Obtaining a natural language description of the jewelry rendering style input by the user; extracting the original image of the jewelry to be displayed from the jewelry image library; The original image of the jewelry to be displayed is semantically extracted to obtain the semantic coding features of the jewelry image; the natural language description of the jewelry rendering style is semantically encoded to obtain the semantic coding representation of the jewelry rendering style; the jewelry image semantic coding features and the jewelry rendering style semantic coding representation are cross-modally combined under the guidance of a jewelry image-style hint gate to obtain the rendering style guided jewelry image semantic coding modulation features, including: calculating the autocorrelation features of the jewelry rendering style semantic coding representation to obtain the jewelry rendering style semantic autocorrelation coding features; feature decoupling the jewelry image semantic coding features to obtain a set of jewelry image semantic local features; cue information joint modulation of the jewelry rendering style semantic autocorrelation coding features and the set of jewelry image semantic local features to obtain the rendering style guided jewelry image semantic coding modulation features; based on the rendering style guided jewelry image semantic coding modulation features, a jewelry rendering image is obtained, and the jewelry rendering image is displayed.

2. The jewelry rendering and display method for the Baowuzhi Mall according to claim 1 is characterized in that: The method performs semantic extraction on the original jewelry image to be displayed to obtain the semantic coding features of the jewelry image, including: inputting the original jewelry image to be displayed into a dynamic convolutional labeling model to obtain a jewelry image semantic coding feature map as the jewelry image semantic coding features.

3. The jewelry rendering and display method for the Baowuzhi Mall according to claim 2 is characterized in that: Inputting the original image of the jewelry to be displayed into a dynamic convolutional labeling model to obtain a jewelry image semantic coding feature map, including: dividing the original image of the jewelry to be displayed into local image blocks to obtain multiple jewelry local image blocks to be displayed; inputting each original local image block of the jewelry to be displayed in the multiple jewelry local image blocks to be displayed into a convolutional neural network model to obtain multiple jewelry local feature maps to be displayed; after splicing the multiple jewelry local feature maps to be displayed to obtain a jewelry splicing feature map to be displayed, the jewelry splicing feature map to be displayed is input into a Transformer model to obtain the jewelry image semantic coding feature map.

4. The jewelry rendering and display method for the Baowuzhi Mall according to claim 3 is characterized in that: The jewelry rendering style natural language description is semantically encoded to obtain a jewelry rendering style semantic encoding representation, including: inputting the jewelry rendering style natural language description into a semantic encoder including a Bert model to obtain a jewelry rendering style semantic encoding representation vector as the jewelry rendering style semantic encoding representation.

5. The jewelry rendering and display method for the Treasure Records Mall according to claim 4 is characterized in that: Calculating the autocorrelation features of the jewelry rendering style semantic coding representation to obtain the jewelry rendering style semantic autocorrelation coding features, including: calculating the product between the jewelry rendering style semantic coding representation vector and the transposed vector of the jewelry rendering style semantic coding representation vector to obtain a jewelry rendering style semantic autocorrelation coding matrix as the jewelry rendering style semantic autocorrelation coding features.

6. The jewelry rendering and display method for the Treasure Records Mall according to claim 5 is characterized in that: The jewelry image semantic coding features are feature decoupled to obtain a set of jewelry image semantic local features, including: feature decoupling the jewelry image semantic coding feature map along the channel dimension to obtain a set of jewelry image semantic local feature matrices as the set of jewelry image semantic local features.

7. The jewelry rendering and display method for the Treasure Records Mall according to claim 6 is characterized in that: The jewelry rendering style semantic autocorrelation coding feature and the jewelry image semantic local feature set are jointly modulated by prompt information to obtain the rendering style guided jewelry image semantic coding modulation feature, including: linearly transforming the jewelry rendering style semantic autocorrelation coding matrix to obtain a jewelry rendering style semantic query coding matrix and a jewelry rendering style semantic value coding matrix; inputting each jewelry image semantic local feature matrix in the set of the jewelry rendering style semantic query coding matrix, the jewelry rendering style semantic value coding matrix and the jewelry image semantic local feature matrix into a cross-modal prompt information encoder based on a converter structure to obtain a set of jewelry rendering style-image semantic cross-modal prompt information semantic coding matrices; The set of jewelry rendering style-image semantic cross-modal prompt information semantic coding matrices is input into a decoder-based information gating unit to obtain a set of jewelry rendering style-image cross-modal semantic interaction attention weights; the set of jewelry rendering style semantic autocorrelation coding matrices and the set of jewelry image semantic local feature matrices are input into a cross-modal interaction unit to obtain a set of jewelry rendering style-image cross-modal interaction local feature matrices; the set of jewelry rendering style-image cross-modal semantic interaction attention weights and the set of jewelry rendering style-image cross-modal interaction local feature matrices are input into a cross-modal interaction optimization unit to obtain a rendering style-guided jewelry image semantic coding modulation feature map as the rendering style-guided jewelry image semantic coding modulation feature.

8. The jewelry rendering and display method for the Treasure Records Mall according to claim 7 is characterized in that: Based on the rendering style guiding the semantic coding modulation feature of the jewelry image, a jewelry rendering image is obtained, and the jewelry rendering image is displayed, including: inputting the rendering style guiding the semantic coding modulation feature map of the jewelry image into a rendering module based on a diffusion model to obtain the jewelry rendering image; and displaying the jewelry rendering image on a screen.

9. A jewelry rendering and display system for the Baowuzhi Mall, characterized in that: include: A rendering style language description acquisition module is used to acquire a jewelry rendering style natural language description input by a user; The original jewelry image acquisition module is used to extract the original jewelry image to be displayed from the jewelry image library; the original image semantic extraction module is used to perform semantic extraction on the original jewelry image to be displayed to obtain the semantic coding features of the jewelry image; the style description semantic coding module is used to perform semantic coding on the natural language description of the jewelry rendering style to obtain the semantic coding representation of the jewelry rendering style; the rendering style jewelry image combination module is used to perform a jewelry image-style hint gate guided cross-modal combination on the jewelry image semantic coding features and the jewelry rendering style semantic coding representation to obtain the rendering style guided jewelry image semantic coding modulation features, wherein the rendering style jewelry image combination module includes: A correlation calculation unit is used to calculate the autocorrelation features of the jewelry rendering style semantic coding representation to obtain the jewelry rendering style semantic autocorrelation coding features; a jewelry rendering image feature decoupling unit is used to perform feature decoupling on the jewelry image semantic coding features to obtain a set of jewelry image semantic local features; a rendering style guided modulation unit is used to perform prompt information joint modulation on the jewelry rendering style semantic autocorrelation coding features and the set of jewelry image semantic local features to obtain the rendering style guided jewelry image semantic coding modulation features; a jewelry rendering module is used to obtain a jewelry rendering image based on the rendering style guided jewelry image semantic coding modulation features, and display the jewelry rendering image.

Images