Method and device for determining presentation effect of live broadcast room, equipment and storage medium

Abstract

The invention provides a live broadcast room presentation effect determination method and device, equipment and a storage medium, and relates to the technical field of computers, in particular to the technical fields of information network technologies, artificial intelligence, intelligent e-commerce and the like. The specific implementation scheme is as follows: determining information of a live broadcast room; performing multi-mode identification on live broadcast commodities in a live broadcast room to determine commodity features of the live broadcast commodities; according to the commodity characteristics, recommending a decoration template for the live broadcast commodity; and determining the presentation effect of the live broadcast room according to the information of the live broadcast room and the decoration template. The attraction of the live broadcast room can be enhanced, the watching experience of the user is optimized, and meanwhile the overall effect of live broadcast is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to technical fields such as information network technology, artificial intelligence, and smart e-commerce. Background Art

[0002] With the continuous development of Internet technology, more and more businesses are selling goods through online live broadcasting. As a key place for real-time interaction between businesses and consumers, the visual and auditory presentation effect of the live broadcast room is directly related to the audience's stay time, purchase intention and brand image shaping. Therefore, how to determine the live broadcast room presentation effect that matches the live broadcast product to improve the display effect and enhance the attractiveness to users is a problem that needs to be solved. Summary of the invention

[0003] The present invention provides a method, device, equipment and storage medium for determining the presentation effect of a live broadcast room.

[0004] According to one aspect of the present disclosure, a method for determining a live broadcast room presentation effect is provided, comprising:

[0005] Determine the live broadcast room information;

[0006] Perform multimodal recognition on live broadcast products in the live broadcast room to determine the product features of the live broadcast products;

[0007] Recommend decoration templates for live broadcast products based on product features;

[0008] Determine the presentation effect of the live broadcast room based on the room's information and decoration template.

[0009] According to another aspect of the present disclosure, a device for determining a presentation effect of a live broadcast room is provided, comprising:

[0010] A first determination module is used to determine the information of the live broadcast room;

[0011] A feature recognition module is used to perform multimodal recognition on the live broadcast products in the live broadcast room to determine the product features of the live broadcast products;

[0012] Template recommendation module, used to recommend decoration templates for live broadcast products based on product features;

[0013] The second determination module is used to determine the presentation effect of the live broadcast room according to the information and decoration template of the live broadcast room.

[0014] According to another aspect of the present disclosure, there is provided an electronic device, comprising:

[0015] at least one processor; and

[0016] a memory communicatively connected to the at least one processor; wherein,

[0017] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute any method in the embodiments of the present disclosure.

[0018] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute any method according to the embodiments of the present disclosure.

[0019] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, which implements any method according to the embodiments of the present disclosure when executed by a processor.

[0020] This disclosure determines the live broadcast room information, identifies the product features of the live broadcast products, and intelligently recommends the corresponding live broadcast room decoration templates based on these features, and finally creates a presentation effect that matches the live broadcast products. The above approach not only significantly enhances the attractiveness of the live broadcast room, but also greatly optimizes the audience's viewing experience, while improving the overall effect of the live broadcast.

[0021] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0023] Figure 1 is a flow chart of a method 100 for determining a live broadcast room presentation effect according to an embodiment of the present disclosure;

[0024] Figure 2 It is a flow chart of an implementation of a method 200 for recommending a decoration template for a live broadcast room according to an embodiment of the present disclosure;

[0025] Figure 3 It is a flow chart of an implementation of a method 300 for recommending a decoration template for a live broadcast room according to an embodiment of the present disclosure;

[0026] Figure 4 is a structural diagram of a device 400 for determining a live broadcast room presentation effect according to an embodiment of the present disclosure;

[0027] Figure 5 is a structural schematic diagram of a device 500 for determining a live broadcast room presentation effect according to an embodiment of the present disclosure;

[0028] Figure 6A schematic block diagram of an example electronic device 600 that may be used to implement embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0029] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0030] The “and / or” in the embodiments of the present disclosure indicates that three relationships may exist. For example, A and / or B may indicate the following three situations: A exists alone, A and B exist at the same time, and B exists alone. The term “at least one” herein indicates any combination of at least two of any one or more of a plurality of types. For example, at least one of A, B, and C may indicate any one or more elements selected from the set consisting of A, B, and C. The terms “first” and “second” herein refer to and distinguish between multiple similar technical terms, and do not mean to limit the order or to limit the meaning to only two. For example, the first feature and the second feature refer to two types / two features. The first feature may be one or more, and the second feature may also be one or more.

[0031] With the continuous development of Internet technology, more and more businesses are selling goods through online live broadcasting. As a key place for real-time interaction between businesses and consumers, the visual presentation effect of the live broadcast room is directly related to the audience's stay time, purchase intention and brand image shaping. Therefore, how to determine and present the live broadcast room effect that matches the live broadcast product has become an important link that businesses cannot ignore.

[0032] At present, the method of determining the presentation effect of the live broadcast room still relies heavily on the fixed decoration template system preset by the live broadcast platform. Although this method provides merchants with a convenient choice, allowing them to select a decoration scheme that better suits their needs from a series of preset templates based on their initial aesthetic or brand style preferences, it has significant limitations, namely, it is unable to flexibly make dynamic adjustments or automatically optimize the presentation effect based on the specific characteristics of the live broadcast products.

[0033] In order to solve the above problems, the embodiments of the present disclosure propose a method and device for determining the presentation effect of a live broadcast room. Among them, the method and device for determining the presentation effect of a live broadcast room proposed in the embodiments of the present disclosure can be applied to electronic devices such as mobile phones, computers, intelligent voice interaction devices, smart home appliances, car terminals, game consoles, e-book readers, multimedia playback devices, wearable devices, etc. The electronic device determines the information of the live broadcast room, and identifies the product characteristics of the live broadcast products, and recommends decoration templates for the live broadcast products according to the product characteristics, and determines the presentation effect of the live broadcast room based on the decoration template and the information of the live broadcast room. Alternatively, the method and device for determining the presentation effect of a live broadcast room proposed in the embodiments of the present disclosure can also be applied to a cloud server. After receiving the product characteristics of the live broadcast products, a decoration template is recommended for the live broadcast products according to the product characteristics, and the decoration template is fed back to the terminal device. The terminal device determines the presentation effect of the live broadcast room based on the information of the live broadcast room and the decoration template.

[0034] Figure 1 1 is a flowchart of a method 100 for determining a live broadcast room presentation effect according to an embodiment of the present disclosure, comprising:

[0035] S110, determining the information of the live broadcast room;

[0036] S120, performing multimodal recognition on the live broadcast products in the live broadcast room to determine product features of the live broadcast products;

[0037] S130, recommending a decoration template for the live broadcast product according to product characteristics;

[0038] S140: Determine the presentation effect of the live broadcast room according to the information and decoration template of the live broadcast room.

[0039] The method for determining the presentation effect of a live broadcast room proposed in the disclosed embodiment can determine the live broadcast room information and extract the product features of the live broadcast products, and intelligently recommend the decoration templates of the live broadcast room based on this, and determine the final presentation effect, thereby enhancing the attractiveness of the live broadcast and the experience of the audience, optimizing the live broadcast effect, and thus increasing revenue.

[0040] In some implementations, multimodal recognition includes at least one of image recognition, text recognition, and speech recognition.

[0041] In the disclosed embodiments, multimodal recognition technology is a comprehensive information processing method that integrates multiple technologies such as image recognition, text recognition and speech recognition. Image recognition mainly analyzes the color, shape, texture and other features in the product image to understand and interpret the image content and extract the product features of the product; text recognition technology is specifically responsible for accurately extracting the text information (such as product name, brand name, etc.) contained in the product image and converting it into an editable and easy-to-search text format; speech recognition uses acoustic models and language models to convert human speech into text or instructions.

[0042] Through multimodal recognition technology, various characteristics of live broadcast products can be captured more comprehensively. This comprehensive recognition method not only takes into account the appearance of the product (such as color, shape, material, etc.), but also combines the host's verbal description and the audience's real-time feedback, thereby greatly improving the accuracy of product recognition.

[0043] Product features can be obtained by using deep learning models to identify multimodal information. In the process of fine training and continuous optimization of deep learning models, the diversity of data samples is highly valued to improve the generalization ability of the model. To achieve this goal, the embodiments of the present disclosure can collect and integrate a large number of product images and use these product images as data samples, thereby improving the accuracy of the deep learning model in identifying product features.

[0044] At the same time, the picture samples used in the embodiments of the present disclosure can be taken from the front, side, back and multiple angles. This design is intended to help the model understand the morphological changes of objects under different viewing angles, so that it can still accurately identify and classify in complex and changeable practical application scenarios. The diversity of background complexity and lighting conditions are also important factors in the construction of the data set. A variety of background pictures from simple solid color backgrounds to complex life scenes, as well as various lighting pictures from bright daylight to dim nights are introduced. In addition, the training data set is further enriched through data enhancement techniques such as random cropping, rotation, scaling and color transformation. These methods can improve the robustness of the model, enabling it to maintain stable performance output when dealing with various complex environments.

[0045] The disclosed embodiment can use a convolutional neural network (CNN) architecture to build a product feature recognition system, such as using CNN models such as ResNet and EfficientNet, to improve the performance of the product feature recognition system. Furthermore, an attention mechanism (AM) can be introduced to make the model more accurate when processing complex backgrounds and multi-character information. The disclosed embodiment uses an organic combination of the attention mechanism and the CNN architecture, so that the product recognition system can improve the accuracy of product feature recognition when facing product images with complex backgrounds, obstructions, light changes, and multi-character logos.

[0046] In some embodiments, the disclosed embodiments adopt a multi-task learning strategy to improve the intelligent recognition ability of the model. For example, the model is trained to perform multiple related but different tasks at the same time, so as to promote the sharing of internal knowledge of the model and the improvement of generalization ability. In one example, optical character recognition (OCR) is used as one of the core tasks of model training, and additional tasks such as product classification and attribute extraction are introduced. By training the OCR task of the model, the model can recognize the text information in the product image, such as key data such as brand name, model, price, etc.; by training the product classification task, the model can classify the product into the correct category according to its appearance, features, etc.; by training the attribute extraction task, the model can extract the detailed attribute information of the product, such as color, size, material, etc.; the above information has a certain impact on the recommendation and determination of the decoration template of the live broadcast room. Through multi-task learning, the overall recognition ability and robustness of the model can be improved.

[0047] Furthermore, for some complex or special scenarios, in addition to using a deep learning model to recognize images, the embodiments of the present disclosure may also use the following methods:

[0048] (1) For images that are difficult for deep learning models to accurately identify, a strategy of combining manual annotation with rule engines can be used. This strategy combines the dual guarantees of pre-set rules and manual review, which can improve the accuracy of product feature recognition.

[0049] (2) Before the recognition task is performed, a series of preprocessing and enhancements are performed on the input information, such as noise reduction and contrast optimization of live product images. This method can improve the clarity and readability of the image, thereby improving the accuracy of recognition.

[0050] (3) Build an iterative feedback loop system that automatically captures samples with recognition errors, then manually reviews and re-labels them, and reincorporates these corrected data into the training set, so that the model can continue to learn and optimize.

[0051] Using the above method, the product information identification system can efficiently and accurately extract product information, provide users with comprehensive product data content, and improve the intelligence level of e-commerce platforms and the retail industry.

[0052] In some embodiments, the commodity characteristics identified by the commodity information identification system include at least one of color characteristics, brand characteristics, material characteristics, and type.

[0053] In some embodiments, the color feature includes at least one of the main color of the product and the color corresponding to the product type; the brand feature includes at least one of the brand style and the brand color; the material feature includes at least one of the material visual effect and the material texture.

[0054] In the disclosed embodiment, product features are a multi-dimensional set that covers not only the external performance of the product, but also its internal attributes and market positioning. Product features are an important basis for recommending live broadcast room decoration templates in the process of determining the live broadcast room presentation effect. These features include at least the following aspects:

[0055] Color features: Color is the most intuitive feature of a product. Different product colors correspond to decoration templates with different themes. For example, if a product uses bright red as the main color, you can recommend a decoration template that incorporates red tones to create a warm and vibrant atmosphere. At the same time, consider the psychological effect of color and ensure that the color of the selected decoration template can stimulate the target audience's desire to buy. For example, when selling environmentally friendly products in a live broadcast, you can recommend a decoration template with green as the main color.

[0056] Brand characteristics: Brand is the core of product identity, and its unique visual elements (such as brand logo, standard color, font, etc.) should be an important reference for decoration template design. By incorporating brand characteristics ingeniously, the brand image can be strengthened and the audience's awareness and favorability of the brand can be enhanced. In addition, the display of brand stories and cultural connotations can also add depth and appeal to the live broadcast room.

[0057] Material characteristics: Material is the basic element that constitutes the product and directly determines the quality of the product. Material visual effects mainly refer to the material characteristics presented through visual perception, including color, gloss, texture, etc. These elements can make the audience feel its texture and beauty based on the presentation effect of the live broadcast room. For example, for jewelry products, a simple and exquisite decoration template can be used, with dark tones and high-quality materials as the background to create a luxurious and professional atmosphere.

[0058] Material texture is more related to tactile perception, that is, the feeling people get by touching the material, including the softness, hardness, smoothness, roughness, coldness and warmth of the material. Although the audience cannot directly feel the material texture of the product when watching the live broadcast, they can adjust the presentation effect of the live broadcast room and use visual means to convey the texture of the material. For example, for silk products, you can use a decoration template with elegant colors to set off the delicate luster of silk.

[0059] Type characteristics: The type of product determines its target audience and market demand. Therefore, when recommending decoration templates, it is necessary to consider the aesthetic preferences and consumption habits of the target audience corresponding to the product type. For example, for fashion apparel products targeting young consumers, you can use decoration templates with rich colors and avant-garde designs; while for household products targeting family users, you can focus more on creating a warm and comfortable atmosphere.

[0060] By using the above method, we can more comprehensively evaluate product features and determine a more matching decoration template to optimize the live broadcast display effect.

[0061] Figure 2 It is a flowchart of an implementation method 200 for recommending a decoration template for a live broadcast room according to an embodiment of the present disclosure, including:

[0062] S201. Use a deep learning model to perform multimodal recognition on the information of the live broadcast products, and extract product features of the live broadcast products based on the multimodal recognition results.

[0063] S202: Based on the product features, search for a pre-saved decoration template list, determine the relevance between each template in the decoration template list and the product features, and set a weight for each product feature.

[0064] S203. Calculate the recommendation score of each template in the decoration template list based on the relevance between each template and the product feature and the weight of the product feature.

[0065] S204: Based on the recommendation score, recommend a decoration template for the live broadcast product.

[0066] In the disclosed embodiments, data from different modalities are preprocessed so that they can be effectively processed by the deep learning model. For example, image data may be resized and normalized; text data may need to be segmented and stop words removed; and speech data may need to be sampled and noise suppressed.

[0067] Using deep learning models for single-modal feature extraction, CNN can be used to extract visual features in images, which may include edges, textures, shapes, etc., and CNN can extract audio features such as pitch, rhythm, timbre, etc. In addition, the Transformer model is used to process text data and extract semantic features such as keywords, sentiment tendencies, etc. After extracting the features of each modality, they need to be fused to form a comprehensive description of the characteristics of the live broadcast product.

[0068] By using deep learning algorithms, we can match products with more suitable decoration templates and realize personalized recommendations for decoration templates.

[0069] In some embodiments, based on the correlation between the decoration template and each product feature of the live broadcast product, and the weight corresponding to each correlation, each correlation is weighted and summed to obtain a recommendation score of the decoration template for the live broadcast product.

[0070] In the disclosed embodiment, attribute labels for each recommendation template for the live product characteristics can be pre-set. These labels cover the basic attributes of the product, such as color, material, type, etc., and establish a correlation relationship between the decoration template and the characteristic elements in the product characteristics, so as to facilitate the subsequent recommendation and scoring of the decoration template. In an example, the decoration template list is shown in Table 1 below:

[0071] Table 1

[0072]

[0073] In one example, the correlation range is 0 to 100, with 0 indicating the lowest correlation and 100 indicating the highest correlation. As shown in Table 1, the correlations of template 1 to the product color features of white, blue, and green are 80, 60, and 50, respectively, which means that the template may be suitable for live broadcast rooms selling white products; the correlations of template 1 to the material features of cotton, metal, and ceramic are 20, 70, and 40, respectively, which means that the template may be suitable for live broadcast rooms selling metal products; the correlations of template 1 to the brand features of sports and leisure are 80 and 30, respectively, which means that the template may be more suitable for live broadcast rooms selling sports products.

[0074] In the disclosed embodiment, it is necessary to assign initial weights to various features of the product. This process requires an in-depth understanding of the essential attributes of the product, market demand, and preferences of the target consumer groups, so that the weight of each product feature can properly reflect its influence in the process of recommending decoration templates.

[0075] In the disclosed embodiment, according to the specific product characteristics of the live broadcast product, each decoration template is searched in the decoration template list to determine the similarity between each decoration template and the product characteristics. For example, if the color characteristic of a live broadcast product is white, the material characteristic is metal, and the brand characteristic is leisure, then the correlation between template 1 in the decoration template list and the product in color characteristics is 80, the correlation in material characteristics is 70, and the correlation in brand characteristics is 30.

[0076] Based on the relevance of each template to the product features and the weight of the product features, each template is scored. The template score calculation formula is:

[0077] Score(T)=w_color*SimColor(T,Product)+w_brand*SimBrand(T,Product)+w_

[0078] material*SimMaterial(T,Product)

[0079] w_color+w_brand+w_material=1

[0080] Among them, w_color, w_brand, and w_material are the weights of color features, brand features, and material features, respectively, and the sum of all weights is 1; SimColor(T,Product), SimBrand(T,Product), and SimMaterial(T,Product) are the similarities between the template and the corresponding color feature elements, brand feature elements, and material feature elements, respectively. Continuing with the above example, the weight values ​​of w_color, w_brand, and w_material are set to 0.3, 0.4, and 0.3, respectively. Taking the example shown in Table 1 as an example, for this live broadcast product, the recommendation score of template 1 for this live broadcast product is:

[0081] 80*0.3+30*0.4+70*0.3=57.

[0082] Using this calculation method, we can obtain the recommendation scores of all decoration templates in Table 1 for the live broadcast product.

[0083] After obtaining the recommendation score calculated based on the correlation between product features and decoration templates and the corresponding weights, this result is used as the core basis to recommend decoration templates for live broadcast products. Specifically, the candidate decoration templates are sorted according to the recommendation scores, and templates with higher scores are recommended first. If there are templates being promoted among the templates with higher scores, these templates will be given priority for recommendation. These templates can not only highlight the characteristics of live broadcast products and enhance the audience's visual experience, but also enhance the audience's willingness to buy.

[0084] Using the above method can greatly save the time and cost of manual screening of decoration templates.

[0085] In some implementations, the weights corresponding to the relevance are adjusted according to business needs and information of the live broadcast room.

[0086] In the disclosed embodiment, the differences in different business needs and specific information of the live broadcast room have an important impact on the recommendation effect. Therefore, in order to further improve the accuracy and effectiveness of the recommendation, the weights corresponding to the relevance of product features can be flexibly adjusted according to business needs and specific information of the live broadcast room.

[0087] For example, in some scenarios, the brand characteristics of the live broadcast products may be more important than the color characteristics. In this case, the weight of the relevance of the brand characteristics can be increased to ensure that the recommended decoration template can accurately convey the brand's visual image. In other scenarios, if color matching is crucial to attracting the target audience or creating a specific atmosphere, the proportion of color matching in the weight can be increased to ensure that the color design of the decoration template can resonate with the audience and enhance the attractiveness and appeal of the live broadcast. By adjusting the weights, it is possible to respond more flexibly to different business needs and actual situations, and provide personalized and precise decoration template recommendations for live broadcast products.

[0088] In some implementations, a decoration template list is determined and saved; wherein the decoration template list includes a plurality of decoration templates and includes the relevance of each decoration template to various commodity features.

[0089] In this way, the decoration template list not only contains multiple decoration templates, but also clearly records the relevance of each template to various product features. This structured data storage method enables the system to quickly find matching decoration templates based on the specific features of the product.

[0090] Figure 3 It is a flowchart of an implementation method 300 for recommending a decoration template for a live broadcast room according to an embodiment of the present disclosure, including:

[0091] S301, based on the product features, generating feature vectors of each product feature of the live broadcast product. The product features can be obtained by using step S201.

[0092] This step is a data preprocessing process, which aims to convert product features into data that the model can understand.

[0093] The color features in product features can be converted into color vectors through color extraction algorithms. This algorithm can accurately identify and quantify the color information in the image and convert it into a data format that is easy for computers to process and understand. Specifically, color vectors can be represented by a variety of color space models, the most common of which are the RGB (red, green, blue) color model and the HSV (hue, saturation, brightness) color model. In the RGB model, the color vector consists of three components, representing the intensity or brightness of red, green, and blue in the image, which together define each color in the image. The HSV model provides a way to describe color that is closer to human visual perception, where hue represents the basic attributes of color, saturation represents the purity or brightness of the color, and value represents the brightness of the color.

[0094] For brand features in product features, brand information can be encoded as a numerical vector or represented using an embedded vector. Specifically, each brand name or identifier can be mapped to a unique numerical value or numerical sequence through a certain preset rule or standard. Alternatively, by learning semantic information, user behavior, etc. from a large amount of brand data, the brand can be mapped to a continuous vector space. In this vector space, similar brands will be mapped to similar positions, thereby preserving the similarity and correlation between brands.

[0095] For the material features in the product features, the word vector generation model (such as Word2Vec, GloVe) can be used to encode the material information. Specifically, the word vector model is trained on a corpus containing a large number of material names and their contextual information. Through the training process, the model is able to learn the vector representation of each material name. These vectors not only contain the semantic information of the material name, but also reflect the similarities and differences between materials to a certain extent. For example, in the Word2Vec model, similar materials will be relatively close in the vector space (such as cotton and linen, which are similar in touch, breathability, etc.), while materials with greater differences will be farther apart (such as "metal" and "cloth").

[0096] For other auxiliary information, such as product category information, the category can be encoded or represented by an embedded vector. The product description can extract semantic features through natural language processing technology (such as BERT model, GPT model), and the text can be understood through the model to automatically extract keywords and phrases in the product description. The effect data of historical recommendations, such as user feedback, click-through rate, etc., can be used to optimize the decoration template recommendation model. By analyzing user feedback, we can identify the recommendation types that are more popular with users and the recommendation types that need to be improved. By monitoring and analyzing the click-through rate data, we can find out which recommendation templates are more suitable for the general public, and based on this, we can tune the model.

[0097] S302: Input the product feature vector into the decoration template recommendation model.

[0098] S303: The decoration template recommendation model processes the feature vector and outputs a recommended decoration template.

[0099] In the disclosed embodiment, a suitable neural network architecture (such as a multi-layer perceptron, CNN or a Transformer-based model, etc.) is selected to build an efficient and accurate decoration template recommendation model. The input layer of the decoration template recommendation model is responsible for receiving the feature vector of the product information.

[0100] In the disclosed embodiment, several convolutional layers and fully connected layers are used to design the hidden layers of the decoration template recommendation model. Among them, the convolutional layer is mainly used to extract local features from the input data. These features can be the edges, textures, shapes, etc. of the image, or local patterns of text or sequence data. By stacking multiple convolutional layers, more advanced and abstract feature representations can be gradually extracted.

[0101] The fully connected layer is mainly used to integrate the local features extracted by the previous convolutional layer into global features and perform tasks such as classification or regression. At the same time, the fully connected layer can also be used as a means of feature fusion to integrate features from different sources or different levels.

[0102] The output layer is the last layer of the decoration template recommendation model, which is responsible for converting the high-level features extracted by the previous layers into task-specific outputs. The output layer of the decoration template recommendation model can be constructed by a Softmax or multi-label classification layer.

[0103] The Softmax layer is a special fully connected layer used for multi-classification problems. It can convert the input feature vector into a probability distribution, that is, the predicted probability of each category. In the decoration template recommendation model, the Softmax layer can output the probability distribution of each template being recommended.

[0104] The multi-label classification layer is an output layer designed for instances that can be classified into multiple categories at the same time. It is used to handle such complex classification tasks. Multi-label classification can be implemented through multiple Sigmoid functions, one Sigmoid function for each category. Each Sigmoid function independently outputs the probability that the category is positive. In the decoration template recommendation model, the multi-label classification layer can be used to output the score or probability of each template, indicating the user's preference for the template.

[0105] In machine learning and deep learning, choosing a suitable loss function is crucial for model training and optimization. The loss function measures the difference between the model's predicted value and the true value, and the optimization process aims to minimize this difference. In one embodiment of the present disclosure, the decoration template recommendation model is optimized by using a cross entropy loss function or a mean square error loss function. Among them, the cross entropy loss function is generally used to deal with classification problems, especially when the output layer uses a Softmax function for multi-classification. Cross entropy loss can handle the difference between probability distributions well, and guide model learning by reducing the KL divergence (Kullback-Leibler Divergence) between the predicted probability distribution and the true probability distribution. The mean square error loss function can provide a method for quantifying the accuracy of predictions for the model by calculating the average of the squares of the differences between the predicted values ​​and the true values.

[0106] During model training, after the predicted value is calculated through forward propagation, the difference between the predicted value and the true value is calculated using the loss function. The loss value can be passed back from the output layer to the input layer layer by layer through the back propagation algorithm, while calculating the gradient of the parameters of each layer. These gradients indicate how to adjust the parameters to minimize the loss. Finally, the model parameters are updated using the optimization algorithm based on the gradient.

[0107] In addition, during the model training process, in addition to the training set, a validation set is also required to evaluate the performance of the model. By monitoring the performance indicators on the validation set (such as accuracy, recall, etc.), it is possible to determine whether the model is overfitting or underfitting, and adjust the training strategy accordingly. At the same time, the performance of the validation set can be used as an evaluation indicator to find the hyperparameter combination that makes the model perform best on the validation set.

[0108] Deploy the trained model to the system that determines the live broadcast room presentation effect, which is used to receive product information in real time and generate a list of recommended templates. Continuously optimize and update the model based on user feedback data.

[0109] The above method can improve the efficiency of decoration template recommendation, thereby meeting users' diverse decoration needs and preferences.

[0110] In some embodiments, the live broadcast room information includes at least one of the anchor position, camera viewing angle, and user comment area information;

[0111] According to the information of the live broadcast room and the decoration template, the presentation effect of the live broadcast room is determined, including:

[0112] Determine the configuration information of elements in the decoration template;

[0113] According to the information of the live broadcast room, the configuration information of the elements in the decoration template is adjusted to generate the presentation effect of the live broadcast room.

[0114] In the disclosed embodiment, the elements in the decoration template include:

[0115] (1) Background design: static images, dynamic images, real-time videos, color gradient designs or pattern designs, etc.

[0116] (2) Decorative elements: pictures, illustrations, icons, animation effects (such as flashing, moving, etc.), virtual items (such as balloons, gifts, etc.), etc.

[0117] (3) Text style: text box, font style, font effects, dynamic text effects, etc.

[0118] (4) Layout design: layout of product display area, anchor position and size, audience interaction area location, etc.

[0119] (5) Brand elements: brand logo, slogan, color, etc.

[0120] (6) Sound effects and background music: music that is suitable for the atmosphere of the live broadcast room, interactive or prompt sound effects, etc.

[0121] Adjust the above elements of the decoration template according to the information of the live broadcast room. For example, if the audience asks questions about the product, the anchor can make the following adjustments to the template: adjust the audience interaction area to the center of the live broadcast room, and add a striking question and answer icon or border to highlight the interactive area; pause or reduce the dynamic effect of the product display area to avoid distracting the audience's attention from the interactive session.

[0122] After the decoration template is determined, in order to ensure that the visual effects of the live broadcast room can be presented to the audience smoothly and with high quality, a powerful graphics rendering library can be used for real-time rendering. Graphics rendering libraries, such as Open Graphics Library (OpenGL), DirectX (Direct eXtension), Web Graphics Library (WebGL), etc., can efficiently process complex calculations such as conversion of graphics data, lighting, texture mapping, and shadow effects, thereby generating realistic three-dimensional images or enhanced two-dimensional visual effects.

[0123] For web-based live broadcast rooms, HyperText Markup Language (HTML) and Cascading Style Sheets (CCS) are the core technologies for building and beautifying the live broadcast room interface. HTML is responsible for defining the structure of the live broadcast room, that is, the skeleton of the page. It organizes content through tags, such as video players, product display areas, audience interaction areas, etc. CSS is responsible for the style rendering of these contents, including layout, color, font, spacing, animation effects, etc., and applies styles to HTML elements through selectors (such as class selectors, ID selectors, element selectors, etc.), thereby achieving visual beautification and layout adjustment. In addition, for complex interactive effects or animations, JavaScript (or libraries such as jQuery) can also be combined to achieve this. JavaScript can process user input (such as clicks, scrolling, etc.) and dynamically modify HTML and CSS based on these inputs, thereby achieving a richer interactive experience.

[0124] In addition, you can also use game engines such as Unity and Unreal Engine (UE) for more complex 3D rendering and interactive design. Unity and Unreal Engine can handle complex scene construction, lighting simulation, physical effects, and high-quality material and texture mapping. In the live broadcast room effect presentation, these features can be fully utilized to create a realistic virtual environment, such as simulating real product display space, building complex stage sets, etc. In addition to 3D rendering, Unity and Unreal Engine also have powerful interactive design capabilities. By writing scripts or using the engine's built-in interactive system, you can easily implement various complex interactive logics, such as audience interaction with virtual products, participation in gamified lottery activities, etc. These interactive designs can not only increase the fun of the live broadcast room, but also enhance the audience's sense of participation.

[0125] By adopting the above methods, the live broadcast room can be more in line with actual needs and enhance the audience's sense of involvement and interactivity.

[0126] The disclosed embodiment also provides a device for determining the presentation effect of a live broadcast room. Figure 4 4 is a schematic diagram of a structure of a device 400 for determining a live broadcast room presentation effect according to an embodiment of the present disclosure, including:

[0127] A first determination module 410 is used to determine the information of the live broadcast room;

[0128] The feature recognition module 420 is used to perform multimodal recognition on the live broadcast commodities in the live broadcast room to determine the commodity features of the live broadcast commodities;

[0129] The template recommendation module 430 is used to recommend decoration templates for live broadcast products according to product characteristics;

[0130] The second determination module 440 is used to determine the presentation effect of the live broadcast room according to the information and decoration template of the live broadcast room.

[0131] In some implementations, multimodal recognition includes at least one of image recognition, text recognition, and speech recognition.

[0132] In some implementations, the product feature includes at least one of a color feature, a brand feature, a material feature, and a type.

[0133] In some embodiments, the color feature includes at least one of the main color of the product and the color corresponding to the product type; the brand feature includes at least one of the brand style and the brand color; the material feature includes at least one of the material visual effect and the material texture.

[0134] In some embodiments, the template recommendation module 430 is used to search a pre-saved list of decoration templates based on product features to determine the relevance of each decoration template to the product features; based on the relevance, determine the recommendation score of each decoration template for the live broadcast product; based on the recommendation score, recommend a decoration template for the live broadcast product.

[0135] In some embodiments, the template recommendation module 430 is used to perform weighted summation of the correlations between the decoration template and each product feature of the live broadcast product and the weights corresponding to each correlation to obtain a recommendation score of the decoration template for the live broadcast product.

[0136] The disclosed embodiment also provides a device for determining the presentation effect of a live broadcast room. Figure 5 is a structural diagram of a device 500 for determining a live broadcast room presentation effect according to an embodiment of the present disclosure, comprising:

[0137] The adjustment module 550 is used to adjust the weight corresponding to the relevance according to business needs and information of the live broadcast room.

[0138] In some implementations, the device 500 for determining the presentation effect of the live broadcast room further includes:

[0139] The saving module 560 is used to determine and save the decoration template list; wherein,

[0140] The decoration template list includes multiple decoration templates and includes the relevance of each decoration template to various product features.

[0141] In some embodiments, the template recommendation module 430 is used to generate a feature vector of each product feature of the live broadcast product based on the product features;

[0142] The feature vector is input into a pre-trained decoration template recommendation model, and the decoration template recommendation model recommends decoration templates for the live broadcast products.

[0143] In some embodiments, the live broadcast room information includes at least one of the anchor position, camera viewing angle, and user comment area information;

[0144] The second determination module 440 is used to determine the configuration information of the elements in the decoration template;

[0145] According to the information of the live broadcast room, the configuration information of the elements in the decoration template is adjusted to generate the presentation effect of the live broadcast room.

[0146] For the description of specific functions and examples of each module and submodule of the device in the embodiment of the present disclosure, reference can be made to the relevant description of the corresponding steps in the above method embodiment, which will not be repeated here.

[0147] In the technical solution disclosed in the present invention, the acquisition, storage and application of personal information of users involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0148] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0149] Figure 6 A schematic block diagram of an example electronic device 600 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0150] like Figure 6 As shown, the device 600 includes a computing unit 601, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the device 600 can also be stored. The computing unit 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0151] A number of components in the device 600 are connected to the I / O interface 605, including: an input unit 606, such as a keyboard, a mouse, etc.; an output unit 607, such as various types of displays, speakers, etc.; a storage unit 608, such as a disk, an optical disk, etc.; and a communication unit 609, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 609 allows the device 600 to exchange data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0152] The computing unit 601 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 601 performs the various methods and processes described above, such as detection methods. For example, in some embodiments, the detection method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the detection method described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform the detection method in any other appropriate manner (e.g., by means of firmware).

[0153] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0154] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0155] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0156] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0157] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0158] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0159] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0160] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the principles of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for determining a live broadcast room presentation effect, comprising: Determine the live broadcast room information; Performing multimodal recognition on the live broadcast commodity in the live broadcast room to determine commodity features of the live broadcast commodity; According to the characteristics of the product, recommend a decoration template for the live broadcast product; The presentation effect of the live broadcast room is determined according to the information of the live broadcast room and the decoration template.

2. The method according to claim 1, wherein: The multimodal recognition includes at least one of image recognition, text recognition and speech recognition.

3. The method according to claim 1 or 2, wherein: The product feature includes at least one of a color feature, a brand feature, a material feature and a type.

4. The method according to claim 3, wherein: The color feature includes at least one of the main color of the product and the color corresponding to the product type; The brand characteristics include at least one of brand style and brand color; The material feature includes at least one of a material visual effect and a material texture.

5. The method according to claim 3 or 4, wherein: The recommending a decoration template for the live broadcast product according to the product characteristics includes: Based on the product characteristics, searching a pre-saved decoration template list to determine the relevance of each decoration template to the product characteristics; Based on the relevance, determining a recommendation score of each decoration template for the live broadcast product; Based on the recommendation score, a decoration template is recommended for the live broadcast product.

6. The method according to claim 5, wherein: Determining a recommendation score of the decoration template for the live broadcast product based on the relevance includes: Based on the correlation between the decoration template and each product feature of the live broadcast product, and the weights corresponding to the each correlation, the each correlation is weightedly summed to obtain the recommendation score of the decoration template for the live broadcast product.

7. The method according to claim 6, further comprising: The weight corresponding to the relevance is adjusted according to business needs and the information of the live broadcast room.

8. The method according to claim 5 further comprises determining and saving the decoration template list; wherein: The decoration template list includes multiple decoration templates and includes the relevance of each decoration template to various product features.

9. The method according to claim 3 or 4, wherein: The recommending a decoration template for the live broadcast product according to the product characteristics includes: Based on the product features, generating a feature vector of each product feature of the live broadcast product; The feature vector is input into a pre-trained decoration template recommendation model, and the decoration template recommendation model recommends a decoration template for the live broadcast product.

10. The method according to any one of claims 1 to 9, wherein: The live broadcast room information includes at least one of the anchor position, camera viewing angle and user comment area information; Determining the presentation effect of the live broadcast room according to the information of the live broadcast room and the decoration template includes: Determine the configuration information of the elements in the decoration template; According to the information of the live broadcast room, the configuration information of the elements in the decoration template is adjusted to generate the presentation effect of the live broadcast room.

11. A device for determining a live broadcast room presentation effect, comprising: A first determination module is used to determine the information of the live broadcast room; A feature recognition module, used to perform multimodal recognition on the live broadcast commodity in the live broadcast room to determine the commodity features of the live broadcast commodity; A template recommendation module, used to recommend a decoration template for the live broadcast product according to the product characteristics; The second determination module is used to determine the presentation effect of the live broadcast room according to the information of the live broadcast room and the decoration template.

12. The device according to claim 11, wherein The multimodal recognition includes at least one of image recognition, text recognition and speech recognition.

13. The device according to claim 11 or 12, wherein: The product feature includes at least one of a color feature, a brand feature, a material feature and a type.

14. The device according to claim 13, wherein: The color feature includes at least one of the main color of the product and the color corresponding to the product type; The brand characteristics include at least one of brand style and brand color; The material feature includes at least one of a material visual effect and a material texture.

15. The device according to claim 13 or 14, wherein: The template recommendation module is used to: Based on the product characteristics, searching a pre-saved decoration template list to determine the relevance of each decoration template to the product characteristics; Based on the relevance, determining a recommendation score of each decoration template for the live broadcast product; Based on the recommendation score, a decoration template is recommended for the live broadcast product.

16. The device according to claim 15, wherein: The template recommendation module is used to: Based on the correlation between the decoration template and each product feature of the live broadcast product, and the weights corresponding to the each correlation, the each correlation is weightedly summed to obtain the recommendation score of the decoration template for the live broadcast product.

17. The apparatus according to claim 16, further comprising: The adjustment module is used to adjust the weight corresponding to the relevance according to business needs and information of the live broadcast room.

18. The apparatus according to claim 15, further comprising: A saving module is used to determine and save the decoration template list; wherein, The decoration template list includes multiple decoration templates and includes the relevance of each decoration template to various product features.

19. The device according to claim 13 or 14, wherein: The template recommendation module is used to: Based on the product features, generating a feature vector of each product feature of the live broadcast product; The feature vector is input into a pre-trained decoration template recommendation model, and the decoration template recommendation model recommends a decoration template for the live broadcast product.

20. The device according to any one of claims 11 to 19, wherein: The live broadcast room information includes at least one of the anchor position, camera viewing angle and user comment area information; The second determining module is used for: Determine the configuration information of the elements in the decoration template; According to the information of the live broadcast room, the configuration information of the elements in the decoration template is adjusted to generate the presentation effect of the live broadcast room.

21. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 10.

22. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-10.

23. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 10.

Images