Personalized publicity map generation method based on stable diffusion image generation in retail e-commerce scene

Abstract

The invention relates to the field of retail e-commerce and image processing, in particular to a personalized publicity map generation method based on stable diffusion image generation in a retail e-commerce scene. Comprising a style template selection module, a text-to-graph and graph-to-graph module, a canvas preview and adjustment module, a post-processing and output module and a data analysis and optimization module. According to the method, a stable diffusion image generation technology is utilized, a user is allowed to customize the content and style of an e-commerce banner or a propaganda map according to own requirements or consumer comment pictures, and the propaganda map meeting personalized requirements is generated. According to the method, the brand image can be enhanced, the product exposure and sales effect can be improved, the user participation and loyalty can be enhanced, and the targets of personalized recommendation and directional marketing can be realized. Through the customized propaganda map, the e-commerce platform can better interact with the user, the personalized demand of the user is met, and more attractive and personalized shopping experience is provided.

Description

Technical Field

[0001] The present invention mainly relates to the field of retail e-commerce and image processing, and in particular to a method for generating personalized promotional images based on stable diffusion image generation in retail e-commerce scenarios. The method can be applied in a variety of e-commerce environments, such as product promotion and sales, marketing activities and special offers, brand promotion and image building, and user-generated content (UGC) display. The present invention can enhance brand image, improve product exposure and sales results, enhance user engagement and loyalty, and achieve the goals of personalized recommendations and targeted marketing. Through customized promotional images, e-commerce platforms can better interact with users, meet users' personalized needs, and provide a more attractive and personalized shopping experience. Background Art

[0002] Currently, promotional graphics for e-commerce platforms are usually designed manually by professional designers. This process is time-consuming and labor-intensive, and it is difficult to achieve large-scale personalized promotional graphics generation. Existing automated promotional graphics generation technologies can only generate promotional graphics in fixed styles, lacking personalized and differentiated design elements, and are difficult to meet users' personalized needs. Summary of the invention

[0003] The present invention provides a method for generating personalized promotional images based on stable diffusion image generation. Specifically, it includes the following modules: 1. Style template selection module: manages the model library and generates the desired image style or layout through style and template selection.

[0004] 2. Text-to-image and image-to-image module: Based on keywords, input pictures, control network parameters, image elements, character postures, and finely control the expected output of generated images.

[0005] 3. Canvas preview and adjustment module: realize image preview, image redrawing, image repair and image filling in the unified canvas to achieve the ideal visual effect.

[0006] 4. Post-processing and output module: Improve image resolution and increase details through magnification technology.

[0007] 5. Data analysis and optimization module: Through data analysis, the most popular comments and pictures are obtained as the basis for promotional pictures, or promotional pictures of different styles are generated according to the different preferences of consumers, so as to achieve personalized promotional push for thousands of people.

[0008] This invention uses stable diffusion image generation technology to achieve personalized promotional image generation, which meets the personalized needs of e-commerce platforms and users, and helps to enhance the brand image of e-commerce platforms, increase product exposure and sales effects, and improve user participation and loyalty. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 Invention example process.

[0010] Figure 2 Style and keyword recommendation generation process.

[0011] Figure 3 The original input element.

[0012] Figure 4 Promotional image for the shopping festival’s electronic products event. DETAILED DESCRIPTION

[0013] The following will describe the specific implementation of the present invention in detail through an example, but this does not limit the scope of the present invention. According to actual needs and application environment, the specific implementation of the present invention can have various changes and adjustments.

[0014] 1. First, collect customer profile data. This data can include the customer's basic information, consumption habits, preferences, historical behavior, etc.

[0015] 2. Use the recommendation system to recommend style templates and prompt keywords based on customer profile data. The specific recommendation process can be as follows: Figure 2 As shown, it includes building a customer portrait, analyzing the portrait data to generate customer preference tags, selecting a user preference model, selecting a recommendation algorithm, and finally generating a recommended keyword prompt.

[0016] 3. If the recommendation system can successfully recommend results, then you can proceed to the next step to receive the recommendation results; otherwise, use the default template and common keywords.

[0017] 4. Based on the recommendation results of the recommendation system, customers can independently adjust the style template and generate keywords. Furthermore, they can adjust the low-rank large language model (LoRA model), adjust the task image or item features to better meet the targeted needs of the product.

[0018] 5. Use the ControlNet and its model parameters to adjust the image elements and character pose expectations. The ControlNet allows users to have more detailed control over the generated images, such as the position, size, color of image elements, character poses, etc. If the customer sharing mode is turned on, select product promotional reference images from the highly praised photos as the user-generated content (UGC) input source.

[0019] 6. Through the text-generated image and image-generated image modules, using the input Prompt keyword, template image and ControlNet to control the general direction of image generation. Users can preview the generated image and adjust the image, such as image filling and partial redrawing, until they are satisfied.

[0020] 7. Enter the post-processing stage, and perform facial processing and resolution enhancement on the generated image. Facial processing includes optimizing the face of the person in the generated image, such as improving facial expressions, skin texture, etc. Resolution enhancement includes increasing the image resolution through magnification technology to make the generated promotional image clearer and more detailed.

[0021] 8. Generate the final personalized promotional image and end the entire process.

[0022] To show the effect, you can Figure 3 The original input element, generated as Figure 4 As shown, this is a promotional picture for the electronic products event during the shopping festival.

[0023] The above is a specific embodiment of the present invention. Although the present invention is described in a specific manner, it does not mean that the embodiment of the present invention is limited to this. Any modification and change to the technical solution, as long as it does not depart from the spirit and scope of the present invention, should be regarded as within the protection scope of the present invention.

Claims

1. A method for generating personalized promotional images based on stable diffusion images in a retail e-commerce scenario, characterized in that: The method includes the following modules: a style template selection module, a text-generated image and image-generated image module, a canvas preview and adjustment module, a post-processing and output module, and a data analysis and optimization module.

2. The method of claim 1, wherein: The style template selection module is used to manage the model library and generate the desired image style or layout through style and template selection.

3. The method according to claim 1 or 2, wherein: The text-generated image and image-generated image modules input images based on keywords, control network parameters, control image elements, and character postures, and finely control the expected output of the generated image.

4. The method according to claim 1 or 2, wherein: The canvas preview and adjustment module implements image preview, image redrawing, image repair and image expansion in a unified canvas to achieve ideal visual effects.

5. The method according to claim 1 or 2, wherein: The post-processing and output modules improve image resolution and increase details through magnification technology.

6. The method according to claim 1 or 2, wherein: The data analysis and optimization module uses data analysis to obtain the most popular comments and pictures as the basis for promotional pictures, or generates promotional pictures of different styles based on the different preferences of consumers, so as to achieve personalized promotional push for thousands of people.

Images