Method for automatically generating multi-part customizable bag line draft by using AI (Artificial Intelligence)

Abstract

The invention provides a method for automatically generating a multi-part customizable bag line draft by using AI. The method comprises the following steps: S1, making a training data set; s2, copying a copy of the Stable Diffusion model; s3, inputting the training data set into a copy of the Stable Diffusion model for training; s4, testing the copy of the trained Stable Diffusion model, and carrying out the test on the copy of the trained Stable Diffusion model; and S5, inputting a specific package description in the tested Stable Diffusion model copy, and outputting a package line draft file.

Description

Technical Field

[0001] The present invention relates to a method for automatically generating line drawings of multi-position customizable bags by using AI. Background Art

[0002] In today's bag design industry, line drawings are the key link from concept creativity to actual product implementation. Traditional line drawing generation methods usually rely on designers to draw manually, which not only requires a lot of time and energy, but also the results are easily affected by subjective experience and professional level, making it difficult to achieve efficient and stable design output in a rapidly changing market environment. Although some automated drawing tools based on computer-aided design (CAD) provide certain conveniences for the design process, most of them can only generate basic geometric outlines, and lack flexible and controllable presentation methods for the functional divisions and personalized details of various parts of the bag (such as shoulder straps, handles, compartments, hardware accessories, etc.). In addition, these tools are usually unable to fully combine multi-dimensional information such as users' personalized preferences, brand positioning, and fashion trends, and often require secondary modifications and multiple iterations by designers to meet the needs of commercial applications.

[0003] With the rapid development of deep learning and generative AI technologies, especially the diffusion model represented by Stable Diffusion, breakthrough progress has been made in the field of image generation. This type of model has outstanding performance in terms of fidelity and detail richness, and has the potential for multimodal conditional constraints, which can integrate various forms of information such as text and images into the image generation process. However, in the subdivided design scenario of bags, there are still many challenges in relying solely on general diffusion models: first, it is difficult for general models to make fine distinctions and functional positioning for different parts of bags (such as bag body, shoulder straps, zippers, hardware, etc.); second, the existing models are still insufficient in the coordinated control of factors such as style, function and structure, and it is difficult to accurately present users' customized needs for details; third, how to effectively integrate brand positioning, market trends and user-specific preferences into the generation process is still a pain point that the industry needs to solve urgently. Summary of the invention

[0004] The purpose of the present invention is to provide a method for automatically generating line drawings of multi-part customizable bags using AI to solve the problems raised in the above background technology. To achieve the above purpose, the present invention provides the following technical solutions:

[0005] A method for automatically generating line drawings of multi-part customizable bags using AI includes the following steps:

[0006] S1: Create a training data set;

[0007] S2: copy the Stable Diffusion model;

[0008] S3: Input the training data set into the Stable Diffusion model copy for training;

[0009] S4: Test the Stable Diffusion model copy after training;

[0010] S5: Input the specific bag description into the Stable Diffusion model copy after testing and output the bag line drawing file.

[0011] In S1, the following steps are included:

[0012] S11: Collect, screen and organize bag design drawings of various styles and structures to form a sample image dataset for training;

[0013] S12: Preprocessing the sample image data set, wherein the preprocessing includes removing duplicate, blurred, or severely distorted images.

[0014] S13: vectorizing or semi-automatically labeling the bag parts of the sample image data set;

[0015] S14: By performing content recognition on each image in the sample image data set, extract or write corresponding Chinese and English keywords and description information according to its style, function or decorative element information;

[0016] S14: Use image augmentation to expand the size of the sample image dataset and improve the generalization ability of the model, while processing all sample image datasets into a unified size and format.

[0017] The pretreatment is performed by manual or semi-automatic cleaning.

[0018] In S2, the following steps are included:

[0019] S21: Create a new copy based on the existing Stable Diffusion base model, which has the same network structure and pre-trained weights as the original model;

[0020] S22: This copy can be further fine-tuned for the bag design task.

[0021] In S3, the following steps are included:

[0022] S31: The keywords associated with each training dataset are used as conditional inputs for the Stable Diffusion model replica to guide the model to learn the relationship between different bag parts and the overall shape;

[0023] S32: Set reasonable loss functions and hyperparameters, observe the generated images and loss rate indicators during the training process, and adjust the model weights in a timely manner so that the model can better generate line drawings of various parts of the bag that meet the requirements.

[0024] In S4, the following steps are included:

[0025] S41: Input the line drawing requirements of bags with different styles, materials and structures, and let the trained StableDiffusion model copy output the design sketch to observe whether the model can present the various parts of the bag as required and ensure the coordination of the overall style;

[0026] S42: If it is found that the model does not adequately recognize some details or there is style confusion, retraining or optimization can be performed by readjusting the distribution of training data, adding or modifying keywords, and fine-tuning hyperparameters.

[0027] Beneficial effects:

[0028] The present invention establishes and trains an AI model that can distinguish different parts of a bag, so that when automatically generating a bag line draft, the details of each part can be flexibly adjusted according to different needs. Compared with the traditional design method, the present invention has the following beneficial effects:

[0029] 1. Significantly improve design efficiency

[0030] With the help of AI's accurate recognition and rapid generation of the structure and function of each part of the bag, the present invention can produce a variety of different design schemes in a short period of time, reducing the repetitive work of designers in manual sketching and repeated modifications, and greatly shortening the transition period from concept to preliminary visual presentation.

[0031] 2. Enhance creativity and customization capabilities

[0032] Through targeted training of AI models, the present invention can generate diversified designs for different styles, uses and detail requirements, including the overall shape of the bag, functional divisions and decorative elements, etc.; it can also flexibly meet and quickly generate exclusive line drafts for personalized customization or small batch production needs, effectively improving the diversity and uniqueness of the design.

[0033] 3. Optimize designer workflow and reduce costs

[0034] This invention can serve as an "automated assistant" for designers, allowing them to focus more on creativity and decision-making, while leaving a large amount of basic and repetitive drawing work to AI. By reducing the time investment in manual drawing and modification, it can reduce labor costs, greatly improve efficiency in the design process, and shorten project cycles.

[0035] 4. Provide rich teaching and training materials

[0036] In the field of education and training, the present invention can generate examples of bag line drawings for different design styles and detail requirements to help students understand and master diverse design ideas and methods; at the same time, it can also automatically transform and modify details to give students the opportunity to compare and refer to them during practice, thereby improving teaching resources and expanding the depth of teaching.

[0037] 5. Improve market competitiveness and user satisfaction

[0038] Through rapid iteration and precise adjustment, the present invention can implement creative ideas more quickly and make rapid modifications based on real-time market feedback or customer customization requirements, thereby seizing the initiative in a fierce market environment and improving the speed of new product launches and product competitiveness. At the same time, it can improve customer satisfaction and enhance cooperation stickiness with line draft solutions that are more in line with user needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a method diagram for automatically generating line drawings of bags that can be customized in multiple parts;

[0040] Figure 2 It is an example graph of the sample image dataset. DETAILED DESCRIPTION

[0041] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] See Figure 1 The present invention provides a method for automatically generating line drawings of multi-part customizable bags using AI, comprising the following steps:

[0043] S1: Create a training dataset:

[0044] S11: Collect, screen and organize bag design drawings of various styles and structures to form a sample image dataset for training (such as Figure 2 (As shown in the figure) This dataset contains clear labels and annotations of various bag parts (such as handles, bag body, zipper, flap, buckle, tassel, metal accessories, etc.) to ensure that the subsequent model can recognize and generate different parts.

[0045] S12: Preprocessing the sample image data set, wherein the preprocessing includes removing duplicate, blurred, severely distorted lighting or perspective images, and images that do not meet design specifications, etc. The preprocessing is performed by manual or semi-automatic cleaning.

[0046] S13: Vectorize or semi-automatically annotate the bag parts of the sample image data set to ensure accurate data annotation.

[0047] S14: By performing content recognition on each image in the sample image dataset, extract or write corresponding Chinese and English keywords and description information (such as "fringe handle cowhide women's bag", "detachable strap travel bag", "metal buckle business briefcase", etc.) based on its style, function or decorative elements. These keywords will serve as an important reference for subsequent Stable Diffusion model training and reasoning.

[0048] S14: Use image augmentation (such as rotation, flipping, scaling, cropping, etc.) to expand the scale of the sample image dataset and improve the generalization ability of the model, while processing all sample image datasets into a unified size and format.

[0049] S2: Copy the Stable Diffusion model

[0050] S21: Create a new copy based on the existing Stable Diffusion base model, which has the same network structure and pre-trained weights as the original model.

[0051] S22: This copy can be further fine-tuned for the bag design task so that more precise details can be output in the subsequent line drawing generation for different parts.

[0052] S3: Input the training data set into the Stable Diffusion model copy for training

[0053] S31: The keywords associated with each training data set (including material, function and appearance description, etc.) will be used as conditional inputs for the Stable Diffusion model copy to guide the model to learn the relationship between different bag parts and the overall shape;

[0054] S32: Set reasonable loss functions and hyperparameters, observe the generated images, loss rate and other indicators during the training process, and adjust the model weights in a timely manner so that the model can better generate line drawings of various parts of the bag that meet the requirements.

[0055] S4: Test the Stable Diffusion model copy after training

[0056] S41: Input the line drawing requirements of bags of different styles, materials and structures, and let the trained StableDiffusion model copy output the design sketch to observe whether the model can present the various parts of the bag as required (such as the curvature of the handle, the structure of the bag body, the details of the buckle) and ensure the coordination of the overall style.

[0057] If it is found that the model does not adequately recognize some details or there is style confusion, it can be retrained or optimized by readjusting the distribution of training data, adding or modifying keywords, fine-tuning hyperparameters, etc.

[0058] When the model copy is trained, the accuracy and clarity of the line drawings generated by the model are verified through a test set or manual testing.

[0059] S5: Input the specific bag description into the Stable Diffusion model copy after testing and output the bag line drawing file.

[0060] Finally, the stably trained and tested model can be put into use: input a specific bag description, including the functional demands of different parts (such as "leather handles", "metal zippers", "detachable shoulder straps", "larger main bag body", etc.), color matching style, decorative elements, etc., and the model will automatically generate the corresponding line drawing.

[0061] At the same time, multiple candidate line drafts can be output according to needs, so that designers or users can obtain multiple styles for comparison and screening in a short time; and some details can be further fine-tuned or manually corrected.

[0062] The above is only a preferred embodiment of the present invention, and does not limit the present invention in any form. The protection scope of the present invention shall be subject to the protection scope of the claims. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technician familiar with this profession can make some changes or modifications to equivalent embodiments of equivalent changes using the technical content disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of the technical solution of the present invention.

Claims

1. A method for automatically generating line drawings of multi-part customizable bags using AI, characterized by: The following steps are involved: S1: Create a training data set; S2: copy the Stable Diffusion model; S3: Input the training data set into the Stable Diffusion model copy for training; S4: Test the Stable Diffusion model copy after training; S5: Input the specific bag description into the Stable Diffusion model copy after testing and output the bag line drawing file.

2. The method for automatically generating line drawings of multi-part customizable bags using AI according to claim 1, characterized in that: In S1, the following steps are included: S11: Collect, screen and organize bag design drawings of various styles and structures to form a sample image dataset for training; S12: Preprocessing the sample image data set, wherein the preprocessing includes removing duplicate, blurred, or severely distorted images. S13: vectorizing or semi-automatically labeling the bag parts of the sample image data set; S14: By performing content recognition on each image in the sample image data set, extract or write corresponding Chinese and English keywords and description information according to its style, function or decorative element information; S14: Use image augmentation to expand the size of the sample image dataset and improve the generalization ability of the model, while processing all sample image datasets into a unified size and format.

3. The method for automatically generating line drawings of multi-part customizable bags using AI according to claim 1, characterized in that: The pretreatment is performed by manual or semi-automatic cleaning.

4. The method for automatically generating line drawings of multi-part customizable bags using AI according to claim 1, characterized in that: In S2, the following steps are included: S21: Create a new copy based on the existing Stable Diffusion base model, which has the same network structure and pre-trained weights as the original model; S22: This copy can be further fine-tuned for the bag design task.

5. The method for automatically generating line drawings of multi-part customizable bags using AI according to claim 1, characterized in that: In S3, the following steps are included: S31: The keywords associated with each training dataset are used as conditional inputs for the Stable Diffusion model replica to guide the model to learn the relationship between different bag parts and the overall shape; S32: Set reasonable loss functions and hyperparameters, observe the generated images and loss rate indicators during the training process, and adjust the model weights in a timely manner so that the model can better generate line drawings of various parts of the bag that meet the requirements.

6. The method for automatically generating line drawings of multi-part customizable bags using AI according to claim 1, characterized in that: In S4, the following steps are included: S41: Input the line drawing requirements of bags with different styles, materials and structures, and let the trained StableDiffusion model copy output the design sketch to observe whether the model can present the various parts of the bag as required and ensure the coordination of the overall style; S42: If it is found that the model does not adequately recognize some details or there is style confusion, retraining or optimization can be performed by readjusting the distribution of training data, adding or modifying keywords, and fine-tuning hyperparameters.

Images