System for providing best-selling product design generation service by using user image-based generative ai

Abstract

A system for providing a best-selling product design generation service by using user image-based generative artificial intelligence (AI) is provided, the system comprising: a planning terminal for selecting the character and the category of a product to be generated, and outputting a generated image in which the character is applied to the category of the product; and a generation service providing server including a product training unit for updating a user image of at least one category of the product on the basis of a preference score for the user image within the at least one category of the product, and training a fine-tuning model by using the updated user image, a character training unit for collecting character images of at least one character, and registering the character images so that the at least one character is added to a character list of the fine-tuning model, and an image generation unit for when the planning terminal selects the character and the category of the product, receiving a generated image in which the character is included in the product from pre-constructed generative AI, and providing the generated image to the planning terminal.

Description

A best-selling product design creation service system using user image-based generative AI.

[0001] The present invention relates to a system for providing a best-selling product design generation service using a user image-based generative AI, and provides a system in which a generative AI generates a product design when a product category and character are specified, using a fine-tuned model learned based on a user image with a high preference score.

[0002] The recent golden age of character goods is truly upon us, with products featuring game and animation characters dominating the retail market. This popularity transcends all industries. From convenience stores and hypermarkets to snack, beverage, liquor, and food companies, everyone is releasing character collaboration products. The popularity of character goods stems from two factors: first, a broad and established fan base. Character IPs attract a wide range of fans, with a strong loyalty base. Second, the diversification of collaborative products. Recent trends, such as keychains, glasses, and cushions, have expanded the product line, and the emergence of pop-up stores has diversified sales channels. Third, the economic downturn. When the economy slows, consumers tighten their purse strings, forcing companies to prioritize their products. This phenomenon is occurring as companies focus on eye-catching marketing strategies amidst depressed consumer sentiment. The growing purchasing power of the MZ generation, with its strong preferences for personalized products, coupled with a prolonged period of low growth, is driving explosive growth in the character collaboration market.

[0003] At this time, a method of investigating preferences for characters and product groups or presenting characters by synthesizing them into products was studied and developed. In relation to this, prior art Korean Patent Registration No. 10-1780042 (announced on September 19, 2017) and Korean Patent Publication No. 2022-0169875 (published on December 28, 2022) disclose a configuration in which, when a productization request is received from a user terminal, a character is specified, a product group is selected, and the character and product group selected by the user are stored and saved as preference survey data, and when a character and product are selected from a user terminal, a configuration in which a character within a product is positioned and synthesized, and when the user terminal inquires about the production price of a product including a character, a licensing cost for this is calculated and provided.

[0004] However, the former configuration only discloses the announcement of a poll regarding the popularity of the game's characters. Even in the latter case, it only discloses a composite composition where the character is overlaid on the product image, rather than a product design incorporating the character. When planning character products, it's crucial to choose the right character, but it's also crucial to determine which product category that character will be applied to. Even if both are selected, the actual product design still needs to be envisioned. If the product design begins as an idea without a clear direction, it sets the stage for endless design revisions. Therefore, research and development are needed to develop a system that can plan products based on user preferences when planning character products.

[0005] One embodiment of the present invention provides a best-selling product design generation service system using a user image-based generative AI, which can generate a product design that is most likely to sell based on preference while providing a visual draft rather than an idea state so that the product design does not have to be infinitely modified, by having a generative AI generate an image so that a character is applied to the product category based on a fine-tuned model trained with user images having high preference scores in the product category and a fine-tuned model trained with character images when a product category and character are designated in a planning terminal. However, the technical problem to be achieved by the present embodiment is not limited to the technical problem described above, and other technical problems may exist.

[0006] As a technical means for achieving the above-described technical task, one embodiment of the present invention includes a generation service providing server including a planning terminal for selecting a category and character of a product to be generated and outputting a generated image in which a character is applied to the product category, a product learning unit for updating a user image of at least one product category based on a preference score for a user image within at least one product category and training a fine-tuning model with the updated user image, a character learning unit for collecting a character image of at least one character and registering the at least one character to be added to a character list of a fine-tuning model, and an image generation unit for receiving a generated image including a character in a product from a pre-established generative artificial intelligence (Generative Artificial Intelligence) when a product category and character are selected from the planning terminal and providing the generated image to the planning terminal.

[0007] According to any one of the problem solving means of the present invention described above, when specifying a product category and character in a planning terminal, by having a generative AI generate an image so that a character is applied to the product category based on a fine-tuned model learned with a user image having a high preference score in the product category and a fine-tuned model learned with a character image, it is possible to generate a product design that is most likely to sell based on preference, while providing a visible draft rather than an idea state so that infinite modification of the product design does not occur.

[0008] FIG. 1 is a diagram illustrating a system for providing a best-selling product design creation service using user image-based generative AI according to one embodiment of the present invention.

[0009] Figure 2 is a block diagram for explaining a generation service providing server included in the system of Figure 1.

[0010] FIG. 3 and FIG. 4 are drawings for explaining an embodiment of a best-selling product design creation service using a user image-based generative AI according to one embodiment of the present invention.

[0011] FIG. 5 is a flowchart illustrating a method for providing a best-selling product design creation service using a user image-based generative AI according to one embodiment of the present invention.

[0012] Below, with reference to the attached drawings, embodiments of the present invention are described in detail so that those skilled in the art can easily implement them. However, the present invention may be implemented in various different forms and is not limited to the embodiments described herein. In the drawings, irrelevant parts have been omitted for clarity of description, and similar reference numerals have been used throughout the specification to indicate similar elements.

[0013] Throughout the specification, when a part is said to be "connected" to another part, this includes not only the case where it is "directly connected" but also the case where it is "electrically connected" with another element in between. Furthermore, when a part is said to "include" a component, this should be understood to mean that, unless specifically stated to the contrary, it may include other components rather than excluding them, and does not preclude the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0014] The terms "about," "substantially," and the like used throughout the specification are used in a sense of degree or in a sense close to the numerical value when manufacturing and material tolerances inherent to the meanings mentioned are presented, and are used to prevent unscrupulous infringers from unfairly exploiting disclosures that mention precise or absolute numerical values ​​to aid understanding of the present invention. The terms "step of doing" or "step of" used throughout the specification of the present invention do not mean "step for doing."

[0015] In this specification, the term "unit" includes a unit realized by hardware, a unit realized by software, and a unit realized using both. In addition, one unit may be realized by using two or more pieces of hardware, and two or more units may be realized by one piece of hardware. Meanwhile, the "unit" is not limited to software or hardware, and the "unit" may be configured to be on an addressable storage medium or may be configured to reproduce one or more processors. Accordingly, as an example, the "unit" includes components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functionality provided within the components and "units" may be combined into a smaller number of components and "units," or further separated into additional components and "units." Additionally, components and '~parts' may be implemented to regenerate one or more CPUs within a device or secure multimedia card.

[0016] Some of the operations or functions described herein as being performed by a terminal, apparatus, or device may instead be performed by a server connected to the terminal, apparatus, or device. Similarly, some of the operations or functions described herein as being performed by a server may also be performed by a terminal, apparatus, or device connected to the server.

[0017] In this specification, some of the operations or functions described as terminal and mapping or matching may be interpreted to mean mapping or matching the terminal's unique number or personal identification information, which is the terminal's identifying data.

[0018] The present invention will be described in detail with reference to the attached drawings below.

[0019] FIG. 1 is a diagram illustrating a best-selling product design generation service providing system using a user image-based generative AI according to one embodiment of the present invention. Referring to FIG. 1, the best-selling product design generation service providing system (1) using a user image-based generative AI may include at least one user terminal (100), a generation service providing server (300), and at least one planning terminal (400). However, the best-selling product design generation service providing system (1) using a user image-based generative AI of FIG. 1 is merely one embodiment of the present invention, and thus the present invention is not limited thereto through FIG. 1.

[0020] At this time, each component of FIG. 1 is generally connected via a network (Network, 200). For example, as illustrated in FIG. 1, at least one user terminal (100) can be connected to a generation service providing server (300) via a network (200). In addition, the generation service providing server (300) can be connected to at least one user terminal (100) and at least one planning terminal (400) via the network (200). In addition, at least one planning terminal (400) can be connected to the generation service providing server (300) via the network (200).

[0021] Here, a network means a connection structure that enables information exchange between each node, such as multiple terminals and servers, and examples of such networks include a local area network (LAN), a wide area network (WAN), the Internet (WWW), a wired and wireless data communication network, a telephone network, and a wired and wireless television communication network. Examples of wireless data communication networks include, but are not limited to, 3G, 4G, 5G, 3GPP (3rd Generation Partnership Project), 5GPP (5th Generation Partnership Project), 5G NR (New Radio), 6G (6th Generation of Cellular Networks), LTE (Long Term Evolution), WIMAX (World Interoperability for Microwave Access), Wi-Fi, the Internet, LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), RF (Radio Frequency), Bluetooth networks, NFC (Near-Field Communication) networks, satellite broadcasting networks, analog broadcasting networks, and DMB (Digital Multimedia Broadcasting) networks.

[0022] In the following, the term "at least one" is defined as a term including both singular and plural, and it will be clear that even if the term "at least one" does not exist, each component can exist in the singular or plural and can mean either the singular or plural. Furthermore, whether each component is provided in the singular or plural may vary depending on the embodiment.

[0023] At least one user terminal (100) may be a user terminal that uploads a user image using a web page, app page, program, or application related to a best-selling product design creation service using user image-based generative AI. In one embodiment of the present invention, the platform (Colley) includes a menu for uploading an image of a character product that the user likes, i.e., a user image, as shown in FIG. 4d , where the user can upload the user image.

[0024] Here, at least one user terminal (100) may be implemented as a computer capable of accessing a remote server or terminal via a network. Here, the computer may include, for example, a notebook, desktop, or laptop equipped with a navigation system or web browser. In this case, at least one user terminal (100) may be implemented as a terminal capable of accessing a remote server or terminal via a network. At least one user terminal (100) may include, for example, a wireless communication device that ensures portability and mobility, and may include all types of handheld-based wireless communication devices such as navigation, PCS (Personal Communication System), GSM (Global System for Mobile communications), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet) terminals, smartphones, smartpads, tablet PCs, etc.

[0025] The generation service providing server (300) may be a server that provides a best-selling product design generation service web page, app page, program, or application using a user image-based generative AI. In addition, the generation service providing server (300) may be a server that collects user images, extracts user images based on user preferences, and then has a product category learned by a fine-tuned model, and has the character images learned by the fine-tuned model in order to add characters to the character list of the fine-tuned model. In addition, the generation service providing server (300) may be a server that, when a product category and character are designated by a planning terminal (400), has a generative AI based on the learned fine-tuned model generate an image, and transmits the generated image, which is the generated image, to the planning terminal (400).

[0026] Here, the generation service providing server (300) may be implemented as a computer capable of connecting to a remote server or terminal via a network. Here, the computer may include, for example, a notebook computer, desktop computer, or laptop computer equipped with a navigation system or web browser.

[0027] At least one planning terminal (400) may be a terminal of a product planner that receives a generated image generated by a creation service providing server (300) after specifying a product category and character using a web page, app page, program, or application related to a best-selling product design creation service using user image-based generative AI. It goes without saying that this may include anyone who is not a planner but is responsible for creating character products.

[0028] Here, at least one planning terminal (400) may be implemented as a computer capable of connecting to a remote server or terminal via a network. Here, the computer may include, for example, a notebook, desktop, or laptop equipped with a navigation system or web browser. In this case, at least one planning terminal (400) may be implemented as a terminal capable of connecting to a remote server or terminal via a network. At least one planning terminal (400) may include, for example, all kinds of handheld-based wireless communication devices such as navigation, PCS (Personal Communication System), GSM (Global System for Mobile communications), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet) terminals, smartphones, smartpads, tablet PCs, etc., as wireless communication devices that ensure portability and mobility.

[0029] FIG. 2 is a block diagram for explaining a generation service providing server included in the system of FIG. 1, and FIGS. 3 and 4 are diagrams for explaining an embodiment in which a best-selling product design generation service using a user image-based generation AI according to one embodiment of the present invention is implemented.

[0030] Referring to FIG. 2, the generation service providing server (300) may include a product learning unit (310), a character learning unit (320), an image generation unit (330), and a feedback management unit (340).

[0031] When the generation service providing server (300) according to one embodiment of the present invention or another server (not shown) operating in conjunction with it transmits a best-selling product design generation service application, program, app page, web page, etc. using user image-based generative AI to at least one user terminal (100) and at least one planning terminal (400), at least one user terminal (100) and at least one planning terminal (400) can install or open the best-selling product design generation service application, program, app page, web page, etc. using user image-based generative AI. In addition, the service program may be driven on at least one user terminal (100) and at least one planning terminal (400) using a script executed in a web browser. Here, a web browser is a program that allows the use of web (WWW: World Wide Web) services, and refers to a program that receives and displays hypertext written in HTML (Hyper Text Mark-up Language), and includes, for example, Chrome, Microsoft Edge, Safari, Firefox, Whale, and UC Browser. In addition, an application refers to an application program on a terminal, and includes, for example, an app running on a mobile terminal (smartphone).

[0032] <Stage 1_Product RoRA Learning Stage>

[0033] The product learning unit (310) updates the user image of at least one product category based on the preference score for the user image within at least one product category, as illustrated in FIG. 3i, and trains a fine-tuning model with the updated user image. In one embodiment of the present invention, when each user takes a photo of a character product they like and uploads it, as illustrated in FIGS. 4d to 4h, the photo is provided to followers as a feed, where followers can then like or leave comments. Since this data is accumulated in the platform (Colley) of the present invention, the platform of the present invention can determine which user images are popular. The reason for identifying and utilizing only popular user images is to enable planners to design products based on popular user images when planning character products. If even unpopular user images are utilized in the training of the fine-tuning model, the generative AI will then use these unpopular user images (character product images) and reflect them in the generation process when generating images. Accordingly, the concept of a preference score is utilized to ensure that only user images that have received positive user responses, as shown in Figure 3i. The preference score can be calculated based on the number of likes and comments for the user image.

[0034] Here, the concept of the applicant's Korean Patent No. 10-2667500 (announced on May 22, 2024) can be further utilized. That is, by having AI predict popular product categories when a planner is planning, rather than having the planner manually specify product categories one by one and use generative AI to create generated images, the planner can extract popular product categories from a predictive model and then create generated images. For example, if it is predicted that character A and product B will be popular, the generative AI can be configured to generate generated images by specifying the categories of character A and product B.

[0035] Returning to fine-tuning, the fine-tuning model may be LoRA (Low-Rank Adaptation), which fine-tunes generative AI. In this case, the generative AI cannot be used as-is; in one embodiment of the present invention, it must be configured to generate the desired product design. This process is called fine-tuning. In other words, fine-tuning is used in a way that it updates itself to match downstream tasks through a retraining process. In other words, it involves providing additional external data to the generative AI. This means providing hundreds or thousands of additional training datasets tailored to the problem being solved, retraining the generative AI, and updating the model's parameters.

[0036] However, as generative AI has been growing increasingly large, the computational cost of updating the parameters of large-scale models has been increasing. Accordingly, methodologies such as Parameter Efficient Fine Tuning (PEFT) have emerged as efficient tuning methods. To this end, in one embodiment of the present invention, DreamBooth and LoRA, which can efficiently perform fine-tuning, are used so that the generative AI can perform product design according to one embodiment of the present invention. For example, setting up a large and complex machine learning model, such as that used in a large-scale language model (LLM) such as ChatGPT, requires considerable time and resources. Since these models have many parameters, this retraining process is costly and time-consuming, but LoRA provides a method for quickly adjusting the model without retraining. The configuration in which the generative AI is fine-tuned using LoRA will be described in detail later after explaining the generative AI, Stable Diffusion, in the image generation unit (330).

[0037] <Stage 2_Character LoRA Learning Stage>

[0038] The character learning unit (320) can collect character images of at least one character, as shown in FIG. 3j, and register at least one character to be added to the character list of the fine-tuning model. In order to add a new character using the above-described LoRA, the process is completed by collecting character images, as shown in FIG. 3j, having LoRA learn them, and registering the learned character in the character list.

[0039] <Step 3_Image Creation Step>

[0040] When a product category and character are selected from the planning terminal (400) as shown in FIG. 3k, the image generation unit (330) can receive a generated image including the character for the product from a pre-established generative AI (Generative Artificial Intelligence) and provide it to the planning terminal (400). That is, the image generation unit (330) can load LoRA corresponding to the category and product and connect it to Stable Diffusion to generate the generated image. The planning terminal (400) can select the product category and character to be generated and output a generated image in which the character is applied to the product category.

[0041] Generative AI can be a Stable Diffusion model, a text-to-image model that converts text into images. The technology for generating images from text is largely based on generative adversarial networks (GANs) and diffusion models, and diffusion models have recently significantly outperformed generative adversarial models. Representative image generation models based on diffusion models include OpenAI's Dalle-2 and Stable AI's Stable Diffusion, and these generative models can produce images that resemble actual human drawings. Furthermore, in addition to image generation, Inpaint, which modifies parts of the generated image, and Image-to-Image, which transforms an image using text as a guide, are also possible. When fine-tuning a trained diffusion model using new images, technologies such as Dreambooth or LoRA (Low-Rank Adaptation of Large Language Models), which can generate images similar to the style of the trained image using a small number of images, can be utilized.

[0042] Accordingly, in one embodiment of the present invention, a method can be used to create a product design by fine-tuning a pre-trained text-to-image model only by specifying a category and product by a planner.

[0043] Diffusion Model

[0044] The diffusion model is a probabilistic modeling method used in image generation. It models the probability distribution of an image and can generate new images based on this. The diffusion model offers the advantage of producing diverse results during image generation and can increase efficiency by reusing information learned during the image generation process. Diffusion model training is achieved through a two-component diffusion process. First, the forward diffusion process diffuses the image by adding noise to the pixel values ​​of the image at each step. Second, the backward diffusion process reversely diffusion the image to restore the original image from the noise. By repeating these two processes, the backward diffusion process can be used to generate new images from noise.

[0045] Diffusion models are trained using large datasets, and during the training process, the model is updated to minimize the difference between the trained image and the actual image through a backward diffusion process. The trained model is then used in the image generation process through a forward diffusion process, and noise is added to generate diverse images. Diffusion models have shown superior performance compared to adversarial generative neural networks in quantitative and qualitative evaluations in the field of image generation. Because they use probabilistic modeling methods for image generation, they can generate a wider variety of images than adversarial generative neural networks. Furthermore, through fine-tuning techniques such as LoRA, they can learn representations from images appearing in relatively small datasets to generate images.

[0046] Deep Learning-Based Text-to-Image

[0047] Text-to-Image technology refers to generating images when text is input into a text-to-image model. Because it can generate images from a description of the image desired by the planner, it is a technology with wide application in various fields. Text-to-Image technology can be divided into two parts: extracting features and analyzing correlations from text and image datasets, and generating images. The part that analyzes features and correlations between text and images uses BERT and VIT (Vision Transformer) as encoders to extract features and learn the relationship between text and images through contrastive learning. The image generation part is based on image generation models such as generative adversarial networks and diffusion models. Therefore, the advancement of text-to-image technology can improve its performance through advancements in models that extract text and image features, models that learn their relationships, and generative models that generate images.

[0048] Text-to-image technology was developed based on the DCGAN model, followed by AttnGAN, StackGAN, and DM-GAN, which introduced the attention mechanism. Recently, models such as Dalle-2 and Stable Diffusion, which use the diffusion model as a generative model, have been mainly used. These two models exhibit a similar structure, and among them, Stable Diffusion consists of a CLIP-based text encoder, an information generator consisting of UNet and Scheduler, and a decoder that generates images. Stable Diffusion generates images in latent space rather than pixel space, enabling fast image generation.

[0049] Fine-tuning

[0050] Text-to-Image Model Based on Diffusion Model

[0051] The Stable Diffusion model is a model that generates images from text based on a diffusion model. Stable Diffusion consists of the CLIP text encoder, UNet, and a Variational Autoencoder (VAE). CLIP (Contrastive Language-Image Pretraining) is a model that extracts text and image features and learns the relationship between the two sets of data through contrastive learning. Stable Diffusion uses CLIP's text encoder to extract text features. When a description of the image to be generated is input, CLIP's text encoder converts it into tokens and passes these tokens to UNet. UNet then removes noise generated from the tokens through a denoising process. Denoising is the process of removing noise from an image to generate an image. Finally, the image decoder uses VAE to convert the image's latent variables into pixels, ultimately generating the image. Stable Diffusion is a model that has been trained in advance on the LAION dataset, which consists of a large amount of text-image data, and can generate images of various shapes for the input text.

[0052] <Text-to-Image 모델의 미세조정>

[0053] When generating images using a diffusion model, fine-tuning to achieve the desired image direction is time-consuming. Therefore, DreamBooth, Textual Inversion, and LoRA can be used to shorten the time. In one embodiment of the present invention, a pre-trained diffusion model, the Stable Diffusion model, is trained on a dataset using DreamBooth and LoRA to generate product designs that are popular with users. DreamBooth is used to reflect characters to product categories. For example, even if there is a character called Mickey Mouse, it must be customized to fit the style of a keychain. Furthermore, if the model does not know Mickey Mouse itself, it cannot be applied. For this purpose, DreamBooth is utilized. Since DreamBooth is a large model, LoRA is used to efficiently perform fine-tuning.

[0054] LoRA (Low-Rank Adaptation of Large Language Models) is a low-rank adaptation technique for large-scale language models. This technique is designed to quickly and effectively fine-tune large-scale language models with a large number of parameters. Instead of retraining the entire layer, it uses fine-tuning data to update only a subset of parameters. Specifically, a learnable rank decomposition matrix is ​​inserted into a pre-trained model with fixed weights, and learnable parameters are inserted into each layer for fine-tuning. The fine-tuned model can then be utilized. Furthermore, LoRA can fine-tune the model based on target labels or target data, enabling designers to achieve desired output values. Due to these advantages, LoRA is used not only in language models but also in models such as image classification and text-to-image. In Stable Diffusion, LoRA can be applied to image representation and cross-attention layers for text input by designers.

[0055] The feedback management unit (340) can determine the number of likes and comments from at least one user terminal (100) for a generated image, then examine the preference score and store the preference score for the product category and character combination of the generated image. This is because even for the same category and character, the preference may vary depending on the design used. For example, even for the same character, the preference may vary depending on the design used. Accordingly, even if the generative AI creates a generated image based on a popular user image, the generative AI can reflect the user feedback by examining the user feedback again. Furthermore, this user feedback can also be used to retrain the predictive model of the applicant's previously registered patent described above.

[0056] Hereinafter, the operation process according to the configuration of the generation service providing server of the above-described FIG. 2 will be described in detail with reference to FIGS. 3 and 4 as examples. However, it will be apparent that the embodiment is only one of various embodiments of the present invention and is not limited thereto.

[0057] A platform according to an embodiment of the present invention was developed for the purpose as in Fig. 3a, and a technology as in Fig. 3b was used. The generative AI of the present invention is as in Fig. 3c, and a fine-tuning model as in Figs. 3d to 3g was used for fine-tuning. The back-end structure of the present invention is as in Fig. 3h, with the first step as in Fig. 3i, the second step as in Fig. 3j, and the third step as in Fig. 3k. In order to use the platform according to an embodiment of the present invention, after logging in as in Fig. 3l, on the main screen as in Fig. 3m, the planner selects the product category and character (Intellectual Property, IP) desired as in Fig. 3n, and presses the image generation button, and then generation begins and the result is shown as in Fig. 3o. In addition, the generated images can be filtered and selected by character or product category as in Fig. 3p.

[0058] A platform according to one embodiment of the present invention, as illustrated in FIG. 4a, provides an all-in-one solution for creating IP (character) products. This platform can alleviate the planner's concerns about what kind of product to create and what kind of character to select, and can encourage quick and accurate decision-making. In one embodiment of the present invention, as illustrated in FIG. 4b, users can upload and share their favorite character products, and other users can also leave responses. Using this, one embodiment of the present invention provides a solution for planning and creating IP products, as illustrated in FIG. 4c. FIGS. 4d to 4h show images of character products uploaded by users, i.e., user images.

[0059] Figure 4i shows a generated image created using generative AI. Clicking on any of these images reveals tags indicating the [Product Category] (e.g., figures) and the [Character] (e.g., Pikachu), as shown in Figure 4j. This allows for an intuitive understanding of the product category and character the image created using generative AI represents.

[0060] Matters not described in the method for providing a best-selling product design creation service using the user image-based generative AI of FIGS. 2 to 4 are the same as or can be easily inferred from the contents described in the method for providing a best-selling product design creation service using the user image-based generative AI of FIG. 1 above, and therefore, description thereof will be omitted below.

[0061] FIG. 5 is a diagram illustrating a process of transmitting and receiving data between each component included in a best-selling product design creation service provision system using the user image-based generative AI of FIG. 1 according to one embodiment of the present invention. Hereinafter, an example of the process of transmitting and receiving data between each component will be described with reference to FIG. 5. However, the present invention is not limited to this embodiment, and it will be apparent to those skilled in the art that the process of transmitting and receiving data illustrated in FIG. 5 may be modified according to various embodiments described above.

[0062] Referring to FIG. 5, the generation service providing server updates a user image of at least one product category based on a preference score for the user image within at least one product category, and trains a fine-tuning model with the updated user image (S5100).

[0063]

[0064] And, the generation service providing server collects the character image of at least one character and registers at least one character to be added to the character list of the fine-tuning model (S5200).

[0065] In addition, when a product category and character are selected from a planning terminal, the generation service providing server receives a generated image including the character in the product from a pre-built generative AI (Generative Artificial Intelligence) and provides it to the planning terminal (S5300).

[0066] The order of the above-described steps (S5100 to S5300) is merely an example and is not limited thereto. That is, the order of the above-described steps (S5100 to S5300) may be mutually changed, and some of the steps may be executed simultaneously or deleted.

[0067] Matters not described in the method for providing a best-selling product design creation service using a user image-based generative AI of FIG. 5 are the same as or can be easily inferred from the contents described in the method for providing a best-selling product design creation service using a user image-based generative AI through FIGS. 1 to 4, and therefore, description thereof will be omitted below.

[0068] The method for providing a best-selling product design creation service using user image-based generative AI according to one embodiment described through FIG. 5 may also be implemented in the form of a recording medium containing computer-executable instructions, such as an application or program module executed by a computer. The computer-readable medium may be any available medium that can be accessed by a computer, and includes both volatile and nonvolatile media, removable and non-removable media. In addition, the computer-readable medium may include all computer storage media. The computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented by any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data.

[0069] The method for providing a best-selling product design creation service using a user image-based generative AI according to an embodiment of the present invention described above can be executed by an application that is installed by default on a terminal (which may include a program included in a platform or operating system installed by default on the terminal), or can be executed by an application (i.e., a program) that the user directly installs on a master terminal through an application providing server such as an application store server, an application, or a web server related to the service. In this sense, the method for providing a best-selling product design creation service using a user image-based generative AI according to an embodiment of the present invention described above can be implemented by an application (i.e., a program) that is installed by default on a terminal or directly installed by a user, and can be recorded on a computer-readable recording medium such as a terminal.

[0070] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will readily appreciate that the present invention can be readily modified into other specific forms without altering the technical spirit or essential characteristics of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single entity may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined manner.

[0071] The scope of the present invention is indicated by the claims described below rather than the detailed description above, and all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts should be interpreted as being included in the scope of the present invention.

[0072] The mode for carrying out the invention has been described together with the best mode for carrying out the invention above.

[0073] The present invention has industrial applicability because it allows a generative AI to create an image so that a character is applied to a product category based on a fine-tuned model learned with a user image having a high preference score in the product category and a fine-tuned model learned with a character image when a product category and character are specified in a planning terminal, thereby creating a product design that is most likely to sell based on preference, while providing a visible draft rather than an idea state so that the product design does not undergo infinite modification.

Claims

1. A planning terminal that selects the category and character of the product to be created and outputs a created image in which the character is applied to the category of the product; and A generation service providing server including a product learning unit that updates a user image of at least one product category based on a preference score for the user image within the at least one product category and trains a fine-tuning model with the updated user image, a character learning unit that collects a character image of at least one character and registers the at least one character to be added to a character list of the fine-tuning model, and an image generation unit that receives a generated image including the character in the product from a pre-established generative artificial intelligence (Generative Artificial Intelligence) when a product category and character are selected in the planning terminal and provides the image to the planning terminal; A system that provides a best-selling product design creation service using user image-based generative AI including .

2. In paragraph 1, The above preference score is, A system for providing a best-selling product design generation service using a user image-based generative AI, characterized in that the design is calculated based on the number of likes and comments for the user image.

3. In paragraph 1, The above generative AI is, Stable Diffusion is a Text-To-Image model that converts text into images. The above fine-tuning model is, A best-selling product design generation service system using user image-based generative AI, characterized by LoRA (Low-Rank Adaptation) that fine-tunes the generative AI.

4. In paragraph 3, The above image generation unit, A best-selling product design generation service providing system using user image-based generative AI, characterized in that it loads LoRA corresponding to the above category and product and connects it to the above stable diffusion to generate the generated image.

5. In paragraph 1, A best-selling product design creation service system using user image-based generative AI, characterized by using DreamBooth to reflect the above character on the above product.

6. In paragraph 1, The above generation service providing server is, A feedback management unit that determines the number of likes and comments of at least one user terminal for the above-mentioned generated image, investigates the preference score, and stores the preference score for the combination of categories and characters of the product of the above-mentioned generated image; A best-selling product design creation service system using user image-based generative AI, characterized by including more.

Images