System and method for generating outdoor advertising contents based on generative AI models while satisfying predefined regulatory requirements and incorporating advertiser specifications

Abstract

The embodiments include: a step of obtaining an advertiser’s basic advertising data for an advertisement describing the basic content the advertiser wishes to reflect in the advertisement; and an advertiser’s advertising display condition indicating an area on a target vehicle where the advertiser will display the advertisement; a step of obtaining a vehicle image taken of the target vehicle, analyzing the vehicle image through pre-trained deep learning models to detect a window area of ​​the target vehicle, and detecting a reference area of ​​the target vehicle based on the detection result of the window area; a step of generating the outdoor advertisement for the target vehicle by applying the basic advertising data and the advertising display condition to a pre-trained advertisement generation model; a step of obtaining area information of the display area of ​​the outdoor advertisement according to the advertiser’s user input indicating the size or area of ​​the outdoor advertisement; a step of determining the display location of the outdoor advertisement according to the display area information according to the advertiser’s user input; and a step of verifying whether the outdoor advertisement having the obtained display area and the determined display location satisfies the guide. The present invention relates to a system and method for generating an outdoor advertisement that satisfies preset constraints that the outdoor advertisement must comply with and reflects the requirements of the advertiser based on a generative AI model, comprising the step of providing the advertiser with an outdoor advertisement for the target vehicle that satisfies the area constraint conditions of the guide.

Description

Technology Field

[0001] The embodiments of the present application relate to a technology that accurately satisfies the legal guidelines for outdoor advertising contents and generates outdoor advertising contents based on a generative AI model. Background Technology

[0002] Generally, buses are a means of public transportation that operates on fixed, fixed routes and is conveniently used by many people. Consequently, the exterior surface of a bus can be exposed to a large number of people across the wide geographical area it travels, making it recognized as a good advertising medium.

[0003] The application of advertisements to the exterior surface of buses is regulated by the Enforcement Decree of the Act on the Management of Outdoor Advertisements and the Promotion of the Outdoor Advertising Industry (abbreviated as: Enforcement Decree of the Outdoor Advertising Act). Under the current Enforcement Decree of the Outdoor Advertising Act, outdoor advertisements applicable to commercial vehicles such as city buses and intercity buses are described in Article 19, Paragraph 1 of the current Enforcement Decree of the Outdoor Advertising Act. According to the current Enforcement Decree of the Outdoor Advertising Act as of 2025 (amended on May 21, 2024), multiple legal restrictions must be satisfied, including i) that the advertisement must be displayed only on the vehicle body or the bus protruding number plate excluding the window portion (hereinafter, Condition i is referred to as the "location restriction condition"), and ii) that the display area must be within 50% of the total area of ​​each side (excluding the window portion) (hereinafter, Condition ii is referred to as the "area restriction condition").

[0004] Due to these legal restrictions, bus companies have established advertising areas in a very rigid and conservative manner and have continued to utilize them without change.

[0005] FIG. 1 illustrates a conventional advertising area where conventional bus outdoor advertisements are generally attached.

[0006] Currently, in the bus advertising field, when producing bus advertisements, the exact calculation of whether the area occupied by the bus advertisement actually reaches 50% is not done. Instead, from a conservative perspective considering legal conditions and based on the designer's experience or sense, a rectangular frame advertising area is designated as an area occupying approximately 30-40% of the total area excluding windows, as shown in Fig. 1, and bus advertisements are produced based on this advertising area.

[0007] Since bus advertising costs are proportional to the area of ​​the advertising space, a fixed advertising space resulted in a situation where advertising costs per area were also fixed.

[0008] As such, the limitations of the current bus advertising sector, which relies on the designer's experience or sensibility or is based solely on fixed advertising areas, cause various problems.

[0009] When bus outdoor advertisements are produced relying on a designer's experience or intuition rather than a fixed advertising area, problems arise where advertisers complain about the loss of space, and / or advertising designers face the inconvenience of being unable to persuade advertisers with accurate grounds regarding requests to increase the advertising area. For this reason, most bus outdoor advertising contracts are conducted based on a fixed advertising area.

[0010] However, if the advertising area is fixed, there are still other problems lurking, such as the advertiser being forced to compromise on the fixed advertising area even if the fixed advertising area is unsuitable for fully containing the advertising content desired by the advertiser or if the rectangular frame is unsuitable for displaying the advertising theme desired by the advertiser. Prior art literature

[0011] Patent Publication No. 10-2006-0122800 (Title of Invention: Advertising display on the upper exterior of a bus door and method thereof, published November 30, 2006) Utility Model Registration No. 20-0230356 (Title of Invention: Advertising screen rotation device, published July 4, 2001) Utility Model Registration No. 20-0434547 (Title of Invention: Rotating advertising device, published December 22, 2006) The problem to be solved

[0012] To solve the aforementioned problems, the embodiments of the present application aim to provide an apparatus and method for generating outdoor advertisements that reflect the requirements of an advertiser based on a generative AI model, while simultaneously complying with the legal guidelines for bus advertisements required by the Enforcement Decree of the Outdoor Advertising Act. means of solving the problem

[0013] A method for generating an outdoor advertisement that satisfies preset constraints that the outdoor advertisement must comply with according to one aspect of the present application and reflects the requirements of an advertiser based on a generative AI model is performed by a computing device. The method comprises: an advertiser’s basic advertising data for an advertisement describing basic content that the advertiser wishes to reflect in the advertisement; and an advertiser’s advertising display condition indicating an area on a target vehicle where the advertiser will display the advertisement; a step of acquiring a vehicle image of the target vehicle, analyzing the vehicle image through pre-trained deep learning models to detect a window area of ​​the target vehicle, and detecting a reference area of ​​the target vehicle based on the detection result of the window area, wherein the reference area of ​​the target vehicle is a part of the vehicle area of ​​the target vehicle for determining whether the advertiser’s outdoor advertisement satisfies the area constraint conditions of a preset guide; a step of generating the outdoor advertisement for the target vehicle by applying the advertising basic data and the advertising display condition to a pre-trained advertisement generation model; and a step of acquiring area information of the display area of ​​the outdoor advertisement according to user input of the advertiser indicating the size or area of ​​the outdoor advertisement. The method may include: a step of determining the display location of an outdoor advertisement based on the display area information according to the user input of the advertiser; a step of verifying whether the outdoor advertisement having the acquired display area and the determined display location satisfies the guide; and a step of providing the outdoor advertisement for the target vehicle that satisfies the area limit condition of the guide to the advertiser.The above-mentioned advertisement generation model is a generative AI model pre-trained to generate an image that visually implements the information of the input data from the input data, and is trained to receive basic data or prompt data in the form of text as the input data, encode the input data into a vector of a latent space, and sequentially generate and combine visual features based on the encoded vector to generate an image that reflects the intent and constraints of the input data as output data, wherein the latent space refers to a multidimensional feature space in which the semantic features of the input data are quantified and expressed. The step of verifying whether the outdoor advertisement having the acquired display area and the determined display location satisfies the guide; is a step of verifying whether the outdoor advertisement according to the display area information satisfies the area limit conditions of the guide, comprising: a step of calculating the area of ​​the reference area in the target vehicle; a step of calculating the ratio of the area of ​​the acquired outdoor advertisement to the ratio of the calculated area; and a step of analyzing whether the ratio of the area of ​​the outdoor advertisement to the ratio of the area exceeds a threshold ratio corresponding to a preset area limit condition. The method may include: a step of determining that the outdoor advertisement for the target vehicle does not satisfy the area restriction condition if the ratio of the area of ​​the outdoor advertisement to the ratio of the above area exceeds a threshold ratio corresponding to a preset area restriction condition; and a step of determining that the advertisement for the target vehicle satisfies the area restriction condition if the ratio of the area of ​​the outdoor advertisement to the ratio of the above area has an area equal to or less than the threshold ratio corresponding to the preset area restriction condition. The step of determining whether the area ratio with respect to the area of ​​the reference area is equal to or less than the threshold ratio corresponding to the preset area restriction condition is to determine whether the following mathematical formula is satisfied.

[0014] [Mathematical Formula]

[0015]

[0016] Here, Rthr may include a preset threshold ratio, A1 representing the area of ​​the reference area, and A2 representing the area of ​​the outdoor advertisement for the target bus.

[0017] In one embodiment, the step of calculating a reference area of ​​the target vehicle from the target vehicle may include: inputting the vehicle image into a pre-trained vehicle recognition model to recognize the target vehicle in the vehicle image; inputting a first sub-image including the recognized target vehicle into a pre-trained vehicle area detection model to detect a vehicle area in which the recognized target vehicle is represented in the selected vehicle image; inputting a second sub-image representing the vehicle area within the selected vehicle image into a pre-trained window area detection model to detect a window area of ​​the target vehicle located on the vehicle surface in a direction satisfying the position constraint condition; and determining the remaining area excluding the window area from the detected vehicle area within the vehicle image as a reference area for determining the area constraint condition based on the detection result for the window area of ​​the target vehicle. The window area detection model may include an ROI detection unit that extracts shape features from an input image and identifies the location of a pre-set part of interest of the vehicle based on the extracted shape features; a color feature extraction unit that extracts color features pixel by pixel from the input image; and a window detection unit configured to identify a window area in the vehicle area based on the location of the identified part of interest.

[0018] In one embodiment, the input data for the window detection unit is a combination of the location of an identified part of interest and the result of extracting color features, wherein the location of the identified part of interest corresponds to an image marked on a color distribution map, and the color distribution map is a distribution map representing the distribution of each color based on the shape of the bus exterior area, and the window detection unit may include a pre-trained artificial neural network model to infer a first correlation between the location of the part of interest and the location of the window area based on the location of the part of interest, and to infer a second correlation between the color distribution map and the location of the window area, and to calculate the location of the window area in the bus exterior area corresponding to the input data based on the inferred first correlation and second correlation. The window area detection model is trained to minimize the following loss function, and

[0019] [Mathematical Formula]

[0020]

[0021] Here, L represents the loss function of the second region detection model, and EROI position and Ewindow position represent the predicted position of the region of interest and the predicted position of the window region, respectively, calculated by the ROI detection unit and the window detection unit from the same sample training image, and TROI position and Twindow position represent the actual position of the region of interest and the actual position of the window region, respectively, from the same sample training image.

[0022] In one embodiment, the advertising basic data includes one or more of advertising theme information, source image, advertising text information, and advertising object information, and the advertising theme information includes an advertising material type, and the advertising material type may be an integrated type that displays a single advertising material on a display surface or a distributed type in which two or more advertising materials are displayed dispersedly on the display surface.

[0023] In one embodiment, the step of calculating the ratio of the area of ​​the obtained outdoor advertisement to the ratio of the calculated area may include, when the outdoor advertisement is of the dispersed type, the step of summing the area of ​​the outdoor advertisement on the display surface of the outdoor advertisement; and the step of calculating the ratio of the summed area of ​​the outdoor advertisement to the ratio of the calculated area.

[0024] In one embodiment, the advertisement generating model may include a prompt layer that converts input data of the advertisement generating model into prompt data according to a prompt format—the prompt layer is configured to understand the context and meaning of the input data and convert it into a prompt format based on a transformer structure—and an image generating layer configured to receive prompt data reflecting basic advertisement content and advertiser display conditions to generate the outdoor advertisement, and to generate an image reflecting the advertiser's intentions, requirements, and characteristics of the vehicle advertisement.

[0025] In one embodiment, the step of verifying whether an outdoor advertisement having the acquired display area and the determined display location satisfies the guide; may include the step of verifying whether the outdoor advertisement according to the display location satisfies the area limitation condition of the guide, comprising: identifying an area where the display of the outdoor advertisement is restricted on the target vehicle; analyzing whether the determined display location of the outdoor advertisement overlaps with the identified limitation area; determining that the outdoor advertisement for the target vehicle does not satisfy the location limitation condition if the display location of the outdoor advertisement overlaps with the limitation area; readjusting the display location of the outdoor advertisement so that the outdoor advertisement for the target vehicle satisfies the location limitation condition; and determining that the outdoor advertisement for the target vehicle satisfies the location limitation condition if the display location of the outdoor advertisement does not overlap with the limitation area. The limitation area may be the window area.

[0026] In one embodiment, the step of obtaining area information for the outdoor advertisement may include: providing initial area information according to a preset area rule to an advertiser; and determining area information according to user input entered thereafter. The preset area rule may be the maximum allowable area capable of satisfying the area limit conditions of the preset guide.

[0027] In one embodiment, the threshold ratio may be 0.5.

[0028] A computer-readable recording medium according to another aspect of the present application may record a program that enables a method of generating an outdoor advertisement that satisfies the preset constraints that the outdoor advertisement according to the embodiments described above must comply with and that generates an outdoor advertisement reflecting the advertiser's requirements based on a generative AI model.

[0029] A system for generating an outdoor advertisement that satisfies preset constraints that an outdoor advertisement must comply with according to another aspect of the present application and reflects the requirements of an advertiser based on a generative AI model comprises a processor and memory. The system comprises: a reference area analysis unit that acquires a vehicle image of the target vehicle, analyzes the vehicle image through pre-trained deep learning models to detect a window area of ​​the target vehicle, and detects a reference area of ​​the target vehicle based on the detection result of the window area; wherein the reference area of ​​the target vehicle is a part of the vehicle area of ​​the target vehicle for determining whether the advertiser's outdoor advertisement satisfies the area constraint conditions of a preset guide; It may include: basic advertising data of an advertiser describing the basic content that the advertiser wishes to reflect in the advertisement; an advertising generation unit that generates the outdoor advertisement for the target vehicle by applying the basic advertising data and the advertising display conditions to a pre-trained advertising generation model when the advertiser obtains the advertising display conditions indicating the area where the advertiser will display the advertisement on the target vehicle; area information of the display area of ​​the outdoor advertisement obtained according to the advertiser's user input indicating the size or area of ​​the outdoor advertisement; and a guide verification unit that verifies whether the outdoor advertisement has the display location of the outdoor advertisement determined according to the advertiser's user input satisfies the guide.The above-mentioned advertisement generation model is a generative AI model pre-trained to generate an image that visually embodies the information of the input data from the input data. It receives basic data or prompt data in text form as input data, encodes the input data into a vector in a latent space, and sequentially generates and combines visual features based on the encoded vector to generate an image as output data that reflects the intent and constraints of the input data. Here, the latent space refers to a multidimensional feature space in which the semantic features of the input data are quantified and expressed. The guide verification unit may be configured to verify whether an outdoor advertisement having the acquired display area and the determined display location satisfies the guide, and to verify whether an outdoor advertisement according to the display area information satisfies the area limit condition of the guide by calculating the area of ​​the reference area in the target vehicle, calculating the ratio of the area of ​​the acquired outdoor advertisement to the ratio of the calculated area, analyzing whether the ratio of the area of ​​the outdoor advertisement to the ratio of the area exceeds a threshold ratio corresponding to a preset area limit condition, and determining that if the ratio of the area of ​​the outdoor advertisement to the ratio of the area exceeds the threshold ratio corresponding to the preset area limit condition, the outdoor advertisement for the target vehicle does not satisfy the area limit condition, and determining that the advertisement for the target vehicle has an area equal to or less than the threshold ratio corresponding to the preset area limit condition, the ratio of the area of ​​the outdoor advertisement to the ratio of the area is satisfied with the area limit condition. The step of determining whether the area ratio with respect to the area of ​​the reference area is equal to or less than the threshold ratio corresponding to the preset area limit condition is to determine whether the following mathematical formula is satisfied.

[0030] [Mathematical Formula]

[0031]

[0032] Here, Rthr may include a preset threshold ratio, A1 representing the area of ​​the reference area, and A2 representing the area of ​​the outdoor advertisement for the target bus. Effects of the invention

[0033] An outdoor advertising device according to one aspect of the present application can generate various outdoor advertisements based on a generative AI model, while not being limited to a fixed area or frame shape.

[0034] As a result, it is possible to satisfy advertisers who wish to maximize the display area of ​​outdoor advertisements to maximize advertising effectiveness. Furthermore, advertising designers can quantitatively provide advertisers making excessive demands for advertising expansion that go beyond legal guidelines with the rationale required to create the aforementioned bus outdoor advertisements (e.g., the maximum value of the necessary outdoor advertisement display area), thereby enabling them to persuade advertisers more easily.

[0035] In addition, it is possible to generate outdoor advertisements that have a frame shape and area suitable for fully containing the advertisement content desired by the advertiser, or have a frame shape suitable for the advertisement theme desired by the advertiser, while simultaneously accurately satisfying legal guidelines.

[0036] The effects of the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from the description in the claims. Brief explanation of the drawing

[0037] FIG. 1 illustrates a conventional advertising area where conventional bus outdoor advertisements are generally attached. FIG. 2 is a schematic diagram of a neural network according to various embodiments of the present application. FIG. 3 is a schematic diagram of a system that generates outdoor advertisements reflecting the requirements of advertisers based on a generative AI model and complies with a preset guide, according to one aspect of the present application. FIG. 4 is a detailed flowchart of a process for detecting a reference area of ​​a target vehicle in a vehicle image according to various embodiments of the present application. FIG. 5 is a schematic diagram of the internal architecture of a window area detection model according to various embodiments of the present application. Figure 6 is a detailed flowchart of the operation of the window area detection model of Figure 5. FIG. 7 illustrates a vehicle standard area used to verify whether an outdoor advertisement satisfies legal guidelines according to various embodiments of the present application. FIG. 8 is a schematic diagram of the internal architecture of a generative AI model according to various embodiments of the present application. FIG. 9 is a flowchart of a method for generating an outdoor advertisement that reflects the requirements of an advertiser based on a generative AI model and complies with a preset guide, according to another aspect of the present application. FIG. 10 is a detailed flowchart of the process for determining whether an outdoor advertisement on a target vehicle satisfies the area limit conditions of the legal guide according to various embodiments of the present application. FIG. 11 illustrates an outdoor advertisement generated from advertising basic data according to one example of the present application. FIG. 12 illustrates an outdoor advertisement generated from advertising basic data according to another example of the present application. FIG. 13 is a schematic diagram of a vehicle including a roller screen on which an outdoor advertisement is installed, according to various embodiments of the present application. FIG. 14 illustrates an exemplary computer system capable of implementing the system (100) of FIG. 2. Specific details for implementing the invention

[0038] Hereinafter, embodiments and examples of the present invention are described in detail with reference to the attached drawings so that those skilled in the art to which this application pertains can easily implement them. However, the present invention may be implemented in various different forms and is not limited to the embodiments and examples described herein.

[0039] Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be modified and implemented in various forms. Accordingly, the embodiments are not limited to the specific disclosed forms, and the scope of this specification should be understood to include various modifications, equivalents, and / or alternatives that fall within the technical concept.

[0040] Terms such as "first" or "second" may be used to describe various components, but these terms should be interpreted solely for the purpose of distinguishing one component from another. For example, the first component may be named the second component, and similarly, the second component may be named the first component. Expressions such as "first," "second," "first," or "second" used in various embodiments may modify various components, regardless of order and / or importance, and do not limit said components.

[0041] A singular expression includes a plural expression unless the context clearly indicates otherwise.

[0042] When it is stated that a component is "connected" to another component, it should be understood that it may be directly connected to or coupled with that other component, or that there may be other components in between.

[0043] As used herein, the expression “configured to” may be replaced, depending on the context, with, for example, “suitable for,” “having the capacity to,” “designed to,” “adapted to,” “made to,” or “capable of.” The term “configured to” may not necessarily mean only that which is “specifically designed to” in hardware. Instead, in some situations, the expression “device configured to” may mean that the device is “capable of” in conjunction with other devices or components. For example, the phrase “processor configured to perform A, B, and C” may mean a dedicated processor for performing the said operations (e.g., an embedded processor), or a generic-purpose processor (e.g., a CPU or an application processor) capable of performing said operations by executing one or more software programs stored in a memory device.

[0044] In this specification, terms such as “comprising” or “having” are intended to specify the existence of the described features, numbers, steps, actions, components, parts, or combinations thereof, and should not be understood as precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0045] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which this application pertains. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined herein.

[0047] Explanation of Terms

[0048] In this specification, "vehicle" refers to a means of transportation equipped with wheels and a vehicle body. The vehicle is a means of transportation subject to restrictions under the Enforcement Decree of the Act on the Management of Outdoor Advertisements and the Promotion of the Outdoor Advertising Industry (abbreviated as: Enforcement Decree of the Outdoor Advertising Act) when advertisements are attached to its exterior surface, and may be an automobile under the Automobile Management Act, a dump truck under the Construction Machinery Management Act, a commercial vehicle under the Passenger Transport Business Act, or a commercial cargo vehicle under the Cargo Transport Business Act.

[0049] In this specification, a bus is a means of transportation subject to the restrictions of the Enforcement Decree of the Act on the Management of Outdoor Advertisements and the Promotion of the Outdoor Advertising Industry (abbreviated as: Enforcement Decree of the Outdoor Advertising Act) when attaching advertisements to its exterior surface, and may be an intercity bus, a city bus, or other bus.

[0050] In this specification, the guide is a legal guide that outdoor advertisements attached to vehicles must comply with, and may be a restriction condition under the Enforcement Decree of the Outdoor Advertising Act or other laws regulating outdoor advertisements attached to vehicles.

[0051] Legal restrictions regarding outdoor advertisements on buses follow the restrictions specified for buses in the Enforcement Decree of the Outdoor Advertising Act. However, the present invention is not limited to the condition values ​​of the examples described above, and may follow corresponding restrictions in accordance with changes to the legal restrictions. Such changes to legal restrictions should not be interpreted as limiting the scope of technical protection of the present invention.

[0052] For example, prior to May 21, 2024, the legal restrictions under the Enforcement Decree of the Outdoor Advertising Act are as follows: i) it must be displayed on the side or rear of the vehicle body excluding the window area, or on the bus protruding number plate (referring to a number plate installed attached to the bus door so that it protrudes when the door is opened); ii) the display area must be within half of the area of ​​each side (excluding the window area).

[0053] On the other hand, as of 2025, the filing date of this application, the legal restriction conditions under the current Enforcement Decree on Outdoor Advertising have been changed as described above, with the location restriction conditions extended to the entire remaining part of the vehicle body excluding the window area from the side, rear, or bus protruding number plate of the vehicle body. As can be seen from this example of changing legal restriction conditions, the values ​​of the location restriction conditions and the area restriction conditions are examples showing that the present invention can flexibly respond to changes in legal restriction conditions, and this implies that the present invention can be implemented and applied with the changed values ​​of various legal restriction conditions that may arise in the future.

[0055] System Architecture

[0056] A system and method for generating vehicle outdoor advertisements according to embodiments of the present application are configured to use one or more artificial intelligence (AI) models. An artificial intelligence (AI) model refers to a computational model capable of learning specific patterns from input data and deriving new results. A neural network is a representative implementation method of such an artificial intelligence model. In the system and method for generating vehicle outdoor advertisements, at least one artificial intelligence model may be implemented as a neural network.

[0057] FIG. 2 is a schematic diagram of a neural network according to various embodiments of the present application.

[0058] A neural network (or referred to as an artificial neural network) can be composed of a set of interconnected computational units, which can generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network is composed of at least one node. The nodes (or neurons) constituting neural networks may be interconnected by one or more links. Within a neural network, one or more nodes connected via links may form a relative relationship between an input node and an output node. The concepts of input and output nodes are relative; any node in an output node relationship with respect to one node may be in an input node relationship with respect to another node, and vice versa. As described above, the input node versus output node relationship can be generated around links. One or more output nodes may be connected to a single input node via links, and vice versa.

[0059] In a relationship between an input node and an output node connected through a single link, the value of the output node's data can be determined based on the data input to the input node. Here, the link interconnecting the input node and the output node may have a weight. The said weight may be a conditional weight. According to embodiments, the link may be configured so that data moves differently to the node based on specific conditions through a conditional link. Such conditions are defined by a user or an algorithm and enable the neural network to perform the desired function. For example, if one or more input nodes are connected to a single output node via respective links, the output node may determine an output value based on the input values ​​of the connected input nodes and the conditions set on each link.

[0060] As described above, a neural network consists of one or more nodes interconnected through one or more links, forming input-output node relationships within the neural network. The characteristics of a neural network can be determined by the number of nodes and links within the network, the relationships between the nodes and links, and the values ​​of the conditions assigned to each link. For example, if two neural networks exist with the same number of nodes and links but different link weight values, the two neural networks can be recognized as being different from each other.

[0061] A neural network can be composed of a set of one or more nodes. A subset of nodes constituting a neural network can form a layer. Some of the nodes constituting a neural network can form a layer based on their distances from an initial input node. For example, a set of nodes with a distance of n from an initial input node can form n layers. The distance from the initial input node can be defined by the minimum number of links that must be traversed to reach that node from the initial input node. However, this definition of a layer is arbitrary for illustrative purposes, and the degree of a layer within a neural network can be defined in a way different from that described above. For example, a layer of nodes may be defined by its distance from a final output node.

[0062] Initial input nodes may refer to one or more nodes within a neural network to which data is directly input without passing through links in relation to other nodes. Alternatively, in terms of link-based relationships within the neural network, they may refer to nodes that do not have other input nodes connected by links. Similarly, final output nodes may refer to one or more nodes within a neural network that do not have output nodes in relation to other nodes. Furthermore, hidden nodes may refer to nodes constituting the neural network that are neither initial input nodes nor final output nodes.

[0063] A neural network according to one embodiment of the present disclosure may have the number of nodes in the input layer equal to the number of nodes in the output layer, and may be a neural network in which the number of nodes decreases and then increases again as it progresses from the input layer to the hidden layer. Additionally, a neural network according to another embodiment of the present disclosure may have the number of nodes in the input layer less than the number of nodes in the output layer, and may be a neural network in which the number of nodes decreases as it progresses from the input layer to the hidden layer. Additionally, a neural network according to yet another embodiment of the present disclosure may have the number of nodes in the input layer greater than the number of nodes in the output layer, and may be a neural network in which the number of nodes increases as it progresses from the input layer to the hidden layer. A neural network according to yet another embodiment of the present disclosure may be a neural network in which the above-described neural networks are combined.

[0064] A deep neural network (DNN) may refer to a neural network that includes multiple hidden layers in addition to input and output layers. Using a deep neural network allows for the identification of the latent structures of data. That is, it is possible to identify the latent structures of photos, text, videos, voice, and music (e.g., what objects are present in a photo, what the content and emotions of a text are, what the content and emotions of a voice are, etc.). Deep neural networks may include convolutional neural networks (CNN), recurrent neural networks (RNN), autoencoders, Generative Adversarial Networks (GAN), restricted Boltzmann machines (RBM), deep belief networks (DBN), Q networks, U networks, Siamese networks, Generative Adversarial Networks (GAN), etc. The description of deep neural networks described above is merely illustrative and the present disclosure is not limited thereto.

[0065] A neural network may be trained in at least one of supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning. Training of a neural network may be a process of applying knowledge to the neural network to perform a specific action. Depending on the embodiments, the neural network may recalculate and re-respond to itself to minimize errors in the output.

[0066] In the training of neural networks, training data is generally a subset of real-world data (i.e., the data intended to be processed using the trained neural network); therefore, a training cycle may exist where errors decrease on the training data but increase on real-world data. Overfitting is a phenomenon in which errors on real-world data increase due to excessive training on the training data. For example, a neural network trained on cats by showing it yellow cats may fail to recognize cats other than yellow ones as cats, which can be a type of overfitting.

[0067] Overfitting can cause an increase in errors in machine learning algorithms. Various optimization methods can be used to prevent such overfitting. To prevent overfitting, methods such as increasing the training data, regularization, dropout (which disables some nodes in the network during training), and the use of batch normalization layers can be applied.

[0069] Generative AI-based outdoor advertising generation

[0070] FIG. 3 is a schematic diagram of a system that generates outdoor advertisements reflecting the requirements of advertisers based on a generative AI model and complies with a preset guide, according to one aspect of the present application.

[0071] Referring to FIG. 3, a system that generates an outdoor advertisement reflecting the advertiser's requirements based on a generative AI model while complying with the above-mentioned pre-set guide (hereinafter, outdoor advertisement generation system) may include a control unit (110), a standard area analysis unit (130), an advertisement generation unit (150), and a guide verification unit (170).

[0072] The control unit (110) performs overall operations to generate outdoor advertisements that reflect the advertiser's requirements based on a generative AI model and in compliance with a preset guide, or controls the operations of other components (130, 150, 170) of the system.

[0073] The above control unit (110) can provide various user interfaces (UI) for generating outdoor advertisements that reflect the advertiser's requirements based on a generative AI model and comply with a preset guide.

[0074] In one example, the control unit (110) may provide basic advertising data of an advertiser for an advertisement that describes the basic content that the advertiser wants to reflect in the advertisement; and a UI screen that induces user input to obtain the advertiser's advertising display conditions, which indicate the area where the advertiser will display the advertisement in a target vehicle.

[0075] The above control unit (110) can provide a UI screen capable of performing editing operations, such as reducing the area of ​​an outdoor advertisement that does not satisfy the legal guide so that it satisfies the legal guide.

[0076] The above control unit (110) can provide an outdoor advertisement that satisfies the legal guide, and / or a UI screen showing the result of applying the outdoor advertisement to a target vehicle.

[0077] The control unit (110) can search for various materials (or information) to generate outdoor advertisements that reflect the advertiser's requirements based on a generative AI model and comply with a preset guide. The control unit (110) can obtain necessary materials (or information) by accessing a data store, an external web server, an internet search engine, or other external information sources that are pre-stored in the system (100).

[0078] A data store may be used to store various types of information. In some embodiments, information stored in a data store may be structured according to a specific data structure. Additionally, in some embodiments, each data store may be a relational, columnar, correlated, or other suitable database. Although this specification describes or illustrates a specific type of database, this specification considers any suitable type of database. In some embodiments, a server may provide an interface that enables the management, retrieval, modification, addition, or deletion of information stored in a data store.

[0079] The control unit (110) may obtain vehicle-related information stored in the internal data store of the system (100) or from an external source, and this may include vehicle profile information, vehicle body information, glass area information, etc.

[0080] The above vehicle profile information may include the vehicle's product name, product code, category, and / or other information. The above body information is information describing the vehicle's body and may include the body design, body area, etc. The above glass area information may include the location of the glass area in the vehicle, the area of ​​the glass area, etc.

[0081] The operation of the above control unit (110) and the information utilized accordingly will be described in more detail with reference to Fig. 9 below.

[0083] The standard area analysis unit (130) calculates a standard area used to determine whether the outdoor advertisement on the target vehicle satisfies the area limit conditions of the legal guide.

[0084] The reference area is a part of the vehicle for determining whether the legal restriction conditions, in particular the area restriction conditions, are satisfied. The reference area may be the remaining area of ​​the vehicle body excluding the windows. The reference area is represented as the remaining image area in the vehicle image, excluding the window area representing the windows from the vehicle area representing the vehicle body. In some embodiments, the reference area is an area located where the position restriction conditions are satisfied on the vehicle. For example, the reference area may be an area located on the left side, right side, or rear of the target vehicle.

[0085] The reference area analysis unit (130) can calculate a reference area in the target vehicle using the retrieved vehicle-related data and / or data provided by the advertiser. Calculation of the reference area may involve identifying the location, area, and / or boundary of the reference area in the vehicle body.

[0086] In specific embodiments, the reference area analysis unit (130) can calculate a reference area by analyzing a vehicle image of a target vehicle. The system (100) can receive one or more vehicle images as data for calculating the reference area.

[0088] FIG. 4 is a detailed flowchart of a process for detecting a reference area of ​​a target vehicle in a vehicle image according to various embodiments of the present application.

[0089] The reference area analysis unit (130) above may be configured to perform the operation (S201) of acquiring a target vehicle image, which is a target vehicle to which an outdoor advertisement is to be applied, in order to detect a reference area of ​​the target vehicle in the vehicle image.

[0090] The above-mentioned target vehicle has a vehicle surface expressed in four surface directions: front, rear, left side, and right side. The reference area analysis unit (130) may receive or input an image set containing one or more of the front, rear, left side, and right side of the vehicle as a target vehicle image. In some embodiments, individual images within the image set are images of the target vehicle captured on the front, rear, left side, or right side of the target vehicle, respectively. The images of the target vehicle represent the exterior surface of the target vehicle in the captured surface direction in the form of an image.

[0091] The above system (100) may acquire image information of a target vehicle that describes the image of the target vehicle along with an image of the target vehicle (e.g., by a control unit (110) or a reference area analysis unit (130)). The image information includes surface direction information that describes the surface direction of the target vehicle shown in the image. Additionally, in some embodiments, the image information further includes vehicle type information that describes the type of vehicle to which the target vehicle belongs. The types of vehicles may include high-speed vehicles, intercity vehicles, city vehicles, airport vehicles, and tourist vehicles.

[0092] In some embodiments, the image of each target vehicle in the image set may be captured through a shooting guide that obtains an image of the target vehicle aligned with each surface orientation so that an advertisement created for the target vehicle can more accurately satisfy legal restrictions.

[0093] The above system (100) can display the shooting guide and provide a UI screen capable of shooting a subject (e.g., by the control unit (110) or the reference area analysis unit (130)).

[0094] The above-mentioned shooting guide may be a front shooting guide, a rear shooting guide, a left-side shooting guide, or a right-side shooting guide. The shooting guide for each vehicle surface direction is configured to support the alignment of the position between the imaging element capturing the image and the subject vehicle based on a preset calibration portion for the vehicle surface in that direction. The above-mentioned shooting guide may be configured to provide a position pointer indicating the shooting position for the preset calibration portion. When a user captures an image of the vehicle with the corresponding part of the subject vehicle positioned at the position pointer, the vehicle image in that surface direction is treated as an image suitable for generating outdoor advertisements for the target vehicle that more accurately satisfies legal restrictions.

[0095] In one example, the front shooting guide and rear shooting guide have calibration parts for aligning the front and rear license plates of the vehicle, and are configured to provide a position pointer indicating the shooting position for the license plate of the subject vehicle. The left side shooting guide and right side shooting guide have calibration parts for aligning the left side and right side wheels of the vehicle, and are configured to provide a position pointer indicating the shooting position for the wheels of the subject vehicle. The image of the vehicle captured in this way may be an image that accurately captures the front, rear, left side, and right side.

[0096] Additionally, the reference area analysis unit (130) includes: an operation (S230) of inputting a vehicle image into a pre-trained vehicle recognition model to recognize the target vehicle in the vehicle image; an operation (S250) of inputting a first sub-image including the recognized target vehicle into a pre-trained vehicle area detection model to detect a vehicle area in which the recognized target vehicle is represented in a selected vehicle image; an operation (S270) of inputting a second sub-image representing the vehicle area within the selected vehicle image into a pre-trained window area detection model to detect a window area of ​​the target vehicle located on a vehicle surface in a direction satisfying the position restriction condition; and an operation (S290) of determining the remaining area excluding the window area from the detected vehicle area within the vehicle image as a reference area for determining the area restriction condition based on the detection result for the window area of ​​the target vehicle. In some embodiments, prior to the operation (S230), the reference area analysis unit (130) includes an operation (S210) of selecting a vehicle image having a direction satisfying the position restriction condition of a pre-set legal guide based on surface direction information within the image information of the target vehicle. It can be further configured to perform.

[0097] The above location restriction conditions may indicate vehicle surface locations that are not permitted under the Enforcement Decree of the Outdoor Advertising Act. As previously mentioned, the above location restriction conditions refer to the side or rear of the vehicle body, excluding the window area, or the area of ​​the vehicle protruding license plate (referring to a license plate installed attached to the vehicle's door so that it protrudes when the door is opened). According to current laws, any vehicle surface in any direction may be a surface that satisfies the location restriction conditions.

[0098] The reference area analysis unit (130) can determine whether each vehicle image in the image set corresponds to a vehicle image in which a surface direction satisfies a preset position constraint condition for a vehicle outdoor advertisement is displayed.

[0099] In the above step (S230), the vehicle recognition model is a vision recognition model configured to recognize whether an object in an input image corresponds to a vehicle by simultaneously identifying the location of at least one object within the input image and the category of the object. The vehicle recognition model is a machine learning model having model parameters learned to recognize a vehicle in an input image.

[0100] The vehicle recognition model described above has various deep learning structures configured to extract shape features from input images and recognize objects based on the results of the shape feature extraction. The vehicle recognition model may, for example, have a CNN (Convolutional Neural Networks) based structure (e.g., R-CNN (Regions with CNN), Fast R-CNN, Faster R-CNN, YOLO (You Only Look Once), SSD (Single Shot Multibox Detextor)). The structure of the vehicle recognition model described above is merely exemplary and is not limited thereto.

[0101] The vehicle recognition result output from the above vehicle recognition model may be implemented in a form that includes the vehicle shape within the input image. The form may, for example, be a bounding box, but is not limited thereto and may be implemented in other forms. The vehicle recognition result indicates whether the object is a vehicle and the location of the recognized vehicle in the input image.

[0102] The above step (S250) includes: a step of generating a sub-image (hereinafter, "first sub-image") containing the vehicle shape from an image of a selected target vehicle based on the recognition result of the above step (S230) as a first input image for a vehicle area detection model; and a step of inputting the first sub-image containing the recognized target vehicle into a pre-trained vehicle area detection model to identify a vehicle area in which the recognized target vehicle is expressed in the selected vehicle image.

[0103] In the above step (S230), a sub-image (hereinafter, "first sub-image") including the vehicle shape can be formed from the image of the target vehicle selected in the above step (S210) (S250).

[0104] The above vehicle region detection model is a deep learning model that is pre-trained to analyze an input image pixel by pixel, identify pixels corresponding to a vehicle, and calculate a vehicle region in which the target vehicle is represented (S250).

[0105] The above vehicle region is an image area showing the appearance of various vehicle components, such as the vehicle body, wheels, and windows. In other words, the above vehicle region detection model can be referred to as a vehicle region detection model.

[0106] In various embodiments of the present application, at least a portion of the vehicle region detection model may be implemented as a segmentation model configured to identify a region in which a vehicle is represented in an input image. The segmentation model is configured to analyze the input image pixel by pixel, classify the corresponding pixels, and identify an object region in which a desired object is represented. The segmentation model may include a segmentation model based on a semantic segmentation method or a segmentation model based on an instance segmentation method.

[0107] The above partitioning model may include, for example, FCN (Fully Convolutional Network), U-Net, DeepLab, Mask R-CNN, but is not limited thereto and may include other deep learning structures for partitioning.

[0108] The selected vehicle image represents the vehicle surface in the direction satisfying the position limit condition. Accordingly, the reference area analysis unit (130) can divide the vehicle area represented on the vehicle surface in the direction satisfying the position limit condition (S250).

[0109] The reference area analysis unit (130) may be configured to detect a window area of ​​the target vehicle located on a vehicle surface in a direction satisfying the position limit condition by generating a sub-image (hereinafter, "second sub-image") representing the vehicle area based on the identification result of the vehicle area in operation (S250) as a second input image for a window area detection model; and input the second sub-image representing the vehicle area into a pre-trained window area detection model to detect the second area of ​​the target vehicle, i.e., the window area, in the vehicle area of ​​the target vehicle.

[0110] The window area of ​​the target vehicle may be a sub-area located on a vehicle area corresponding to the vehicle surface in the direction satisfying the position limit condition. If the target vehicle includes a plurality of windows, some windows and other windows may be installed discontinuously from each other. Then, the window area identified in step (S270) includes the area occupied by the windows, excluding the discontinuous area between the windows.

[0111] In some embodiments, the second sub-image may be an image composed of vehicle regions divided in step (S250). The second sub-image may represent an object of a vehicle region extracted along the boundary of the divided vehicle region.

[0112] The step of generating the second input image may involve generating the second input image by dividing the vehicle region identified in the first sub-image or the selected corresponding vehicle image.

[0113] The window area detection model is a segmentation model configured to perform segmentation operations, similar to the vehicle area detection model. Since the window area detection model performs operations similar to the vehicle area detection model, the description will focus on the differences.

[0114] The above window area detection model is a deep learning model that is pre-trained to analyze an input image pixel by pixel, identify pixels corresponding to windows, and calculate a window area in the input image (S250). That is, the above window area detection model can also be referred to as a window area detection model.

[0115] The window area detected by the window area detection model is an area representing a window portion located on the left side, right side, or rear of a vehicle, and is an area where a transparent material (e.g., glass, plastic, etc.) is exposed. In certain embodiments, the window area may be an area from which the window frame of the window is excluded.

[0116] Identifying the window area may involve determining the location of the area defined as the window area. That is, the result of identifying the window area indicates the shape in which the window area is represented within the vehicle area, and / or the location of the window area. The result of identifying the window area can be used to calculate the area of ​​the window area.

[0117] FIG. 5 is a schematic diagram of the internal architecture of a window area detection model according to various embodiments of the present application. FIG. 6 is a detailed flowchart of the operation of the window area detection model of FIG. 5.

[0118] Referring to FIG. 5, the window area detection model may be configured to identify the location of a preset part of interest of the vehicle in the exterior area of ​​the vehicle, and to identify the window area based on the identified part of interest location. In various embodiments of the present application, the window area detection model may include an ROI detection unit (301) and a window detection unit (307). Additionally, in some embodiments, the window area detection model may be further configured to extract color features on a pixel-by-pixel basis from the exterior area of ​​the vehicle. In some embodiments, the window area detection model may further include a color feature extraction unit (303) that extracts color features from an input image.

[0119] Referring to FIG. 6, the reference area analysis unit (130) may include: an operation (S271) of identifying the location of a pre-set part of interest in the vehicle area by inputting a second sub-image showing the vehicle area having a surface satisfying the position limit condition to the ROI detection unit (301) in order to detect the window area of ​​the target vehicle located on the vehicle surface in a direction satisfying the position limit condition; an operation (S273) of extracting the color features of the target vehicle pixel by pixel in the vehicle area by inputting the second sub-image showing the vehicle area to the color feature extraction unit (303); an operation (S275) of generating input data for the window detection unit (307) based on the identified part of interest location and the result of extracting color features; and an operation (S277) of identifying the window area of ​​the target vehicle by inputting the input data of the operation (S275) to the window detection unit (307).

[0120] The ROI detection unit (301) may include a region of interest (ROI) detection model configured to identify the location of a preset part of interest of the vehicle in an input image. The ROI detection unit (301) may identify the location indicated by the output result of the ROI detection model as the location of the preset part of interest of the vehicle.

[0121] Specifically, the ROI detection model may include a shape feature extraction unit that extracts shape features from an input image. The shape features are geometric features that describe the shape of an object, and may be, for example, edges, corners, curves, etc., but are not limited thereto. The ROI detection unit (301) extracts shape features from an input image and can identify the location of a pre-set part of interest of a vehicle based on the extracted shape features.

[0122] In some embodiments, the ROI detection unit (301) may extract shape features from the input image using one or more sampling layers and / or pooling layers, and identify a preset region of interest based on the extracted shape features, thereby also identifying the location of the region of interest. The sampling layer may be, for example, a convolution layer having a convolution filter, but is not limited thereto. The pooling layer may be a layer configured to average pool the input data, but is not limited thereto.

[0123] Such ROI detection models can be implemented in the form of various object detection models. For example, the ROI detection model can have various CNN-based structures (e.g., R-CNN (Regions with CNN), Fast R-CNN, Faster R-CNN, YOLO (You Only Look Once), SSD (Single Shot multibox Detextor)), similar to vehicle recognition models.

[0124] However, while the vehicle recognition model is trained to recognize vehicle objects in the input image, the ROI detection model is trained to recognize specific components of the vehicle, described below, as objects of interest.

[0125] The above-mentioned part of interest may be a component of a vehicle that provides a relative positional relationship capable of calculating the position of a window portion. In various embodiments of the present application, the component of the vehicle set as the part of interest may include one or more of a wheel, a license plate, a wiper, a lower boundary line of the vehicle, an upper boundary line of the vehicle, and a side boundary line of the vehicle. The above-mentioned part of interest may be a component of a vehicle that provides a relative positional relationship capable of calculating the position of a window portion. In various embodiments of the present application, the component of the vehicle set as the part of interest may include one or more of a wheel, a license plate, a wiper, a lower boundary line of the vehicle, an upper boundary line of the vehicle, and a side boundary line of the vehicle.

[0126] If a second sub-image is input, the ROI detection unit (301) can identify the location of at least one part of interest of the vehicle that is pre-set in the exterior area of ​​the vehicle. This is because the second sub-image is an image in which at least one part of interest is expressed. The window area detection model can identify the locations of the wheel, the lower part of the side, the upper part of the side, and the side boundary line in the exterior area of ​​the vehicle as the locations of the part of interest when, for example, the exterior area of ​​the vehicle on the side is input. Additionally, the window area detection model can identify the locations of the wheel, the rear license plate, the rear wiper, the lower part of the rear, the upper part of the rear, and the side boundary line in the exterior area of ​​the vehicle on the rear as the locations of the part of interest when, for example, the exterior area of ​​the vehicle on the rear is input (S271).

[0127] The color feature extraction unit (303) can extract color features for each pixel in the input image using one or more frequency filters. The color features for each pixel can be expressed as frequency values ​​corresponding to the corresponding color. The result of extracting the color features can be implemented as a color distribution map (S273). The color distribution map is a distribution map showing the distribution of each color based on the shape of the exterior area of ​​the vehicle.

[0128] In the above operation (S273), color features of the target vehicle can be extracted pixel by pixel from the exterior area of ​​the vehicle in the second sub-image to obtain a color distribution map.

[0129] Due to the transparent nature of windows that transmit light, color features for the windows may not be extracted. For example, from the true window region in an image, color features corresponding to the interior color located behind the window may be extracted, rather than color features corresponding to the window itself.

[0130] On the other hand, since the part of the vehicle body surrounding the windows is opaque, color features of the body part can be extracted relatively accurately. For example, color features extracted from the body region correspond to the actual color expressed on the vehicle body.

[0131] The window detection unit (307) may be configured to identify a window area in the exterior area of ​​the vehicle based on the location of the identified part of interest, or the result of extracting the location of the identified part of interest and color features.

[0132] In the above operation (S275), the input data for the window detection unit (307) is a combination of the location of the identified part of interest and the extraction result of color features, and the location of the identified part of interest may correspond to an image marked on the color distribution map. In some embodiments, the window area detection model may generate the input data by marking the part of interest identified in the operation (S271) at the corresponding location on the color distribution map based on the shape of the exterior area of ​​the vehicle in the color distribution map (S275). The input data may be a feature map recording color features and the location of the part of interest in pixel units. The size of the feature map may correspond to the size of the image marked on the color distribution map where the location of the identified part of interest is marked.

[0133] The window detection unit (307) may include a pre-trained artificial neural network model that infers a first correlation between the location of the part of interest and the location of the window area based on the location of the part of interest, and calculates the location of the window area in the exterior area of ​​the vehicle corresponding to the input data. The window detection unit (307) can identify the location of the part of interest based on the location of the part of interest calculated by the artificial neural network model (S277).

[0134] In some embodiments, when the window area detection model includes the color feature extraction unit (303), the window detection unit (307) may include a pre-trained artificial neural network model to infer a first correlation between the location of the part of interest and the location of the window area based on the location of the part of interest, and to infer a second correlation between the color distribution and the location of the window area, and to calculate the location of the window area in the exterior area of ​​the vehicle corresponding to the input data based on the inferred first correlation and second correlation. The window detection unit (307) can identify the location of the part of interest based on the location of the part of interest calculated by the artificial neural network model (S277).

[0135] The above artificial neural network model may include deep learning models in various forms of artificial neural networks.

[0136] In various embodiments of the present application, a training data set for a vehicle recognition model, a vehicle area detection model, and a window area detection model may consist of one or more sets, such as a first training data set containing images of a target vehicle with its exterior captured before outdoor advertisements were applied to the vehicle, and a second training data set containing images of a target vehicle with its exterior captured after outdoor advertisements were applied to each of the same vehicles within the first training data set. Training images within the first training data set and training images within the second training data set may be obtained from the same sample vehicle depending on whether outdoor advertisements are applied to the vehicle. The same set of training images may be used to train multiple models.

[0137] Meanwhile, the window area detection model including the ROI detection unit (301) and the window detection unit (307) can be trained using a training data set consisting of a plurality of training samples. The training samples in the training data set have training images representing the exterior area of ​​a sample vehicle. In some embodiments, the training samples may further include first label data representing the actual location of a part of interest in the exterior area of ​​the sample vehicle, and second label data representing the actual location of an actual window area. As described above, the training data set for the window area detection model may consist of one or more sets of a first training data set containing images of a target vehicle with its exterior captured before outdoor advertisements were applied to the vehicle, and a second training data set containing images of a target vehicle with its exterior captured after outdoor advertisements were applied to each of the same vehicles in the first training data set.

[0138] The above window area detection model can be trained to more accurately identify parts of interest, such as wipers, and to more accurately identify window areas.

[0139] Specifically, the model parameters of the ROI detection unit (301) are trained to more accurately identify a part of interest, such as a wiper. In some embodiments, the ROI detection unit (301) is trained to minimize the error between the predicted location of the part of interest and the actual location. The predicted location of the part of interest is obtained from a training image using the ROI detection unit (301). The actual location is obtained from the corresponding first label data included in the same training sample as the training image that provided the predicted location.

[0140] The model parameters of the window detection unit (307) are trained to more accurately identify the window area. Specifically, the model parameters of other deep learning models, such as a vehicle recognition model or a vehicle area detection model, are trained to be specialized in identifying various visual features (e.g., surface, shape, texture, etc.) of a relatively opaque object (e.g., vehicle), rather than being specialized in identifying a relatively transparent window area. These deep learning models recognize objects or detect object areas by extracting features of objects, and features extracted from opaque areas contain rich information for recognizing information about the opaque area or identifying the extracted area. For example, the more opaque it is, the more clearly textures and boundaries appear. Therefore, other deep learning models, other than the window area detection model including the window detection unit (307), cannot accurately detect a relatively transparent window area in the target image. On the other hand, features extracted from a relatively transparent area contain relatively poor information. For example, the more transparent it is, the more indistinctly textures and boundaries appear. A window area detection model including a window detection unit (307) can accurately detect the window area even if features with poor information are extracted in a relatively transparent area.

[0141] In various embodiments of the present application, the window detection unit (307) is trained such that the error between the predicted location of the window area and the actual location is minimized. The predicted location of the window area is obtained from a training image using the window detection unit (307). The actual location is obtained from the corresponding second label data included in the same training sample as the training image that provided the predicted location.

[0142] In some embodiments, the window area detection model may be trained to minimize the loss function of the window area detection model, which can be expressed in the form of the following mathematical formula.

[0143]

[0144] Here, L represents the loss function of the window region detection model. And E ROI position , E window position represents the predicted location of the part of interest and the predicted location of the window area, respectively, calculated by the ROI detection unit (301) and the window detection unit (307) from the same sample training image (i.e., the exterior area of ​​the same sample vehicle). And T ROI position , T window position represents the actual location of the region of interest and the actual location of the window area, respectively, in the same sample training image (i.e., the exterior area of ​​the same sample vehicle).

[0145] By the window detection unit (307) of this window area detection model, the location of the window area can be identified more accurately based on the location of the identified part of interest and the extracted color features. In particular, transparent window areas that are difficult to detect compared to opaque areas can be accurately detected.

[0146] The remaining area excluding the window area identified in the operation (S250) from the exterior area of ​​the vehicle for a specific surface direction identified in the vehicle image through the operation (S230) is defined as a reference area for determining the area limit condition. The reference area is not limited to the area between the window area, where vehicle outdoor advertisements are typically attached, and the lower boundary line of the vehicle, but includes the gap area between the window side, where vehicle outdoor advertisements are typically not attached, and the side boundary line of the vehicle, and the gap area between the upper surface of the window and the upper boundary line of the vehicle. Additionally, the reference area may further include the gap area between the windows.

[0147] FIG. 7 illustrates a vehicle standard area used to verify whether an outdoor advertisement satisfies legal guidelines according to various embodiments of the present application.

[0148] Referring to FIG. 7, the reference area detected in the left-side vehicle image satisfying the position limit condition includes a first area (300a), a second area (300b), and a third area (300c).

[0149] The first area (300a) is the area between the window area, where conventional outdoor advertisements on vehicles are generally attached, and the lower boundary line of the vehicle. The second area (300b) is the gap area between the window side of the target vehicle and the side boundary line of the vehicle. The third area (300c) is the gap area between the upper surface of the window of the target vehicle and the upper boundary line of the vehicle.

[0150] Furthermore, as explained below, the aforementioned reference area is used not only to verify whether outdoor advertisements comply with the area restriction conditions of the legal guidelines, but also corresponds to the space where the display area of ​​the outdoor advertisement can be located. This is because it was detected in consideration of location restriction conditions. In other words, the advertisement layout can be placed within the reference area.

[0151] In this way, the reference area analysis unit (130) can detect the reference area. The result of detecting the reference area can be expressed as the location and / or boundary of the reference area in the vehicle area.

[0153] The above system (100) can calculate the area of ​​the reference area of ​​the target vehicle in the reference area analysis unit (130) or the guide verification unit (170). The area of ​​the reference area is a portion of the area on the surface of the vehicle in a direction that satisfies the positional restriction condition on the vehicle.

[0154] For clarity of explanation, the process of calculating the area of ​​the reference region is explained in more detail through examples performed by the reference region analysis unit (130).

[0155] The above reference area analysis unit (130) can calculate the area of ​​the reference area of ​​the target vehicle in various ways.

[0156] In one example, the reference area analysis unit (130) may calculate the area of ​​the reference area of ​​the vehicle body by obtaining size information of the vehicle body from vehicle-related information, or calculate the area of ​​the reference area of ​​the vehicle body by obtaining area information of the vehicle body from vehicle-related information.

[0157] In another example, it can be obtained through user input from an advertiser indicating the size or area of ​​the vehicle body.

[0158] In another example, the reference area analysis unit (130) can analyze a vehicle image to calculate the area of ​​the reference area of ​​the vehicle body. Specifically, the reference area analysis unit (130) can recognize the depth of the shooting space through a camera device that captures the vehicle image, track the position of the subject in three-dimensional space, and estimate the size of the subject based on this. In this process, the estimated size of the subject can be corrected to a more accurate value by considering the physical characteristics of the camera device (e.g., focal length of the lens) and image distortion. The reference area analysis unit (130) can calculate the area of ​​the vehicle area based on the estimated size of the vehicle area.

[0159] The use of the above reference area is described in more detail with reference to Figure 9 below.

[0161] In alternative embodiments, the reference area analysis unit (130) may be configured to perform some or all of the operations (S201 to S290) as a single operation. In some embodiments, the operations (S230 and S250) may be performed as a single operation. In this case, the operation (S200) includes an operation (S230) of recognizing a vehicle and dividing the vehicle's location and vehicle's area from a vehicle image selected to satisfy the location restriction condition, using a single model configured to perform all operations of the vehicle recognition model and the vehicle area detection model after the operation (S210), and the operation (S250) of FIG. 4 may be omitted.

[0163] The above advertising material generation unit (150) can generate outdoor advertising materials desired by the advertiser through a pre-learned advertising material generation model.

[0164] The above advertising material generation unit (150) can obtain the advertiser's basic advertising data and advertising display conditions, and use them to generate an outdoor advertising material.

[0165] Advertising basic data is material that describes the fundamental content the advertiser wishes to reflect in the advertisement.

[0166] In various embodiments of the present application, the advertising basic data may include one or more of advertising theme information, source image, advertising text information, and advertising object information.

[0167] Advertising theme information may be information describing the nature of the advertising content displayed on an outdoor advertisement. The above advertising theme information may be, for example, an advertising target, an advertising topic, an advertising concept, an advertising target audience, and / or other theme information.

[0168] Additionally, in some embodiments, the advertising theme information may further include a draft of the advertising layout and / or a draft of the background color. The draft of the background color may be an opaque or transparent color in various visible light regions. Then, an outdoor advertisement having a layout structure identical or similar to the draft of the advertising layout is generated. Alternatively, an outdoor advertisement having a background color identical or similar to the draft of the background color is generated.

[0169] An ad layout refers to the arrangement structure between ad components, such as ad text and ad images. The ad layout provides the total surface area of ​​the advertisement to which the ad text, ad images, and ad background are applied.

[0170] The specification of the above-mentioned advertisement layout, the specification of advertisement components (advertisement text, advertisement image), and the selection of the background color can be performed based on user input to the computing device.

[0171] Advertising target information describes the advertising target and may be a name, category (e.g., product category), profile image, product specifications, activity history, or other information.

[0172] A source image is an image that is identical or similar to the source image and is included and displayed in an outdoor advertisement. The source image may represent an advertisement target or an advertisement background. One or more source images may be used to generate an outdoor advertisement.

[0173] Advertising object information is various information describing an object displayed on an outdoor advertisement. The object displayed on the outdoor advertisement may be the advertised target or another object. For example, if the advertised target is a mobile phone, the advertising object may be a mobile phone object and / or a user object using the mobile phone.

[0174] The ad display conditions indicate the location on the target vehicle where the advertiser wishes to display the ad.

[0175] In various embodiments of the present application, the advertising display conditions may include one or more of a display surface indicating the surface direction where the outdoor advertisement is to be displayed, an advertisement category, an advertisement type, and an initial display location superficially or roughly indicating the location where the outdoor advertisement is to be placed on the display surface.

[0176] The marking surface may be some or all of the vehicle surfaces that satisfy the positional constraint conditions. The marking surface may be a single surface or two or more consecutive surfaces. The marking surface may be, for example, any one of the rear, left side, and right side, or a combination of the rear and left side, a combination of the rear and right side, or a combination of the rear, left side, and right side.

[0177] The advertisement category indicates whether the advertisement is a fixed advertisement that is fixed to a surface and displayed, or a rotating advertisement that is installed on a roller screen (200) described below and rotates to be exposed to the outside for a certain period of time, and then hidden from the outside after this certain period of time has passed.

[0178] The above-mentioned advertising material type is a classification of types in which an advertising material is displayed on each display surface. The above-mentioned advertising material type may be an integrated type in which a single advertising material is displayed on the display surface, or a distributed type in which two or more advertising materials are distributed across the display surface. Distributed advertising materials may be series advertising materials that are related in content regarding the same advertiser or target audience, but differ in some or all of their advertising components.

[0179] In some embodiments, the advertisement generating unit (150) may generate an outdoor advertisement by additionally utilizing the display conditions of a preset advertising media provider. The advertising media provider is a user who provides the vehicle on which the outdoor advertisement is installed, and may be, for example, a vehicle company or a vehicle operator, but is not limited thereto.

[0180] The initial display location of the above display conditions is the draft location prior to receiving the completed outdoor advertisement. The above initial display location may refer to a point or area located on the remaining area of ​​the vehicle area, excluding the window area, corresponding to the reference area below. For example, the above initial display location may refer to the display area where the conventional outdoor advertisement of FIG. 1, which is familiar to the advertiser, is installed, but is not limited thereto.

[0181] The final display position of the above outdoor advertisement may be maintained as the display position specified in the display conditions, or it may be adjusted according to input from the advertiser or other users after the creation of the outdoor advertisement.

[0182] The display conditions of the aforementioned advertising media provider are various restriction conditions set according to user input. The display conditions of the aforementioned advertising media provider may, for example, be locations where outdoor advertisements are not to be displayed (e.g., vehicle license plate display areas, route / bus stop display areas, etc.) in addition to location restriction conditions, but are not limited thereto.

[0183] The above-mentioned advertisement generating unit (150) generates input data for a model that includes at least an advertiser display condition and an advertiser display condition among basic advertisement content, an advertiser display condition, and an advertising media provider display condition; and applies the input data of the model to a pre-trained advertisement generating model to visually display at least partially the basic advertisement content and generate an outdoor advertisement according to the advertiser display condition.

[0184] The above-mentioned advertisement generation model may be various generative AI models capable of generating images based on input data. Some or all of the above-mentioned generative AI models may be implemented, for example, as a Generative Adversarial Network (GAN), a diffusion model, a Variational AutoEncoder (VAE), an autoregressive model, a flow-based model, etc., and may also be implemented in combinations or modified forms thereof.

[0185] It is a deep learning model that has been pre-trained to generate an image that visually implements such input data (basic content of advertisements, advertiser display conditions, etc.). The generative AI model receives basic data or prompt data in text form as input data, converts it into feature representations in a specific space, generates an image based on this, and generates an image as output data that reflects the intent and constraints of the input data.

[0186] In specific embodiments, the generative AI model may be based on a diffusion model structure. A diffusion model-based model encodes input data into a vector in a latent space and generates a desired image by sequentially restoring a noisy image from this space. Specifically, the advertisement generation model is a deep learning model pre-trained to generate an image that visually implements input data (basic content of an advertisement, advertiser display conditions, etc.). The generative AI model receives basic data or prompt data in text form as input data, encodes it into a vector in a latent space, and sequentially generates and combines visual features (e.g., color, shape, texture, etc.) based on the encoded vector to generate an image as output data that reflects the intent and constraints of the input data. Here, the latent space refers to a multidimensional feature space in which the semantic features of the input data are quantified and expressed.

[0187] In other specific embodiments, the generative AI model may be based on a Generative Adversarial Network (GAN) structure. The GAN-based model operates in an adversarial learning manner in which a generator generates an image from a noise vector and a discriminator evaluates the quality of the generated image.

[0188] The embodiments of the present invention are not limited to the models exemplified above, and various forms of generative models capable of generating a desired image from input data may be utilized. Additionally, the advertisement generation model may be implemented internally or by utilizing one or more external generative APIs.

[0189] In various embodiments of the present application, the advertisement generation model may be implemented as a transformer-based model.

[0190] FIG. 8 is a schematic diagram of the internal architecture of a generative AI model according to various embodiments of the present application.

[0191] Referring to FIG. 8, the advertisement generation model includes an image generation layer 830. Additionally, in some embodiments, the advertisement generation model may further include a prompt layer 810.

[0192] Prompt Layer 810 is a layer that converts the input data of the advertisement generation model into prompt data according to the prompt format when the input data of the advertisement generation model is not in the prompt format. Prompt Layer 810 can be configured to understand the context and meaning of the input data and convert it into the prompt format based on a transformer structure.

[0193] In some embodiments, the prompt layer 810 consists of an encoder and a decoder; the encoder converts the input ad base content and advertiser display conditions into vectorized features, and the decoder reconstructs these features into prompt sentences suitable for image generation. In this process, the relationships between each element of the input data are learned through a self-attention mechanism, and based on the learned relationships, an optimal prompt is generated to produce an outdoor advertisement that reflects the advertiser's requirements.

[0194] Image Generation Layer 830 is a layer that generates visually attractive outdoor advertisements by receiving prompt data reflecting basic advertising content and advertiser display conditions, and reflecting the advertiser's intentions, requirements, and the characteristics of vehicle advertising.

[0195] In some embodiments, the image generation layer 830 may be based on a diffusion model structure that progressively generates an image from prompt data. This layer starts from initial noise in the latent space and gradually removes noise while reflecting the semantic information of the prompt to generate a clear image. In some embodiments, the image generation layer 830 generates an image that considers both fine details and the overall composition by extracting and combining features at various resolutions. For example, the image generation layer 830 may extract features for each resolution separately through a U-Net structure or other structures and combine them. Additionally, in some embodiments, the image generation layer 830 closely links the text information of the prompt with the image generation process through various attention mechanisms, such as a cross-attention mechanism, to ensure that the advertiser's requirements are visually and accurately implemented.

[0196] In specific embodiments, the advertising material generation unit (150) can convert the input data of the model into prompt data through a prompt layer 810 and apply the prompt data to an image generation layer 830 to generate an outdoor advertising material image that reflects the requirements of the advertiser, in order to apply the input data of the model to a pre-trained advertising material generation model.

[0197] Prompt data is text-based data that includes key elements necessary for ad creation in a structured form. The prompt data may include key elements of the basic ad content (e.g., ad target, ad concept, background, etc.), constraints on advertiser display conditions (e.g., display location, ad type, etc.), and ad style information (e.g., image tone, mood, etc.).

[0198] For example, if an advertiser wants to "display the latest smartphone products on the rear of a vehicle targeting young people," the prompt layer 810 can convert this input data into prompt data such as "modern sleek smartphone, dynamic composition, youthful design, vibrant colors, perfect for bus rear display, professional advertising style." This prompt data is implemented as an actual advertisement image in the image generation layer 830.

[0199] In other specific embodiments, the advertisement generation model includes an image generation layer 830, and the input data of the model may be prompt data representing basic advertisement content, advertiser display conditions, etc. The prompt data may be generated according to the user input of the advertiser, which is induced after the system (100) provides the basic advertisement content, advertiser display conditions, etc. to the advertiser. Then, the advertisement generation unit (150) can generate an outdoor advertisement by directly inputting the prompt data into the image generation layer 830 without passing through the prompt layer 810. Specifically, the outdoor advertisement is generated by inputting prompt data written directly by the advertiser or the advertisement designer into the image generation layer 830. This may be useful when an advertising expert wishes to directly control the desired advertisement image through prompt engineering.

[0200] The method of implementing the aforementioned advertising material generation model in the above system (100) can be implemented in various ways by implementing the model having the aforementioned operation and / or components within the system (100) or by utilizing an external generative AI API service.

[0201] For example, in the advertisement generation model according to the specific embodiments above, the prompt layer 810 may be implemented using the Large Language Model API, and the image generation layer 830 may be implemented using various image generation APIs such as the Stable Diffusion API and the DALL-E API. Similarly, in the advertisement generation model according to the other specific embodiments above, the image generation layer 830 may be implemented using one or more external image generation APIs.

[0202] Furthermore, the aforementioned advertisement generation model may be implemented by combining a self-implementation method with a method utilizing external APIs. For example, it is possible to implement the prompt layer 810 in-house while utilizing an external API for the image generation layer 830, or vice versa. This flexibility in implementation can enhance the scalability and efficiency of the system.

[0203] The outdoor advertisement generated by the above advertisement generation unit (150) is provided to the above guide verification unit (170).

[0205] The guide verification unit (170) determines whether the outdoor advertisement generated by the advertisement generation unit (150) satisfies the legal guide. The guide verification unit (170) can verify whether the outdoor advertisement generated by the advertisement generation unit (150) satisfies the area limit conditions of the legal guide based on the area of ​​the standard area calculated by the standard area analysis unit (130).

[0206] Additionally, the guide verification unit (170) may provide a warning message to the advertiser regarding outdoor advertisements that do not satisfy the area restriction conditions of the legal guide, informing the advertiser of the fact that the area restriction conditions are not satisfied. The warning message may include a fact notification and / or a message inducing reduction.

[0207] Meanwhile, the above system (100) can obtain size information or area information of the display area of ​​an outdoor advertisement generated by the advertisement generating unit (150). Such size information or area information can be obtained through user input indicating the size or area desired by the advertiser. Additionally, the above system (100) can determine the final display location of the display area of ​​an outdoor advertisement generated by the advertisement generating unit (150) by obtaining the location of the display area. The final display location is located on the vehicle area.

[0208] In various embodiments of the present application, the guide verification unit (170) can obtain area information of the display area of ​​an outdoor advertisement generated by the advertisement generation unit (150).

[0209] In various embodiments of the present application, the guide verification unit (170) may be configured to verify whether an outdoor advertisement having the corresponding area satisfies the area limit conditions of the legal guide based on the area information of the display area of ​​the outdoor advertisement and the area information of the vehicle area.

[0210] The area limit condition value of the above guide verification unit (170) may be updated according to the revision of relevant laws. For example, as described above, the area limit condition may be 50% or less of the standard area, but may be changed according to the revision of relevant laws.

[0211] Additionally, the guide verification unit (170) may be further configured to verify whether an outdoor advertisement displayed at a given location satisfies the location restriction conditions of the legal guide based on the display location information of the outdoor advertisement.

[0212] The position restriction condition of the guide verification unit (170) can be pre-set according to the vehicle type. The value of the position restriction condition of the guide verification unit (170) can be updated in accordance with amendments to relevant laws. For example, for commercial vehicles and privately owned vehicles, all sides excluding the window portion may be allowed as areas where outdoor advertisements can be displayed. That is, the entire remaining part of the vehicle body excluding the window may be determined as a display location that satisfies the position restriction condition.

[0213] The verification operation of the above-mentioned guide verification unit (170) is described in more detail with reference to Fig. 9 below.

[0215] A method for generating an outdoor advertisement that reflects the requirements of an advertiser based on a generative AI model and in compliance with a preset guide according to another aspect of the present application can be performed by a computing system including a processor and memory, such as the system (100) of FIG. 2.

[0216] FIG. 9 is a flowchart of a method for generating an outdoor advertisement that reflects the requirements of an advertiser based on a generative AI model and complies with a preset guide, according to another aspect of the present application.

[0217] Referring to FIG. 9, a method for generating an outdoor advertisement that reflects the requirements of an advertiser based on a generative AI model while complying with a pre-set guide (hereinafter, the outdoor advertisement generation method) comprises: a step (S100) of obtaining an advertiser’s basic advertising data for an advertisement that describes the basic content the advertiser wishes to reflect in the advertisement, and an advertiser’s advertising display condition indicating an area where the advertiser will display the advertisement on a target vehicle; a step (S200) of calculating a standard area of ​​the target vehicle to determine whether the advertiser’s outdoor advertisement satisfies a pre-set legal guide from vehicle-related data and / or data provided by the advertiser; a step (S300) of generating the outdoor advertisement for the target vehicle by applying the advertising basic data and the advertising display condition to a pre-trained advertisement generation model; a step (S400) of obtaining area information of the display area of ​​the outdoor advertisement according to the advertiser’s user input indicating the size or area of ​​the outdoor advertisement; and a step (S500) of determining the display location of the outdoor advertisement according to the display area information according to the advertiser’s user input. It may include a step of verifying whether the outdoor advertisement based on the acquired display area and the determined display location satisfies the legal guide (S600); and a step of providing the outdoor advertisement for the target vehicle that satisfies the area limitation conditions of the legal guide to the advertiser (S700).

[0218] The above basic advertising data, display conditions, etc. can be entered through an advertising creation request (S100).

[0219] In various embodiments of the present application, the advertising basic data may include one or more of advertising theme information, source images, advertising text information, and advertising object information. Additionally, the advertising display conditions may include one or more of a display surface, an advertising material type, and a location. As these information has been described above, a detailed description is omitted.

[0220] Vehicle-related data may be information stored in a data store or obtained by accessing external web servers, internet search engines, or other external information sources.

[0221] The information provided by the advertiser is obtained through an input device of the system (100), a connected camera device, or an external computing device, and may be vehicle-related data or vehicle image(s).

[0222] The step of calculating the reference area of ​​the above-mentioned target vehicle (S200) may be to determine the remaining area obtained by subtracting the glass area from the vehicle area of ​​the above-mentioned target vehicle as the reference area.

[0223] In specific embodiments, the step of calculating a reference area of ​​the target vehicle (S200) may be to calculate the reference area of ​​the target vehicle by analyzing the vehicle image through pre-trained deep learning models. Specifically, the step of calculating a reference area of ​​the target vehicle (S200) may include the step of acquiring a vehicle image of the target vehicle (S201); and the step of detecting a window area of ​​the target vehicle by analyzing the vehicle image through pre-trained deep learning models, and detecting a reference area of ​​the target vehicle based on the detection result of the window area (S230 to S290).

[0224] The vehicle image is an image representing the target vehicle to which the outdoor advertisement is to be applied. The vehicle image may be an image of one or more surfaces of the vehicle prior to the application of the outdoor advertisement, on which the outdoor advertisement has not yet been displayed. In some embodiments, the vehicle image may be an image of one surface among the front, rear, left side, and right side, or a set of images of two or more surfaces.

[0225] In some embodiments, the step of acquiring a vehicle image of the target vehicle may include the step of acquiring image information corresponding to the vehicle image, which describes the image of the target vehicle. The image information may be acquired through an input field displayed on a UI screen that captures the vehicle image or through another UI screen where image information can be entered.

[0226] Image information corresponding to the above vehicle image includes surface direction information describing the direction of the vehicle surface appearing in the vehicle image. Additionally, in some embodiments, the image information further includes vehicle type information describing the vehicle type to which the target vehicle belongs. If the vehicle is a bus, the bus type may be an express vehicle, an intercity vehicle, a city vehicle, an airport vehicle, or a tourist vehicle.

[0227] In some embodiments, the image of each target vehicle in the image set may be captured through a shooting guide that obtains an image of the target vehicle aligned with each surface direction so that an advertisement created for the target vehicle can more accurately satisfy legal restrictions. The step of obtaining a vehicle image of the target vehicle may include the step of providing the shooting guide; and the step of obtaining a vehicle image captured through the shooting guide.

[0228] As the shooting guide has been described above with reference to the reference area analysis section (130) of FIG. 2, a detailed explanation is omitted.

[0229] In other specific embodiments, the step (S200) of calculating a reference area of ​​the target vehicle may include: obtaining vehicle area and glass area information of the target vehicle from vehicle-related information corresponding to the target vehicle; determining the remaining area obtained by subtracting the glass area from the vehicle area of ​​the target vehicle as the reference area; and calculating the area of ​​the reference area by subtracting the area of ​​the glass area from the area of ​​the vehicle area of ​​the target vehicle. Additionally, in some embodiments, if vehicle area and glass area information of the target vehicle is not retrieved from vehicle-related information corresponding to the target vehicle (e.g., by the reference area analysis unit (130)), the method may further include: obtaining a vehicle image of the target vehicle; analyzing the vehicle image through pre-trained deep learning models to detect the window area of ​​the target vehicle, and detecting the reference area of ​​the target vehicle based on the detection result of the window area.

[0230] In these embodiments, the process of detecting the window area of ​​the target vehicle by analyzing the vehicle image through pre-trained deep learning models is described in more detail below.

[0231] In various embodiments of the present application, the step (S200) of calculating a reference area of ​​the target vehicle from the target vehicle may include: inputting the vehicle image into a pre-trained vehicle recognition model to recognize the target vehicle in the vehicle image (S230); inputting a first sub-image including the recognized target vehicle into a pre-trained vehicle area detection model to detect a vehicle area in which the recognized target vehicle is represented in the selected vehicle image (S250); inputting a second sub-image representing the vehicle area within the selected vehicle image into a pre-trained window area detection model to detect a window area of ​​the target vehicle located on the vehicle surface in a direction satisfying the position limit condition (S270); and determining the remaining area excluding the window area from the detected vehicle area within the vehicle image as a reference area for determining the area limit condition based on the detection result for the window area of ​​the target vehicle (S290). In some embodiments, the step (S200) may further include the step (S210) of selecting a vehicle image having a direction that satisfies a pre-set legal guide position restriction condition based on surface direction information within the image information of the target vehicle.

[0232] The first sub-image above may be a vehicle image or a smaller image consisting of a portion of the vehicle image.

[0233] In specific embodiments, the window area detection model used to determine the window area may include an ROI detection unit (301) that extracts shape features from an input image and identifies the location of a preset part of interest of the vehicle based on the extracted shape features, and a window detection unit (307) configured to identify a window area in the vehicle area based on the location of the identified part of interest.

[0234] Then, the step (S270) of inputting a second sub-image representing the vehicle area within the selected vehicle image into a pre-trained window area detection model to detect the window area of ​​the target vehicle located on the vehicle surface in a direction satisfying the position constraint condition may include: inputting the second sub-image showing the vehicle area into the ROI detection unit (301) to identify the location of a pre-set part of interest in the vehicle area; generating input data for the window detection unit (307) based on the identified part of interest location; and inputting the input data into the window detection unit to identify the window area of ​​the target vehicle.

[0235] Additionally, in other specific embodiments, the window area detection model may further include a color feature extraction unit (305) that extracts color features pixel by pixel from the input image. Then, the step (S270) of detecting the window area of ​​the target vehicle located on the vehicle surface in a direction satisfying the position constraint condition by inputting a second sub-image representing the vehicle area within the selected vehicle image into a pre-trained window area detection model may include: inputting the second sub-image showing the vehicle area into the ROI detection unit to identify the location of a pre-set part of interest in the vehicle area; inputting the second sub-image showing the vehicle area into the color feature extraction unit to extract color features of the target vehicle pixel by pixel in the vehicle area; generating input data for the window detection unit based on the identified part of interest location and the extraction result of color features; and inputting the input data into the window detection unit to identify the window area of ​​the target vehicle.

[0236] Additionally, in some embodiments, the input data for the window detection unit is a combination of the location of the identified part of interest and the extraction result of color features, and the location of the identified part of interest corresponds to an image marked on a color distribution map, and the color distribution map may be a distribution map showing the distribution of each color based on the shape of the vehicle area. Additionally, in some embodiments, the window detection unit (307) may include a pre-trained artificial neural network model to infer a first correlation between the location of the part of interest and the location of the window area based on the location of the part of interest, infer a second correlation between the color distribution map and the location of the window area, and calculate the location of the window area in the vehicle area corresponding to the input data based on the inferred first correlation and second correlation.

[0237] The loss function used to train the artificial neural network of the window detection unit (307) may include a first term (L1) corresponding to the first correlation and a second term (L2) corresponding to the second correlation.

[0238] In some embodiments, the artificial neural network of the window detection unit (307) in the window area detection model can be trained to minimize the loss function expressed by Equation 1.

[0239] A window region detection model learned through the above loss function can be trained using a training data set consisting of multiple training data. Specifically, the training data set may consist of one or more sets of a first training data set containing images of a target vehicle with its exterior captured before outdoor advertisements were applied to the vehicle, and a second training data set containing images of a target vehicle with its exterior captured after outdoor advertisements were applied to each of the same vehicles within the first training data set.

[0240] In various embodiments of the present application, the part of interest may be a component of a vehicle that provides a relative positional relationship capable of calculating the position of a window portion. In various embodiments of the present application, the component of the vehicle set as the part of interest may include one or more of a wheel, a license plate, a wiper, a lower boundary line of the vehicle, an upper boundary line of the vehicle, and a side boundary line of the vehicle. The part of interest may be a component of a vehicle that provides a relative positional relationship capable of calculating the position of a window portion. In various embodiments of the present application, the component of the vehicle set as the part of interest may include one or more of a wheel, a license plate, a wiper, a lower boundary line of the vehicle, an upper boundary line of the vehicle, and a side boundary line of the vehicle.

[0241] The above vehicle recognition model, vehicle area detection model, and window area detection model have been described above with reference to FIGS. 2, FIGS. 4 to 6, so a detailed explanation is omitted.

[0243] The advertisement generation unit (150) can generate an outdoor advertisement that reflects the advertiser's requirements, such as basic advertising data and advertising display conditions, through a pre-learned advertisement generation model (S300).

[0244] In various embodiments of the present application, the step of generating the outdoor advertisement for the target vehicle (S300) may include the step of storing input data used to generate the outdoor advertisement for the target vehicle as metadata of the outdoor advertisement in the system (100).

[0245] The metadata of the above outdoor advertisement can be used to generate information describing the image of the outdoor advertisement, such as a summary.

[0246] The advertisement generation model is a generative AI model pre-trained to generate an image that visually implements the information of input data (basic advertisement content, advertiser display conditions, etc.). The generative AI model has a deep learning model network structure.

[0247] The aforementioned generative AI model receives basic data or prompt data in text form as input, encodes it into vectors in a latent space, and sequentially generates and combines visual features (e.g., color, shape, texture, etc.) based on the encoded vectors to produce an image as output data that reflects the intent and constraints of the input data. Here, the latent space refers to a multidimensional feature space where the semantic features of the input data are quantified and expressed. In this context, sequential generation means performing the task of visually restoring features of the resolution dimension corresponding to the hierarchy sequentially according to the hierarchy.

[0248] In some embodiments, the generative AI model may be configured to process input data and generate images based on a transformer structure. The transformer structure maps the meaning of the input data to visual elements by utilizing a self-attention mechanism that identifies the relationships between each element of the input data and a cross-attention mechanism that converts text information into image features.

[0249] Specifically, the self-attention mechanism calculates the strength of the association between each element in the input data and other elements, expresses this as a weight, and identifies the contextual meaning of each element based on these weights. For example, in the input "smartphone advertisement targeting the young generation," the influence of the element "young" on the elements "smartphone" and "advertisement" is quantified and reflected in the determination of the visual style.

[0250] The cross-attention mechanism incorporates this contextual meaning into each stage of the image generation process. Starting from an initial noisy state, the image is progressively refined by referencing the semantic information of the input data at each denoising stage. During this process, visual features of various resolutions are extracted and combined through the U-Net structure, generating an image that accurately reflects the intent of the input data, from the overall composition to minute details.

[0251] Through this structure, generative AI models go beyond simply converting input text into images; by understanding the semantic context of the input data and appropriately implementing it into visual elements, they can generate advertising images that accurately reflect the advertiser's intentions and requirements.

[0252] The process of generating outdoor advertisements using the above-mentioned advertisement generation model has been described above with reference to Fig. 8, so a detailed explanation is omitted.

[0254] The area information regarding the above outdoor advertisement indicates the area on the vehicle body where the outdoor advertisement is actually displayed.

[0255] In one embodiment, the area information of the outdoor advertisement may be based on the area of ​​a reference area for the target vehicle. Specifically, the step of obtaining area information for the outdoor advertisement (S400) may include: providing initial area information according to a preset area rule to an advertiser; and, upon obtaining a user input (confirmation input) indicating the advertiser's confirmation of the provided initial area, determining the initial area information as the area information of the outdoor advertisement. Additionally, upon obtaining a user input (correction input) indicating the advertiser's correction of the provided initial area, determining the area information modified according to the correction input as the area information of the outdoor advertisement may further include:

[0256] In some embodiments, the preset area rule may be the maximum allowable area that satisfies the area restriction conditions of the legal guide preset in the system (100). As described below, the initial area is the maximum allowable area according to the area restriction conditions, and may be, for example, 50% of the reference area area, but is not limited thereto.

[0257] In other embodiments, the system (100) may be configured to immediately receive the size or area desired by the advertiser without providing an initial area. In step (S400), the user input of the advertiser indicating the size or area of ​​the outdoor advertisement may be obtained in various ways.

[0258] In another embodiment, the system (100) can obtain information on the size and area of ​​an outdoor advertisement by providing a UI screen (area input screen) that induces user input from an advertiser regarding the value or ratio of the size or area of ​​an outdoor advertisement draft generated (e.g., by the control unit (110)).

[0259] When an advertiser inputs the size of an outdoor advertisement through the area input screen above, the system (100) can calculate the area of ​​the outdoor advertisement based on the input size. Inputting the size of the outdoor advertisement may involve adjusting the size of the frame surrounding the outdoor advertisement. This adjustment of size may be implemented through dragging or arrows, but is not limited thereto.

[0260] In another embodiment, the area input screen may be configured to display the exterior of the target vehicle, and when user input from an advertiser regarding the size or area value or ratio of an outdoor advertisement draft is entered, the outdoor advertisement draft according to the size or area entered by the advertiser may be displayed on the exterior of the target vehicle.

[0261] The exterior of the target vehicle displayed on the area input screen may be in the direction of the vehicle surface where the outdoor advertisement is installed, but is no longer limited. The exterior of the target vehicle may be displayed in a 2D or 3D form.

[0262] This supports advertisers in more easily editing the size and area of ​​outdoor advertisements.

[0263] In some embodiments, the advertiser or other user may select a method for calculating the area of ​​the outdoor advertisement. Specifically, the system (100) may calculate the area of ​​the outdoor advertisement by the frame area of ​​the display area of ​​the outdoor advertisement, or the area of ​​the outdoor advertisement may be calculated by the area of ​​the surface of the vehicle, i.e., the part distinguished from the vehicle area.

[0264] Specifically, the size of the outdoor advertisement may be the size of the frame of the display area. The area of ​​the outdoor advertisement calculated from the size of the outdoor advertisement may be the frame area of ​​the display area. In the case where the background color is transparent, the area occupied by the transparent background portion within the frame is included in the area of ​​the outdoor advertisement.

[0265] In another embodiment, the area of ​​the outdoor advertisement may be the area of ​​the part distinguished from the surface of the vehicle, i.e., the vehicle area. For example, if a background color is not selected and the background portion of the outdoor advertisement for the target bus is made transparent, the area occupied by the transparent background portion may be excluded from the area of ​​the outdoor advertisement for the target bus.

[0266] The step of determining the display location of the outdoor advertisement (S500) is a step of determining the final location where the outdoor advertisement having the area information is to be displayed on the surface of the target vehicle according to the user input of the advertiser.

[0267] The processes for inducing user input to determine the final display location of an outdoor advertisement can be implemented in different ways depending on whether the system (100) has obtained the initial display location of the advertisement display conditions in step (S100).

[0268] When an initial display location is obtained with the above-mentioned advertising display conditions, the step of determining the display location of the outdoor advertisement (S500) may include: providing a location setting screen that displays the outdoor advertisement having the above-mentioned area information by placing it at the display location of the above-mentioned advertising display conditions; and determining the final display location according to the advertiser's location adjustment input obtained through the location setting screen.

[0269] In some embodiments, the position setting screen may be provided with an outdoor advertisement having the area information displayed at an initial position (display position of the advertisement display condition) along with the exterior of the target vehicle. The position setting screen may include various UI elements that allow the position of the displayed outdoor advertisement to be adjusted through user interaction. For example, the UI elements may provide functions to move the outdoor advertisement by dragging it, moving it using arrow keys, or moving it to a specific position by pointing at a specific point on the vehicle surface.

[0270] If an initial display location is not obtained under the above advertising display conditions, the step of determining the display location of the outdoor advertisement (S500) may include: requesting an initial location designation from the advertiser through a location input screen displaying the exterior of the target vehicle; displaying the outdoor advertisement having the area information at the designated initial location; and determining the final display location according to the advertiser's location adjustment input obtained through the location input screen.

[0271] In some embodiments, the position input screen may include UI elements that allow an initial position to be specified through pointing to a vehicle surface. Once the initial position is specified, the position input screen may provide a position adjustment function in the same or similar manner as the position setting screen.

[0272] If a location adjustment input is provided through the location input screen or the location setting screen, the adjusted location may be determined as the final display location (S500). If the advertiser does not provide a location adjustment input through the location input screen or the location setting screen, the initial display location may be determined as the final display location (S500).

[0274] The guide verification unit (170) verifies whether the outdoor advertisement, whose area and location are determined in steps (S400, S500), satisfies the legal guide.

[0275] As described above, the values ​​of various legal restriction conditions in the legal guide depend on current values ​​valid at the time the system (100) operates. For example, according to the current Outdoor Advertising Enforcement Decree (amended on May 21, 2024) as of 2025, the time of filing of this application, the location restriction condition of the legal guide indicates that display is permitted only on the vehicle body or bus protruding number plate excluding the window portion, and the area restriction condition of the legal guide indicates that display must be within 50% of the total area of ​​each side (excluding the window portion).

[0276] In various embodiments of the present application, the step (S600) may include a step (S610) of verifying whether the outdoor advertisement according to the display area information satisfies the area limitation conditions of the legal guide. Additionally, in some embodiments, it may further include a step (S620) of verifying whether the outdoor advertisement according to the display location satisfies the area limitation conditions of the legal guide.

[0277] The step (S620) of verifying whether an outdoor advertisement according to the above display location satisfies the location restriction conditions of the legal guide may include: a step (S621) of identifying an area where the display of the outdoor advertisement is restricted on the target vehicle; a step (S622) of analyzing whether the display location of the outdoor advertisement determined in step (S500) overlaps with the identified restricted area; a step (S623) of determining that the outdoor advertisement for the target vehicle does not satisfy the location restriction conditions if the display location of the outdoor advertisement overlaps with the restricted area; a step (S624) of readjusting the display location of the outdoor advertisement so that the outdoor advertisement for the target vehicle satisfies the location restriction conditions; and a step (S625) of determining that the outdoor advertisement for the target vehicle satisfies the location restriction conditions if the display location of the outdoor advertisement does not overlap with the restricted area.

[0278] The above-mentioned restricted area depends on legal restrictions. The above-mentioned restricted area may, for example, be a window area, but is not limited thereto. The above-mentioned window area may be a transparent portion, for example, an area where transparent plastic is exposed. In some embodiments, the window frame is excluded from the above-mentioned window area.

[0279] In some embodiments, the step of identifying the restricted area (S621) may include: identifying the window area of ​​the target vehicle; and identifying the remaining area of ​​the vehicle body excluding the window area as an area where advertisements can be displayed.

[0280] Additionally, in some embodiments, the readjusting step (S624) may involve moving the display position of the outdoor advertisement to the nearest location where the outdoor advertisement does not overlap with the restricted area. For example, if the outdoor advertisement partially overlaps with the window area, the outdoor advertisement may be moved to the nearest location within the vehicle body area, avoiding the window area.

[0281] FIG. 10 is a detailed flowchart of the process for determining whether an outdoor advertisement on a target vehicle satisfies the area limit conditions of the legal guide according to various embodiments of the present application.

[0282] Referring to FIG. 10, the step (S610) of verifying whether the outdoor advertisement according to the display area information satisfies the area limit conditions of the legal guide comprises: the step (S611) of calculating the area of ​​the reference area in the target vehicle; the step (S612) of calculating the ratio of the area of ​​the outdoor advertisement obtained in the step (S400) regarding the ratio of the calculated area; the step (S613) of analyzing whether the ratio of the area of ​​the outdoor advertisement to the ratio of the area exceeds a threshold ratio (50%) corresponding to the pre-set area limit conditions; the step (S614) of determining that the outdoor advertisement for the target vehicle does not satisfy the area limit conditions if the ratio of the area of ​​the outdoor advertisement to the ratio of the area exceeds the threshold ratio (50%) corresponding to the pre-set area limit conditions; and the step (S615) of reducing and correcting the area of ​​the outdoor advertisement for the target vehicle so that the outdoor advertisement for the target vehicle satisfies the area limit conditions. The method may include a step (S616) of determining that an advertisement for a target vehicle satisfies the area limit condition, wherein the ratio of the area of ​​the outdoor advertisement to the ratio of the above area has an area of ​​less than or equal to a threshold ratio (50%) corresponding to a preset area limit condition.

[0283] The area of ​​the reference area of ​​the vehicle can be obtained by obtaining information on the size or area of ​​the vehicle body from vehicle-related information, by obtaining user input from an advertiser indicating the size or area of ​​the vehicle body, or by calculating the area of ​​the reference area of ​​the vehicle by analyzing the vehicle image (S611).

[0284] The area of ​​the reference area in the above-mentioned target vehicle is the area value of the reference area on the display surface where an outdoor advertisement is to be displayed. For example, if the display surface of the advertisement subject to the area limit condition judgment is the right side, the area of ​​the reference area on the right side is calculated (S611).

[0285] When one outdoor advertisement is installed on one surface (i.e., integrated type), the ratio of the area of ​​one outdoor advertisement to the area of ​​the standard area is calculated (S612).

[0286] When multiple outdoor advertisements are installed on a single surface (i.e., distributed type), the ratio of the total area of ​​multiple outdoor advertisements to the area of ​​a reference area based on the same surface is calculated (S612).

[0287] The step (S612) of calculating the ratio of the area of ​​the outdoor advertisement obtained in the step (S400) for the ratio of the area calculated above may include: a step of summing the area of ​​the outdoor advertisement on the display surface of the outdoor advertisement when the outdoor advertisement is of the dispersed type; and a step of calculating the ratio of the summed area of ​​the outdoor advertisement to the ratio of the area calculated above.

[0288] An outdoor advertisement satisfies the above area limitation condition, meaning that the ratio of the area of ​​the outdoor advertisement to the area of ​​the reference area (i.e., area ratio) is less than or equal to a preset threshold ratio. The threshold ratio corresponds to the preset area limitation condition and is the threshold ratio used to satisfy the above area limitation condition. The threshold ratio is set based on the Enforcement Decree of the Outdoor Advertising Act. In various embodiments of the present application, the threshold ratio may be 0.5.

[0289] If the above reference area has an area (A1) and the outdoor advertisement for the above target vehicle has an area (A2), the area ratio may be A2 / A1. That is, whether the outdoor advertisement satisfies the area limitation condition can be expressed in the form of the following mathematical formula.

[0290]

[0291] Here, R thr It may be a threshold ratio (e.g., 0.5) set in relation to the above area restriction conditions based on the Enforcement Decree of the Outdoor Advertising Act.

[0292] If the area ratio of the outdoor advertisement for the target vehicle generated in step (S400) based on the above reference area is less than or equal to a threshold ratio (e.g., 0.5), the outdoor advertisement for the target vehicle having the said area is determined to be an outdoor advertisement satisfying the area limit condition (S616).

[0293] Meanwhile, if the ratio between the area of ​​the reference area and the area of ​​the outdoor advertisement for the target vehicle generated in step (S400) exceeds a threshold ratio (e.g., 0.5), it is determined that the outdoor advertisement for the target vehicle does not satisfy the area limitation condition (S614). Then, a step (S615) of editing the outdoor advertisement for the target vehicle so that the outdoor advertisement for the target vehicle that does not satisfy the area limitation condition satisfies the area limitation condition is provided.

[0294] In various embodiments of the present application, the step (S615) may be to reduce and correct at least one of the area size of the entire area of ​​the advertisement layout within the outdoor advertisement, the size of the advertisement image, and the size of the advertisement text.

[0295] In some embodiments, the reduction correction step (S615) may include: providing a user interface screen configured to input a reduction correction size value for at least one of the size of the frame of the outdoor advertisement, the size of the advertisement image, and the size of the advertisement text; and changing the size of the component of the outdoor advertisement, into the reduction correction size value.

[0296] Size correction of various ad components may involve adjusting the width, height, or diagonal size.

[0297] For example, if a reduction correction value for the size of the advertisement text is entered in the above step (S615), the advertisement text is changed from the existing size value to a size corresponding to the entered reduction correction value.

[0298] The above user interface screen may include a first sub-screen for selecting a reduction correction target, a second sub-screen for displaying the current size value of the reduction correction target, and a third sub-screen for inputting the reduced-corrected value of the reduction correction target. The third sub-screen may be implemented in a UI form identical or similar to the area input screen.

[0299] When the above outdoor advertisement is reduced and corrected, steps (S400, S500) are repeated to redetermine the display area and display location of the reduced outdoor advertisement. Specifically, the display area of ​​the reduced outdoor advertisement is re-obtained (S400), and the display location of the reduced outdoor advertisement having the obtained display area is redetermine (S500). Subsequently, whether the legal guide is satisfied with the redetermine display area and display location is verified again (S600). During the re-verification, if the area of ​​the reference area has already been calculated, step (S611) is omitted without needing to be performed again.

[0300] Outdoor advertisements satisfying the above area limit conditions may be output to a computing device or transmitted to an external device and provided to an advertiser (S700).

[0301] Conventionally, since advertisements were produced without considering the spacing between windows, the spacing between windows and the upper boundary line of the bus roof, and the spacing between the window sides and the bus side (or rear) boundary line, the standard area for producing advertisements was relatively small. However, outdoor advertisements on the aforementioned target buses that satisfy these area limitation conditions can have the maximum area permitted by law, thereby satisfying the demands of advertisers who wish to produce advertisements as large as possible.

[0302] In various embodiments of the present application, the step (S700) includes providing an image of an outdoor advertisement for a target vehicle that accurately satisfies legal restriction conditions. Additionally, in some embodiments, the step (S700) may include providing advertising information describing an image of an outdoor advertisement for a target vehicle that accurately satisfies legal restriction conditions.

[0303] The above advertising information may include a ratio value of the advertising area occupied by the outdoor advertisement on the target vehicle relative to the standard area of ​​the target vehicle. Additionally, in some embodiments, the advertising information may further include an advertising area value occupied by the outdoor advertisement on the target vehicle and an area value of the standard area of ​​the target vehicle.

[0304] The advertising designer can provide the advertising information, which indicates the value satisfying the conditions required to create the outdoor advertisement (e.g., area value), as a quantitative basis for persuasion to an advertiser who makes an excessive demand for advertising expansion beyond legal restrictions, so that if there is a dispute with the advertiser regarding the size of the outdoor advertisement provided in step (S700), the advertiser can be more easily persuaded.

[0305] In addition, the frame shape of the outdoor advertisement can deviate from the conventional fixed form and be freely transformed into various shapes.

[0306] In addition, advertisers can receive outdoor advertisements equivalent to the work of a professional designer simply by inputting the desired image file and basic text into the system (100).

[0307] FIG. 11 illustrates an outdoor advertisement generated from advertising basic data according to one example of the present application. FIG. 12 illustrates an outdoor advertisement generated from advertising basic data according to another example of the present application.

[0308] Referring to FIGS. 11 and 12, an outdoor advertisement is generated that includes an image identical or similar to the source image of the basic advertising material.

[0309] Additionally, two or more outdoor advertisements can be generated from multiple source images and advertising texts, displayed at two or more locations on the same vehicle surface. These two or more outdoor advertisements are generated when the condition is distributed or when two or more initial display locations are selected, and are series outdoor advertisements following the same advertising theme.

[0311] In alternative embodiments, the system (100) and method transmit the outdoor advertisement provided to the advertiser in step (S700) to an external device connected to the system (100) to request feedback from an expert (e.g., a professional designer) who provides user input to the external device; and receive the outdoor advertisement corrected according to the expert's feedback.

[0312] Additionally, some or all of the guide verification operations may be additionally performed on the outdoor advertisement corrected according to the expert's feedback received from the external device (S600). For example, the guide verification unit (170) can verify whether the outdoor advertisement corrected according to the expert's feedback satisfies the area limit condition through the operation of steps (S612 to S613).

[0314] Generation of outdoor advertisements considering driving conditions

[0315] A system (100) and method according to another aspect of the present application can generate an outdoor advertisement by further considering vehicle operation situation information.

[0316] The above vehicle operation situation describes the overall situation in which the vehicle is in operation. The above vehicle operation situation relates to the exposure environment of outdoor advertisements displayed while attached to the vehicle.

[0317] The above vehicle operation status information may include one or more of the vehicle's driving time information and driving location information.

[0318] The driving time information of the aforementioned vehicle may be one or more of the time period, day of the week, month, or other time information during which the vehicle is driven with outdoor advertisements attached. Additionally, the driving time of the vehicle is related to the attributes of passengers boarding or alighting from the vehicle. This is because the attributes of people who primarily board the vehicle or pass by it vary with a tendency depending on the time of day. These attributes may be gender, age group, or other attributes.

[0319] The above driving location information describes the locations where a vehicle travels while attached to outdoor advertisements. The above driving location information may include one or more of a route number or route path and a driving area. For example, the route number or path is obtained from bus numbers and geographical location information of bus stops. The driving area is obtained from taxi operating areas.

[0320] The above vehicle operation status may be obtained prior to step (S300) of FIG. 9. In certain embodiments, the method of FIG. 9 may further include a step (S101) of obtaining the vehicle operation status. The vehicle operation status is obtained from operation-related information.

[0321] The above control unit (110) can obtain operation-related information stored in the internal data store of the system (100) or from an external source.

[0323] Creating outdoor advertisements suitable for rotating advertising system vehicles, taking into account driving conditions.

[0324] Additionally, the system (100) and method may be further configured to produce a set of outdoor advertisements installed on a roller screen installed on one surface of a vehicle.

[0325] FIG. 13 is a schematic diagram of a vehicle including a roller screen on which an outdoor advertisement is installed, according to various embodiments of the present application.

[0326] Referring to FIG. 13, the vehicle to which the outdoor advertising set is applied may be a vehicle equipped with various types of rotating advertising systems (200). The vehicle may include a screen (210) that provides a display area of ​​the outdoor advertising (10); and one or more rolling parts (220) that rotate the screen to change the exposure area of ​​the screen.

[0327] The rolling unit (220) may include a roller and at least one motor that controls the rotational speed of the roller. The roller may extend along the transverse direction of the vehicle as shown in FIG. 13, but is not limited thereto. A vertical roller may be installed on the vehicle.

[0328] The above-described rotary advertising system (200) can be implemented through various technologies known at the time of filing of this patent. For example, the above-described rotary advertising system (200) can be implemented using Utility Model Registration Publication No. 20-0230356 "Advertising Screen Rotation Device" or Utility Model Registration Publication No. 20-0434547 "Rotating Advertising Device" or other technologies capable of rotating a screen.

[0329] In certain embodiments, the rolling unit (220) may be configured to receive a control signal from the control unit (110) of FIG. 2 or another control unit connected thereto, and to drive a motor based on the control signal.

[0330] The above-described rotary advertising system (200) may be configured to control the rotation speed and rotation cycle of the rolling part (220). The outdoor advertisement (10) exposed through the screen (210) is controlled by the rotation speed and rotation cycle of the rolling part (220).

[0331] The rotation period can be defined as the time during which the entire set of outdoor advertisements displayed on the screen (210) is sequentially exposed.

[0332] In various embodiments of the present application, the area of ​​the screen (210) may be the maximum allowable area according to the area limit condition.

[0334] The above system (100) can obtain a request for the creation of an advertisement for such a rotating advertisement system (200) (S100). The request for the creation of an advertisement may include an advertisement category representing the rotating advertisement, the number of requests for the rotating advertisement, and conditions for displaying advertisements for each rotating advertisement. As the conditions for displaying advertisements have been described above, a detailed explanation is omitted.

[0335] The number of requests for rotating advertisements represents the number of advertisement configurations of the outdoor advertisement set that the advertiser wishes to display on the roller screen (210).

[0336] Each outdoor advertisement (10) included in the outdoor advertisement set generated in the above step (S300) may follow the same or similar advertising theme, or each may follow an individual advertising theme. For example, the outdoor advertisement set may consist of outdoor advertisements (10) that display various products of a single brand (e.g., mobile phones, TVs, refrigerators, etc.). Alternatively, the outdoor advertisement set may consist of multiple advertisements.

[0337] Each outdoor advertisement (10) in the outdoor advertisement set may be an integrated or dispersed outdoor advertisement image as described above.

[0338] In the step (S500) of determining the display location of an outdoor advertisement according to the display area information based on the user input of the advertiser above, the display location of the outdoor advertisement can be determined as a location on the exposure area of ​​the screen (210).

[0339] A vehicle displaying the above-mentioned rotating advertising system (200) can be displayed on a UI screen such as a location input screen (S500).

[0340] In various embodiments of the present application, the verification step (S610) may be to verify whether all outdoor advertisements in the outdoor advertisement set satisfy an area limit condition. Specifically, the system (100) may perform the verification operation of FIG. 10 for each of the n outdoor advertisements (10) in the outdoor advertisement set.

[0341] When it is determined that all n outdoor advertisements (10) satisfy the area limit condition (S616), the set of outdoor advertisements is provided to the user (S800).

[0342] In various embodiments of the present application, the verification step (S600) may further include a verification step (S620) of whether an outdoor advertising set satisfying an area limit condition can be installed on a roller screen (210). Specifically, the verification step (S630) of an outdoor advertising set for the roller screen (210) comprises: a step of confirming whether all outdoor advertisings (10) of an outdoor advertising set can be displayed on the roller screen (210) based on the length of each outdoor advertising (10) within the outdoor advertising set, which has a determined area and a rotational length of the roller screen (210), before providing the outdoor advertising (10) to the advertiser; and a step of providing the outdoor advertisings (10) of the outdoor advertising set to the advertiser if it is confirmed that all outdoor advertisings (10) of the outdoor advertising set can be displayed on the roller screen (210). And if it is confirmed that all outdoor advertisements (10) of the above outdoor advertisement set cannot be displayed on the roller screen (210), the method may include the step of obtaining user input for selecting outdoor advertisements (10) to be displayed from the advertiser through an outdoor advertisement (10) selection screen and providing the selected outdoor advertisements (10) to the advertiser.

[0343] The guide verification unit (170) obtains the length extended in the rotational direction of the roller screen (210) according to the user input of the advertiser (e.g., the vertical screen length of the advertisement in FIG. 13), sums the length of the outdoor advertisement (10) within the set corresponding to the rotational direction of the roller screen (210) (e.g., the horizontal length based on the vehicle in FIG. 13), and if the sum of the lengths of the outdoor advertisement (10) is less than or equal to the extended length of the roller screen (210), it is determined that all outdoor advertisements (10) of the outdoor advertisement set can be displayed on the roller screen (210). Conversely, if the sum of the lengths of the outdoor advertisement (10) exceeds the extended length of the roller screen (210), it is determined that all outdoor advertisements (10) satisfying the area limit condition of the outdoor advertisement set cannot be displayed on the roller screen (210).

[0344] Then, the system (100) selects at least one outdoor advertisement (10) from an existing set and adjusts the composition of the set of advertisements so that all outdoor advertisements (10) of the outdoor advertisement set can be displayed on the roller screen (210). This selection may be exclusion from the set or selection into the final set.

[0345] In some embodiments, the outdoor advertisement (10) selection screen in step (S620) may include UI elements that display each outdoor advertisement (10) within the outdoor advertisement set and allow selecting an outdoor advertisement (10) to be displayed on the roller screen (210) from among the displayed outdoor advertisements (10). For example, the UI elements may provide a checkbox for each outdoor advertisement (10) or a drag-and-drop selection function. The selected outdoor advertisement (10) may be excluded from the initial set or determined as the final set.

[0346] Additionally, the guide verification unit (170) may be further configured to check whether one or more outdoor advertisements (10) can be added to the outdoor advertisement set when the sum of the lengths of the outdoor advertisements (10) is less than or equal to the extended length of the roller screen (210), and if it is confirmed that the outdoor advertisements (10) can be added, to notify the advertiser of the number of outdoor advertisements that can be added.

[0347] Specifically, the guide verification unit (170) calculates the length in a direction perpendicular to the rotation direction of the roller screen (210) (e.g., horizontal length), calculates the minimum possible length of the outdoor advertisement (10) on the roller screen (210) through the length in a vertical direction calculated from the maximum allowable area satisfying the area limit condition, calculates the difference between the extension length of the roller screen (210) and the sum of the corresponding lengths of the outdoor advertisement (10) in the set, and calculates the number of additional outdoor advertisements (10) by comparing the calculated difference length with the minimum possible length.

[0348] The guide verification unit (170) can calculate the minimum possible length of an advertisement (10) that can be installed on the roller screen (210) by dividing the maximum allowable area corresponding to 50% of the reference area by the length in the direction perpendicular to the rotation direction of the roller screen (210).

[0349] For example, for a roller screen (210) in which the length of the rotation direction of the roller screen (210) is 6m, the length of the direction perpendicular to the rotation direction of the roller screen (210), that is, the length of the direction corresponding to the spacing between the rolling parts (220) (vehicle width in FIG. 13), is 5m, and the minimum possible length of the advertisement (10) that can be installed on the roller screen (210) is 1m, the sum of the lengths of the outdoor advertisements (10) in the set is 5m, the guide verification unit (170) can inform the advertiser that one additional outdoor advertisement (10) having a minimum possible length of 1m can be added.

[0350] Next, the system (100) may receive a request from an advertiser to create additional advertisements. At this time, the request for additional creation has a number of advertisements that is less than or equal to the number of outdoor advertisements (10) that can be added, which is calculated in advance by the guide verification unit (170).

[0351] In addition, in some embodiments, the outdoor advertisement (10) selection screen may display the rotational length of the roller screen (210) and the total length of the currently selected outdoor advertisements (10) in real time, thereby supporting the advertiser in selecting an outdoor advertisement (10) within a displayable range.

[0353] An outdoor advertising set consisting of n (n is a natural number) outdoor advertisements (10) that are determined in this way and provided to the advertiser is sequentially exposed to the outside through the rotation of the screen (210).

[0354] Additionally, the method of FIG. 9 may further include the step (S800) of providing a set of exposure frequencies for a set of outdoor advertisements displayed on a screen (210) that is rotated by a rolling part (220).

[0355] The exposure frequency set is a data set for controlling the exposure time and frequency of n outdoor advertisements (10) within the outdoor advertisement set. The exposure frequency set can be calculated or stored as a data structure in the form of a table or matrix.

[0356] The above exposure time can be expressed as a ratio to the rotation cycle. For example, if a set contains three outdoor advertisements and the exposure time for each outdoor advertisement is the same, the exposure time for each outdoor advertisement in the exposure frequency set can be expressed as {T / 3, T / 3, T / 3}.

[0357] The above exposure frequency set can be generated through various methods. For example, it can be implemented using rule-based calculation, statistical-based calculation, optimization-based calculation, or prediction model-based calculation, and can also be implemented in combinations or modified forms thereof.

[0358] In specific embodiments, the exposure frequency set may be implemented through a prediction model that takes vehicle operation situation information as input data and, for each outdoor advertisement within the exposure frequency set, outputs exposure time and number of exposures according to pairs of driving time and location.

[0359] Specifically, the prediction model may be a generative AI model that is pre-trained to convert vehicle operation situation information into feature representations in a specific space and generate exposure frequency values ​​based thereon. The prediction model receives operation situation information, such as the vehicle's route, time of day, weather, passengers, and region, as input data, encodes this information into a vector in a latent space, and generates optimal exposure frequency values ​​for each outdoor advertisement (10) based on the encoded vector. Here, the latent space refers to a multi-dimensional feature space in which the features of the operation situation information are quantified and expressed.

[0360] In some embodiments, the prediction model may utilize a self-attention mechanism based on a transformer structure to identify the relationships between each element of the input operation status information. For example, the optimal exposure frequency can be determined by considering the impact of a combination of specific time periods and weather conditions on the number of passengers.

[0361] The above exposure frequency set can be transmitted to a rotary advertising system (200). Based on the exposure frequency set, the rotary advertising system (200) controls the rotational speed and rotational cycle of the motor of the rolling part (220) to control the outdoor advertisement (10) that is exposed and displayed to the outside through the roller screen (210).

[0363] This system (100) and method can provide the following effects by performing the process of generating, verifying, and providing outdoor advertisements.

[0364] First, the present invention can integrate and process the existing complex advertising production process—including consultation between advertisers and sales representatives, design work, bus installation schedule coordination, government permits, and installation—into an automated online process. This allows for a significant reduction in the time and cost required for advertising production.

[0365] Second, the automatic design generation function utilizing the generative AI model of the present invention can provide various design options that immediately reflect the advertiser's requirements. This can increase the efficiency of advertising production by minimizing the modification and feedback processes that occur during the existing manual design process.

[0366] Third, the present invention automatically verifies and complies with legal guidelines, thereby confirming the feasibility of installing advertisements in advance and facilitating the related licensing process. This can prevent delays in advertisement production or modification costs caused by legal issues.

[0367] Fourth, the automated system of the present invention supports advertisers in efficiently producing various series of advertisements with the same concept. Through this, advertisers can conduct more consistent and effective advertising campaigns.

[0368] Consequently, the present invention realizes cost reduction and improved work efficiency through the automation of the outdoor advertising production process, which enables the provision of high-quality advertising services to advertisers at more competitive unit prices.

[0370] FIG. 14 illustrates an exemplary computer system capable of implementing the system (100) of FIG. 2.

[0371] In a specific embodiment, one or more included computer systems (800) perform one or more steps of one or more methods described or illustrated herein. In a specific embodiment, one or more computer systems (800) provide a function described or illustrated herein. In a specific embodiment, software running on one or more computer systems (800) performs one or more steps of one or more methods described or illustrated herein, or provides a function described or illustrated herein. A specific embodiment includes one or more parts of one or more computer systems (800). In this specification, a reference to a computer system may include a computing device where appropriate. Furthermore, a reference to a computer system may include one or more computer systems where appropriate.

[0372] The present specification considers any suitable number of computer systems (800). The present specification considers computer systems (800) taking any suitable physical form. For example, without limitation, a computer system (800) may be an embedded computer system, a system-on-chip (SOC), a single-board computer system (SBC) (e.g., a computer-on-module (COM) or a system-on-module (SOM)), a desktop computer system, a laptop or notebook computer system, an interactive kiosk, a mainframe, a computer system mesh, a mobile phone, a personal information terminal (PDA), a server, a tablet computer system, or a combination of two or more of these. Where appropriate, a computer system (800) may include one or more computer systems (800); may be monolithic or distributed; may be spread across multiple locations; may be spread across multiple machines; may be spread across multiple data centers; or may reside in a cloud that may include one or more cloud components in one or more networks. Where appropriate, one or more computer systems (800) may execute one or more steps of one or more methods described or illustrated herein without substantially spatial or temporal limitations. For example, without limitation, one or more computer systems (800) may execute one or more steps of one or more methods described or illustrated herein in real time or in batch mode. Where appropriate, one or more computer systems (800) may execute one or more steps of one or more methods described or illustrated herein at different times or at different locations.

[0373] In a specific embodiment, the computer system (800) includes a processor (802), memory (804), storage (806), an input / output (I / O) interface (808), a communication interface (810), and a bus (812). Although this specification describes and illustrates a specific computer system having a specific number of specific components in a specific arrangement, this specification considers any suitable computer system having any suitable number of any suitable components in any suitable configuration.

[0374] In a specific embodiment, the processor (802) includes hardware for executing instructions, such as instructions that constitute a computer program. For example, without limitation, to execute an instruction, the processor (802) may retrieve (or fetch) the instruction from an internal register, internal cache, memory (804), or storage (806); and after decoding and executing the instruction; may write one or more results to an internal register, internal cache, memory (804), or storage (806). In a specific embodiment, the processor (802) may include one or more internal caches for data, instructions, or addresses. The present specification considers a processor (802) that includes any appropriate number of any appropriate internal caches where appropriate. For example, without limitation, the processor (802) may include one or more instruction caches, one or more data caches, and one or more translation lookaside buffers (TLBs). Instructions in the instruction cache may be copies of instructions in memory (804) or storage (806), and the instruction cache may increase the speed of searching for such instructions by the processor (802). Data in the data cache may be the result of previous instructions executed by the processor (802) to be accessed by subsequent instructions executed by the processor (802) or written to memory (804) or storage (806); or a copy of data in memory (804) or storage (806) for instructions executed by the processor (802) to operate on other appropriate data. The data cache may increase the speed of read or write operations by the processor (802). TLBs may increase the speed of virtual address translation by the processor (802). In a specific embodiment, the processor (802) may include one or more internal registers for data, instructions, or addresses.The present specification considers a processor (802) comprising any suitable number of any suitable internal registers where appropriate. Where appropriate, the processor (802) may include one or more arithmetic logic units (ALUs); may be a multi-core processor; or may include one or more processors (802). Although the present specification describes and illustrates a specific processor, the present specification considers any suitable processor.

[0375] In a specific embodiment, memory (804) includes main memory for storing instructions executed by the processor (802) or data operated by the processor (802). For example, without limitation, a computer system (800) may load instructions into memory (804) from a storage (806) or another source (e.g., another computer system (800)). Subsequently, the processor (802) may load instructions from memory (804) into internal registers or an internal cache. To execute instructions, the processor (802) may retrieve instructions from internal registers or an internal cache and decode them. During or after the execution of instructions, the processor (802) may write one or more results (which may be intermediate or final results) to internal registers or an internal cache. Subsequently, the processor (802) may write one or more of these results to memory (804). In a specific embodiment, the processor (802) executes only instructions in one or more internal registers or internal caches or in memory (804) (in contrast to storage (806) or elsewhere), and operates only data in one or more internal registers or internal caches or in memory (804) (in contrast to storage (806) or elsewhere). One or more memory buses (which may each include an address bus and a data bus) may connect the processor (802) to memory (804). As described below, the bus (812) may include one or more memory buses. In a specific embodiment, one or more memory management units (MMUs) reside between the processor (802) and memory (804) and facilitate access to memory (804) requested by the processor (802). In a specific embodiment, memory (804) includes random access memory (RAM). Where appropriate, such RAM may be volatile memory. Where appropriate, such RAM can be dynamic RAM (DRAM) or static RAM (SRAM).Furthermore, where appropriate, such RAM may be a single-port or multi-port RAM. The present specification considers any suitable RAM. Where appropriate, memory (804) may include one or more memories (804). Although the present specification describes and illustrates specific memories, the present specification considers any suitable memory.

[0376] In a specific embodiment, the storage (806) includes a mass storage for data or instructions. For example, without limitation, the storage (806) may include a hard disk drive (HDD), a floppy disk drive, flash memory, an optical disk, a magneto-optical disk, a magnetic tape, a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the storage (806) may include a removable or non-removable (or fixed) medium. Where appropriate, the storage (806) may be inside or outside the computer system (800). In a specific embodiment, the storage (806) is a non-volatile, solid-state memory. In a specific embodiment, the storage (806) includes a read-only memory (ROM). Where appropriate, such a ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically modifiable ROM (EAROM), flash memory, or a combination of two or more of these. The present specification considers a mass storage (806) taking any suitable physical form. Where appropriate, the storage (806) may include one or more storage control units that facilitate communication between the processor (802) and the storage (806). Where appropriate, the storage (806) may include one or more storages (806). Although the present specification describes and illustrates a specific storage, the present specification considers any suitable storage.

[0377] In a specific embodiment, the I / O interface (808) comprises hardware, software, or both, providing one or more interfaces for communication between a computer system (800) and one or more I / O devices. Where appropriate, the computer system (800) may include one or more of these I / O devices. One or more of these I / O devices may enable communication between a person and the computer system (800). For example, without limitation, the I / O devices may include a keyboard, keypad, microphone, monitor, mouse, printer, scanner, speaker, still camera, stylus, tablet, touch screen, trackball, video camera, another suitable I / O device, or a combination of two or more of these. The I / O devices may include one or more sensors. The present specification considers any suitable I / O device and a suitable I / O interface (808) for it. Where appropriate, the I / O interface (808) may include one or more device or software drivers that enable the processor (802) to drive one or more of these I / O devices. Where appropriate, the I / O interface (808) may include one or more I / O interfaces (808). Although this specification describes and illustrates specific I / O interfaces, this specification considers any suitable I / O interface.

[0378] In a specific embodiment, the communication interface (810) comprises hardware, software, or all of the same that provide one or more interfaces for communication (e.g., packet-based communication) between the computer system (800) and one or more other computer systems (800) or one or more networks. For example, without limitation, the communication interface (810) may include a network interface controller (NIC) or network adapter for communicating with Ethernet or other wired-based networks, or a wireless NIC (WNIC) or wireless adapter for communicating with a wireless network, such as a Wi-Fi network. The present specification considers any suitable network and any suitable communication interface (810) for such a network. For example, without limitation, the computer system (800) may communicate with an ad hoc network, a personal area network (PAN), a local area network (LAN), a wide area network (WAN), a metropolitan network (MAN), one or more parts of the Internet, or a combination of two or more of these networks. One or more parts of one or more of these networks may be wired or wireless. For example, the computer system (800) may communicate with a wireless PAN (WPAN) (e.g., Bluetooth WPAN), a Wi-Fi network, a Wi-MAX network, a cellular phone network (e.g., a GSM (Global System for Mobile Communication) network), other suitable wireless networks, or a combination of two or more of these networks. Where appropriate, the computer system (800) may include any suitable communication interface (810) for any of these networks. Where appropriate, the communication interface (810) may include one or more communication interfaces (810). Although this specification describes and illustrates specific communication interfaces, this specification considers any suitable communication interface.

[0379] In a specific embodiment, the bus (812) includes hardware, software, or both that connect components of the computer system (800) to one another. For example, without limitation, the bus (812) may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a front-side bus (FSB), a Hypertransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infiniband interconnect, a low-pin-count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCIe (PCI-Express) bus, a serial advanced technology attachment (SATA) bus, a Video Electronics Standard Association local (VLB) bus, another suitable bus, or a combination of two or more such buses. Where appropriate, the bus (812) may include one or more buses (812). Although this specification describes and illustrates specific buses, this specification considers any suitable bus or interconnect.

[0381] The present invention described above has been explained with reference to the embodiments illustrated in the drawings, but this is merely illustrative, and those skilled in the art will understand that various modifications and variations of the embodiments are possible therefrom. However, such modifications should be considered to be within the technical scope of protection of the present invention. Accordingly, the true technical scope of protection of the present invention should be determined by the technical spirit of the appended claims.

Claims

Claim 1 A method for generating an outdoor advertisement that satisfies preset constraints that the outdoor advertisement must comply with, based on a generative AI model and reflects the requirements of an advertiser, performed by a computing device, comprising: an advertiser’s basic advertising data for an advertisement describing the basic content that the advertiser wishes to reflect in the advertisement; and a step of obtaining an advertiser’s advertising display condition, which indicates an area on a target vehicle where the advertiser wishes to display the advertisement, wherein the advertising display condition includes a display surface of the target vehicle where the outdoor advertisement is to be displayed, and the display surface is a single surface or two or more consecutive surfaces of the target vehicle; a step of obtaining a vehicle image taken of the target vehicle, analyzing the vehicle image through pre-trained deep learning models to detect a window area of ​​the target vehicle, and detecting a reference area of ​​the target vehicle based on the detection result of the window area, wherein the reference area of ​​the target vehicle is a part of the vehicle area of ​​the target vehicle for determining whether the advertiser’s outdoor advertisement satisfies the area constraint conditions of a preset guide. A step of generating the outdoor advertisement for the target vehicle that reflects the requirements of the advertiser by applying the above basic advertising data and the above advertising display conditions to a pre-trained advertising generation model; a step of obtaining area information of the display area of ​​the outdoor advertisement according to user input of the advertiser indicating the size or area of ​​the outdoor advertisement; a step of determining the display location of the outdoor advertisement according to the display area information according to the user input of the advertiser; a step of verifying whether the outdoor advertisement having the obtained display area and the determined display location satisfies the guide; and a step of providing the outdoor advertisement for the target vehicle that satisfies the area limit conditions of the guide to the advertiser.The method includes, wherein the advertising material generation model is a generative AI model pre-trained to generate an image that visually implements the information of the input data from the input data, receiving basic data or prompt data in the form of text as the input data, encoding the input data into a vector of a latent space, and sequentially generating and combining visual features based on the encoded vector to generate an image that reflects the intent and constraints of the input data as output data, wherein, the latent space refers to a multidimensional feature space in which the semantic features of the input data are quantified and expressed, and the step of verifying whether an outdoor advertising material having the acquired display area and the determined display location satisfies the guide; is a step of verifying whether the outdoor advertising material according to the display area information satisfies the area limit conditions of the guide, comprising: a step of calculating the area of ​​the reference area in the target vehicle; a step of calculating the ratio of the area of ​​the acquired outdoor advertising material to the ratio of the calculated area; and a step of analyzing whether the ratio of the area of ​​the outdoor advertising material to the ratio of the area exceeds a threshold ratio corresponding to a preset area limit condition. A step of determining that the outdoor advertisement for the target vehicle does not satisfy the area restriction condition if the ratio of the area of ​​the outdoor advertisement to the ratio of the above area exceeds a threshold ratio corresponding to a preset area restriction condition; a step of determining that the advertisement for the target vehicle satisfies the area restriction condition if the ratio of the area of ​​the outdoor advertisement to the ratio of the above area has an area less than or equal to the threshold ratio corresponding to a preset area restriction condition; wherein the step of determining whether the area ratio with respect to the area of ​​the reference area is less than or equal to the threshold ratio corresponding to a preset area restriction condition is to determine whether the following mathematical formula is satisfied, [mathematical formula]; Here, R thr is a preset threshold ratio, A1 represents the area of ​​the reference region, and A2 represents the area of ​​the outdoor advertisement for the target bus; the step of detecting the reference region of the target vehicle in the target vehicle comprises: inputting the vehicle image into a pre-trained vehicle recognition model to recognize the target vehicle in the vehicle image; inputting a first sub-image including the recognized target vehicle into a pre-trained vehicle region detection model to detect the vehicle region in which the recognized target vehicle is represented in the selected vehicle image; and inputting a second sub-image representing the vehicle region within the selected vehicle image into a pre-trained window region detection model to detect the window region of the target vehicle located on the vehicle surface in a direction satisfying the positional restriction condition of a preset guide. A method comprising: a step of determining, based on a detection result for a window area of ​​the target vehicle, the remaining area excluding the window area in the detected vehicle area within the vehicle image as a reference area for determining the area limit condition; wherein the window area detection model comprises: an ROI detection unit that extracts shape features from an input image and identifies the location of a pre-set part of interest of the vehicle based on the extracted shape features—wherein the part of interest is a component of the vehicle that provides a relative positional relationship capable of calculating the location of the window part—a color feature extraction unit that extracts color features pixel by pixel from the input image; and a window detection unit configured to identify a window area in which a transparent window part is expressed in the vehicle area based on the location of the identified part of interest and the extraction result of color features. Claim 2 delete Claim 3 In claim 1, the input data for the window detection unit is a combination of the location of the identified part of interest and the result of extracting color features, wherein the location of the identified part of interest corresponds to an image marked on a color distribution map, and the color distribution map is a distribution map representing the distribution of each color based on the shape of the bus exterior area, and the window detection unit includes a pre-trained artificial neural network model to infer a first correlation between the location of the part of interest and the location of the window area based on the location of the part of interest, and infer a second correlation between the color distribution map and the location of the window area, and to calculate the location of the window area in the bus exterior area corresponding to the input data based on the inferred first correlation and second correlation, and the window area detection model is trained to minimize the following loss function, [Mathematical Formula] Here, L represents the loss function of the second region detection model, and E ROI position , E window position represents the predicted location of the region of interest and the predicted location of the window region, respectively, calculated by the ROI detector and the window detector from the same sample training image, and T ROI position , T window position A method characterized by representing the actual location of the region of interest and the actual location of the window region, respectively, in the same sample training image. Claim 4 A method according to claim 1, wherein the advertising basic data comprises one or more of advertising theme information, source image, advertising text information, and advertising object information, and the advertising theme information comprises an advertising material type, and the advertising material type is an integrated type that displays a single advertising material on a display surface or a distributed type in which two or more advertising materials are displayed dispersedly on the display surface. Claim 5 A method according to claim 4, wherein the step of calculating the ratio of the area of ​​the obtained outdoor advertisement to the ratio of the calculated area comprises: a step of summing the area of ​​the outdoor advertisement on the display surface of the outdoor advertisement when the outdoor advertisement is of the dispersed type; and a step of calculating the ratio of the summed area of ​​the outdoor advertisement to the ratio of the calculated area. Claim 6 The method according to claim 1, wherein the advertising material generation model comprises a prompt layer that converts input data of the advertising material generation model into prompt data according to a prompt format—the prompt layer is configured to understand the context and meaning of the input data based on a transformer structure and convert it into a prompt format—and an image generation layer configured to receive prompt data reflecting basic advertising content and advertiser display conditions to generate the outdoor advertising material, and to generate an image reflecting the advertiser's intentions, requirements, and characteristics of the vehicle advertisement. Claim 7 A method according to claim 1, wherein the step of verifying whether an outdoor advertisement having the acquired display area and the determined display location satisfies the guide comprises: a step of verifying whether the outdoor advertisement according to the display location satisfies the area limitation condition of the guide, wherein the outdoor advertisement is identified in an area where the display of the outdoor advertisement is limited in the target vehicle; a step of analyzing whether the determined display location of the outdoor advertisement overlaps with the identified limitation area; a step of determining that the outdoor advertisement for the target vehicle does not satisfy the location limitation condition of the guide if the display location of the outdoor advertisement overlaps with the limitation area; a step of readjusting the display location of the outdoor advertisement so that the outdoor advertisement for the target vehicle satisfies the location limitation condition; and a step of determining that the outdoor advertisement for the target vehicle satisfies the location limitation condition if the display location of the outdoor advertisement does not overlap with the limitation area, wherein the limitation area is the window area. Claim 8 The method according to claim 1, wherein the step of obtaining area information for the outdoor advertisement comprises: a step of providing initial area information according to a preset area rule to an advertiser; and a step of determining area information according to user input entered thereafter; wherein the preset area rule is the maximum allowable area capable of satisfying the area limit conditions of the preset guide. Claim 9 A method according to claim 1, characterized in that the threshold ratio is 0.

5. Claim 10 A computer-readable recording medium having a program that enables a method of generating an outdoor advertisement based on a generative AI model that satisfies the pre-set limiting conditions that an outdoor advertisement must comply with according to any one of claims 1, 3 through 9, and reflects the requirements of an advertiser. Claim 11 A system comprising a processor and memory that generates an outdoor advertisement reflecting the requirements of an advertiser based on a generative AI model, satisfying preset constraints that the outdoor advertisement must comply with, wherein the system acquires a vehicle image of a target vehicle, analyzes the vehicle image through pre-trained deep learning models to detect a window area of ​​the target vehicle, and detects a reference area of ​​the target vehicle based on the detection result of the window area - the reference area of ​​the target vehicle is a part of the vehicle area of ​​the target vehicle for determining whether the advertiser's outdoor advertisement satisfies the area constraint conditions of a preset guide -; Advertiser's basic advertising data describing the basic content the advertiser wishes to reflect in the advertisement; and an advertising generation unit that, upon obtaining the advertiser's advertising display conditions, which indicate an area where the advertiser wishes to display an advertisement on the target vehicle, applies the basic advertising data and the advertising display conditions to a pre-trained advertising generation model to generate the outdoor advertisement for the target vehicle that reflects the advertiser's requirements—wherein the advertising display conditions include a display surface of the target vehicle where the outdoor advertisement is desired to be displayed, and the display surface is a single surface or two or more consecutive surfaces of the target vehicle—, area information of the display area of ​​the outdoor advertisement obtained according to the advertiser's user input indicating the size or area of ​​the outdoor advertisement, and a guide verification unit that verifies whether the outdoor advertisement has a display location of the outdoor advertisement determined according to the advertiser's user input satisfies the guide;The advertising material generation model includes, wherein the advertising material generation model is a generative AI model pre-trained to generate an image that visually implements the information of the input data from the input data, receiving basic data or prompt data in the form of text as the input data, encoding the input data into a vector in a latent space, and sequentially generating and combining visual features based on the encoded vector to generate an image that reflects the intent and constraints of the input data as output data, wherein the latent space refers to a multi-dimensional feature space in which the semantic features of the input data are quantified and expressed, and the guide verification unit verifies whether the outdoor advertising material having the acquired display area and the determined display location satisfies the guide by verifying whether the outdoor advertising material according to the display area information satisfies the area limit conditions of the guide, by calculating the area of ​​the reference area in the target vehicle, calculating the ratio of the area of ​​the acquired outdoor advertising material to the ratio of the calculated area, analyzing whether the ratio of the area of ​​the outdoor advertising material to the ratio of the area exceeds a threshold ratio corresponding to a preset area limit condition, and regarding the ratio of the area The method is configured such that if the ratio of the area of ​​the outdoor advertisement to the above-mentioned area exceeds a threshold ratio corresponding to a preset area limit condition, it is determined that the outdoor advertisement for the above-mentioned target vehicle does not satisfy the area limit condition, and if the ratio of the area of ​​the outdoor advertisement to the ratio of the above-mentioned area has an area less than or equal to the threshold ratio corresponding to the preset area limit condition, it is determined that the advertisement for the above-mentioned target vehicle satisfies the area limit condition, and the step of determining whether the area ratio with respect to the area of ​​the above-mentioned reference area is less than or equal to the threshold ratio corresponding to the preset area limit condition is to determine whether the following mathematical formula is satisfied, [mathematical formula]; Here, R thr ... is a preset threshold ratio, A1 represents the area of ​​the reference region, and A2 represents the area of ​​the outdoor advertisement for the target bus; the reference region analysis unit is configured to input the vehicle image into a pre-trained vehicle recognition model to recognize the target vehicle in the vehicle image in order to detect the reference region of the target vehicle in the target vehicle, input a first sub-image including the recognized target vehicle into a pre-trained vehicle region detection model to detect the vehicle region in which the recognized target vehicle is expressed in the selected vehicle image, input a second sub-image representing the vehicle region within the selected vehicle image into a pre-trained window region detection model to detect the window region of the target vehicle located on the vehicle surface in a direction satisfying the positional restriction condition of a preset guide, and determine the remaining area excluding the window region from the vehicle region detected within the vehicle image as the reference region for determining the area restriction condition based on the detection result for the window region of the target vehicle, and the window region detection model is configured to extract shape features from the input image and to identify the location of a preset part of interest of the vehicle based on the extracted shape features - the part of interest can calculate the location of the window portion A system comprising: a component of a vehicle that provides a relative positional relationship; a color feature extraction unit that extracts color features pixel by pixel from an input image; and a window detection unit configured to identify a window area in which a transparent window portion is expressed in the vehicle area based on the position of the identified part of interest and the result of extracting color features.

Images