Geolocation-based background generation for object images

Abstract

The invention relates to geo-location-based background generation for an image of an object. Systems and methods are provided for generating a geographic location-based image for a target object. A geo-location module receives a set of geo-locations associated with a geographic region of interest. Each geographic location in the set of geographic locations is mapped to contextual data associated with the geographic location. A cue generation module generates a plurality of cues based on the set of geographic locations and contextual data. The cue generation module includes a first generative artificial intelligence (AI) model. The image generation module generates a plurality of composite images based on the plurality of cues. The image generation module includes a second generative AI model. Each composite image depicts the target object in a background generated based on the cue.

Description

Technical Field

[0001] The present disclosure generally relates to generative artificial intelligence (AI). More specifically, but not by way of limitation, the present disclosure relates to geo-location-based background generation for object images using generative AI. Background Art

[0002] Advances in digital technology and platforms allow users to perform various actions, such as attending virtual exhibitions or interacting, from anywhere in digital space. The object images provided on these online platforms are a key part of how online platforms communicate with their users and showcase their offerings regarding objects. Online platforms strive to provide visually appealing object images that resonate with users' preferences, identities, and interests, thereby increasing user engagement and interaction with online platforms. Summary of the invention

[0003] Certain embodiments relate to geographic location-based background generation for object images using generative AI. In one example, an image generation server receives a set of geographic locations associated with a geographic area of ​​interest. Each geographic location in the set of geographic locations is mapped to context data associated with the geographic location. The image generation server generates multiple prompts based on the set of geographic locations and the context data using a first generative AI model. The image generation server generates a background image set based on multiple prompts using a second generative AI model. The image generation server generates multiple composite images. Each composite image includes a background image from the set of background images and an image of a target object.

[0004] These illustrative embodiments are not mentioned to limit or define the present disclosure, but to provide examples to aid understanding of the present disclosure.Additional embodiments are discussed in the Detailed Description, and further description is provided there. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] The features, embodiments, and advantages of the present disclosure may be better understood when the following detailed description is read with reference to the accompanying drawings.

[0006] Figure 1 An example of a computing environment is depicted in which an image generation server generates a composite image with a geographic location-based background for a target object in accordance with certain embodiments of the present disclosure.

[0007] Figure 2 An example of a process for automatically generating a composite image of a target object with a geographically-relevant background is depicted in accordance with certain embodiments of the present disclosure.

[0008] Figure 3 is an example workflow for generating a composite image for a marketplace platform based on geographic location according to certain embodiments of the present disclosure.

[0009] Figure 4 is another example workflow for generating a composite image for a marketplace platform based on geographic location in accordance with certain embodiments of the present disclosure.

[0010] Figure 5 is an example workflow for generating optimal prompts for an image generation AI model for generating composite images for a marketplace platform based on geographic location, according to certain embodiments of the present disclosure.

[0011] Figure 6 Depicted is an example of a GUI for picking an existing object image to generate a new composite image with a different background, in accordance with certain embodiments of the present disclosure.

[0012] Figure 7 Depicted is an example of a GUI for providing descriptive hints for background generation for a new composite image to be generated, in accordance with certain embodiments of the present disclosure.

[0013] Figure 8 Describes a display based on some embodiments of the present disclosure Figure 7 Example of a GUI for generating new composite images based on the prompts provided in .

[0014] Fig. 9 Depicted is an example of a GUI for selecting multiple products using various filters to generate a new composite image with a background based on a geographic location, in accordance with certain embodiments of the present disclosure.

[0015] Fig.10 Described is a method for providing a method for providing a Fig. 9 An example of a GUI for generating a new composite image by selecting multiple products in the GUI shown in FIG.

[0016] Fig.11 Depicted is another example of a GUI for providing descriptive hints for background generation for a new composite image to be generated, in accordance with certain embodiments of the present disclosure.

[0017] Fig.12 Describes a display based on some embodiments of the present disclosure Fig.11 Example of a GUI for generating new composite images based on the prompts provided in .

[0018] Fig.13 Depicted is an example of a GUI displaying a composite image with a geographic location-based background on a marketplace platform in accordance with certain embodiments of the present disclosure.

[0019] Fig.14Depicted are examples of an object image having a white background and a corresponding composite image generated using a background based on a geographic location, in accordance with certain embodiments of the present disclosure.

[0020] Fig.15 Depicted is an example of a GUI displaying a composite image with a geo-location-based model's face deployed on a marketplace platform, in accordance with certain embodiments of the present disclosure.

[0021] Fig.16 Another example of a GUI displaying a composite image with a geo-location-based model's face deployed on a marketplace platform is depicted in accordance with certain embodiments of the present disclosure.

[0022] Fig.17 Depicted is an example of a GUI displaying an original object image with a white background on a marketplace platform in accordance with certain embodiments of the present disclosure.

[0023] Fig.18 Depicted is an example of displaying a composite image GUI with a geo-location-based background and a geo-location-based model's face deployed on a marketplace platform for a specific geographic market, in accordance with certain embodiments of the present disclosure.

[0024] Fig.19 An example of a computing system for implementing certain embodiments of the present disclosure is depicted. DETAILED DESCRIPTION

[0025] The present disclosure provides a technique for using generative AI to generate geo-location-based backgrounds for object images. Most object images on online platforms are very simple, like objects against a plain white background. Some object images are taken at various locations to feature objects against visually appealing backgrounds to capture the user's attention. The process of traveling to different locations to take photos is costly, non-scalable, and may not even resonate with users in different regions. Some object images are taken with human models or props (e.g., mannequins) on flat backgrounds, and different background images are taken separately. Then, in a photo editor (e.g., Adobe Photoshop), the background image is automatically generated based on the location of the object. ). This is less expensive, but still requires manual processing of each object image. Therefore, maintaining multiple variations of the same object image for users from different regions remains a problem.

[0026] In an embodiment of the present disclosure, an image generation server may use generative AI to generate a geographic location-based background for an object image. The object image shows the object against a flat (e.g., white) background. The image generation server then combines each generated background with a copy of the object image to create a composite image, each composite image showing the object against a different generated background.

[0027] In one example, an operator of the online platform selects at least one object image and at least one geographic region of interest. The operator may select the geographic region by interacting with a map or other suitable input component. The online platform provides information related to the (multiple) geographic regions, such as (multiple) names or coordinates, to the image generation server. The image generation server accesses a data repository to retrieve certain contextual information associated with the (multiple) geographic locations, such as places of interest, landmarks, seasons, weather, terrain, events, population composition, etc.

[0028] The image generation server uses a text generation AI model to generate multiple prompts based on the geographic location and associated context information. The prompt may include certain context data already associated with the geographic location and / or any new context data or other information output by the text generation AI model. Then, the image generation server inputs one or more of the generated prompts into the image generation AI model to generate a composite image of the target object and the background based on one or more of the generated prompts and the image of the target object. The background is generated based on one or more prompts in the generated prompts. As an example, the composite image may show the target object in the context of depicting certain landmarks or other context features associated with the selected geographic location. The composite image may be displayed in a graphical user interface (GUI) so that the operator of the online platform can select the most desired composite image. The selected composite image may be deployed on an online platform, which in some examples may be an online directory or include an online directory.

[0029] Contextual data associated with a geographic location input to a generative AI model can be weighted and re-weighted in a dynamic manner based on various factors, such as behavioral data collected in response to a user's interaction with a deployed synthetic image. User behavioral data may include information such as how long a user interacts with a synthetic image and whether the user further engages with an online platform for an object depicted in the synthetic image, for example, by further conducting a search for a target object or purchasing a target object. In some examples, a reinforcement learning model is implemented to learn user behavioral data associated with a deployed synthetic image. By weighting and re-weighting the contextual data, a text generation AI model can generate optimal prompts for users on an online platform in a specific geographic area.

[0030] Certain embodiments of the present disclosure overcome the shortcomings of the prior art by using generative AI to generate geographic location-based backgrounds for object images. The process used uses a text generative AI model to automatically generate prompts using weighted context data associated with the geographic location. Different backgrounds are generated based on prompts using an image generative AI model. Objects provided by the online platform can be depicted using images of different backgrounds associated with the geographic location. Therefore, the proposed image generation server enables the online platform to expand object image generation by deploying a cost-effective generative AI model, thereby generating a wide array of images starting from a single object image. The operator of the online platform does not need to go to different locations to take pictures or manually process images of different backgrounds. At the same time, the proposed image generation server enables the online platform to provide users with personalized experiences using different backgrounds depending on the user's geographic location. In turn, users with personalized experiences on the online platform can also improve user engagement, user conversion rate, or other performance indicators of the online platform.

[0031] As used herein, the term "geographic region" or "geographic area of ​​interest" refers to an area on the Earth. The online platform may describe a geographic region by its name or boundaries in a map. For example, the geographic region is Argentina.

[0032] As used herein, the term "geographic location", "geographic location set" or "geographic location data" refers to one or more locations within a geographic area. One or more locations can be represented by coordinates. For example, within the geographic area of ​​Argentina, there are multiple locations.

[0033] As used herein, the term "object" or "target object" is used to refer to something visible or perceptible provided by an online platform. An example of an object may be a product provided on a marketplace platform.

[0034] As used herein, the term "object image" is used to refer to an image or picture featuring an object. The object image may have a white background or a different background with a different color or scene. For example, the object image may be a product image featuring a product on a marketplace platform.

[0035] As used herein, the term "background" is used to refer to the overall scene behind the main object in the image. For example, the background of the target image is just white. For another example, the background of the target image is some scenery, such as a forest.

[0036] As used herein, the term "background image" is used to refer to an image that serves as an overall scene that may feature an object. The background image may be generated by an image-generating AI model based on prompts that describe the overall scene. The background image may be combined with an object image into a new image such that the background image provides a background for the object in the new image.

[0037] As used herein, the term "facial image" is used to refer to an image of a human face. A facial image may be generated by an image-generating AI model based on prompts describing a human face. A facial image may feature the face of a male in a particular geographic location (e.g., China or India).

[0038] As used herein, the term "synthetic image" is used to refer to an image generated by an image-generating AI model or certain image editing tools. The synthetic image features an object extracted from an object image with a unique background provided by a background image or on a model whose face is provided by a facial image.

[0039] As used herein, the term "deployed images" is used to refer to certain composite images selected by an operator of the online platform to be displayed on the online platform or automatically transmitted to the online platform by an image generation server and used by the online platform. For example, the image generation server generates multiple composite images, and the operator of the online platform selects a subset of the multiple composite images to be used on the online platform. The subset of the multiple composite images is a deployed image.

[0040] As used herein, the terms "generative artificial intelligence," "generative AI," or "generative AI models" are used to refer to types of AI systems that are capable of generating text, images, or other media in response to prompts. Examples of text-generating AI systems include the Generative Pretrained Transformer (GPT), the Text-to-Text Transfer Transformer (T5), the Open Pretrained Transformer (OPT), the Bidirectional Autoregressive Transformer (BART), the Language Model for Conversational Applications (LaMDA), or any variant. Examples of image or art-generating AI systems include Stable Diffusion, Midjourney, and DALL-E.

[0041] As used herein, the term "prompt" is used to refer to text or information that is provided to an AI model as an input or "command" to generate its output. An example prompt for generating a composite image of an object is "Woman camping in the California wilderness". The prompt may specify which part of the image needs to be generated, e.g., background, face. Alternatively, the prompt is a description of the final image that the operator wants, and the AI ​​model interprets which parts of the object image need to be regenerated.

[0042] As used herein, the terms "mask", "background mask" or "face mask" are used to refer to a type of filter that operates directly on an image. The filter can be in the form of a grayscale image or a binary image. For example, a background mask can be overlaid on an image to remove background pixels in the image.

[0043] Now refer to the accompanying drawings, Figure 1 An example of a computing environment 100 according to certain embodiments of the present disclosure is depicted, wherein an image generation server 102 generates a composite image 110 for a target object with a background based on a geographic location. In various embodiments, the computing environment 100 includes an image generation server 102, which is connected to an online platform 124 and user computing devices 130A, 130B, and 130C (which may be individually referred to as user computing devices 130 or collectively referred to as user computing devices 130 herein) via a network 132. The user computing device 130 may be a desktop computer, a laptop computer, a tablet computer, a smart phone, or any suitable computing device. The network 132 may be a local area network ("LAN"), a wide area network ("WAN"), the Internet, or any other network topology known in the art that connects the user computing device 130 to the online platform 124.

[0044] The online platform 124 includes a platform server 126 and a database 128. Examples of the online platform 124 may include a market platform that provides certain products or services, or may include any other suitable online platform that uses images to showcase the objects provided. The database 128 stores object images 120 and deployed images 134. The object image 120 includes an image of a target object on a white background or any other flat background. The object image 120 can be taken using a white or other flat background, or can be created by applying image editing techniques to remove a non-flat background. The deployed image 134 includes at least some composite images 110 that depict a target object with a geographically relevant background generated by the image generation server 102. The deployed image 134 is displayed on the online platform 124 for viewing by users of the online platform 124, for example, via a user computing device 130.

[0045] The image generation server 102 includes a geographic location module 104, a prompt generation module 106, an image generation module 108, and a data repository 112. The geographic location module 104 is configured to receive information about one or more geographic areas of interest. The information may take the form of a geographic area name or geographic location coordinates or other identifiers. An operator of the online platform 124 may define or select a geographic area of ​​interest via a GUI such as an interactive map. The GUI may be provided by the image generation server 102 or may be provided by the online platform 124. In other words, in some embodiments, the operator may directly access the image generation server 102 through an interface provided by the image generation server 102, or indirectly access the image generation server 102 via an interface provided by the online platform 124. In yet other embodiments, the operator may access the online platform 124 and / or the image generation server 102 using a user computing device 130.

[0046] In some embodiments, the geolocation module 104 receives the name of a geographic area of ​​interest from an operator of the online platform 124. The geolocation module 104 matches the geographic area with the geolocation data 114 to identify a set of geographic locations within the boundaries of the geographic area of ​​interest. For example, the geographic locations within the boundaries of the geographic area of ​​interest may be sub-regions, such as states, counties, cities, towns, etc. The geolocation module 104 maps the geographic locations to contextual data 116 that is also stored in the data repository 112 or otherwise accessible to the image generation server 102. In some examples, a machine learning model is implemented to learn the contextual data 116 associated with the identified geographic location from a set of available data sources, such as information on the Internet. The mapped geolocation data 114 and contextual data 116 may be stored in separate databases or within the same database 136, for example as two or more tables in the database 136.

[0047] The selected geographic location data 114 and the associated contextual data 116 are input to the prompt generation module 106, which is configured to generate a plurality of prompts using a text generation AI model. The text generation AI model is trained to generate prompts based on the geographic location data 114 and the contextual data 116 associated with the selected set of geographic locations. In some examples, the prompts generated by the prompt generation module 106 are displayed to an operator of the online platform 124 so that the operator can select one or more prompts to be provided to the image generation module 108. In other embodiments, the generated prompts are directly input to the image generation module 108 without operator input.

[0048] In some examples, prompt generation module 106 implements a reinforcement learning model (e.g., a Markov decision process) to determine and assign weights to contextual data 116 based on various factors, including whether backgrounds generated using various prompts are included in synthetic images 110 selected by an operator for deployment to online platform 124 and user behavior data 138 associated with the synthetic images deployed to online platform 124. As will be appreciated by those skilled in the art, contextual data 116 with heavier weights can have a greater impact on the output of the text generation AI model used to generate prompts.

[0049] The prompts generated by the prompt generation module 106 are stored as prompt data 118 within the data repository 112 provided to the image generation module 108, and the image generation module 108 is configured to generate a background, a face, or any suitable item based on the prompts. The operator of the online platform 124 can select one or more prompts from the prompt data 118 for the image generation module 108 to generate a background image using the selected prompts. Alternatively, the prompts generated by the prompt generation module 106 are automatically provided to the image generation module 108 without intervention from the operator of the online platform 124. The image generation module 108 can implement an image generation AI model trained to generate a set of background images based on one or more prompts. In some examples, the image generation module 108 also generates a composite image 110 by combining each background image 122 with a copy of the object image 120. The background image 122 and the object image 120 can be combined by the same or different image generation AI modules used to generate the background image 122 or by any suitable image editing application. The background mask can be used to remove the background from the object image 120 to leave only the target object. The background image generated using the prompt is used as a background for the target object in the composite image 110. In some examples, the image generation module 108 generates a background for the object in the object image based on the selected prompt without generating a separate background image 122, thereby creating the composite image 110. The composite image 110 depicts the object in the background generated based on the selected prompt.

[0050] The composite images 110 are displayed to an operator of the online platform 124 via the GUI. The operator of the online platform 124 may select one or more composite images 110 to be deployed on the online platform 124. In response to the selection, the image generation module 108 then transmits the one or more selected composite images to the online platform 124 for deployment, which are stored in the database 128 of the online platform 124 as deployed images 134.

[0051] In some embodiments, an operator of the online platform 124 constructs multiple geo-specific presentations of the composite image 110, such as a geo-specific catalog. The geo-specific presentation of the composite image 110 may be a subset of the image 134 deployed on the online platform 124. Examples may include different presentations of the composite image 110 for each country, each state, each city, etc. There is no limitation on the manner in which the operator can define the geographic region. Each geo-specific presentation of the composite image 110 includes a composite image 110 having a background generated based on contextual data associated with the geographic region and the geographic location data 114. The geo-specific presentation of the composite image 110 is stored in the data repository 112 of the image generation server 102, in the database 128 of the online platform 124, or in some other data repository accessible to the online platform 124. The composite image 110 is retrieved and displayed to users accessing the online platform 124 from the relevant geographic region. For example, the user's geographic region may be detected based on an IP address associated with the user computing device 130 or based on other markers such as customer profile information.

[0052] In other embodiments, an operator of the online platform 124 may select a number of composite images 110 covering multiple geographic areas and store the collection of composite images 110 in a data repository 112 of the image generation server 102, in a database 128 of the online platform 124, or in some other data repository accessible to the online platform 124. Then, when a user accesses the online platform 124 via a user computing device 130 or by other means, a geo-specific presentation of the composite images 110 may be generated and presented to the user in real-time or near real-time based on the user's geographic area.

[0053] In some examples, the image generation module 108 is configured to generate a collection of facial images 140 opposite to or in addition to the background image 122. The facial image is a facial image of a human model that can be added to the composite image. For example, the prompt generation module 106 can be used to generate prompts describing the characteristics of people living in the geographic area (e.g., country). Then, the image generation module 108 can use the prompts to generate the facial image 140, which can be combined with the object image 120 to produce the composite image 110. In some examples, the composite image includes a person wearing a target object (e.g., a piece of clothing). The composite image 110 may also include a background associated with the relevant geographic area. The image generation module 108 can implement an image generation AI model that is trained to generate a collection of facial images based on multiple prompts. The image generation AI model for facial images can be the same as the model used to generate the background image 122. Alternatively, the image generation AI model for facial images is a separate AI model. In some examples, the image generation module 108 is configured to generate the background image set 122 and the face image set based on a plurality of cues and generate a composite image by combining the background image 122 and the face image with the object image 120 .

[0054] In some examples, the image generation module 108 is configured to generate a composite image by generating a background and / or a face in the object image based on a selected prompt, without generating a separate background image or a face image. The selected prompt can fully describe the composite image. For example, the selected prompt is "young people on the streets of Paris". The image generation module 108 can implement a generative AI model that can be configured to interpret the selected prompt to generate a background or a face or both based on the selected prompt. For this example prompt, the generative AI model can generate a background depicting a Paris street and a face of a young Parisian, and merge it into the object image.

[0055] Figure 2An example of a process 200 for automatically generating a composite image 110 with a geographically relevant background for a target object according to certain embodiments of the present disclosure is depicted. At block 202, the geographical location module 104 determines a set of geographical locations associated with a geographical area of ​​interest. Each geographical location in the set of geographical locations is mapped to contextual data 116 associated with the geographical location. The image generation server 102 may provide a GUI to an operator of the online platform 124. In some embodiments, the geographical location module 104 of the image generation server 102 may provide a map via the GUI for the operator to select a geographical area of ​​interest, for example, by drawing the boundaries of the area on the map. Alternatively or additionally, the operator may type, select, or otherwise enter text describing the geographical area of ​​interest (e.g., Argentina) via the GUI. The geographical location module 104 receives geographical location information associated with the geographical area of ​​interest, such as a boundary or a place name. In some examples, the data repository 112 of the image generation server 102 stores predefined contextual data 116 for various geographical locations. The predefined contextual data 116 may be specified by a subject matter expert or pre-learned by a machine learning model. The contextual data 116 may include places of interest, landmarks, terrain, climate, weather, events, population composition, income level, urbanization level, culture, etc. In some examples, the geolocation module 104 includes a machine learning model trained to dynamically extract contextual data 116 associated with a geographic location identified by the online platform 124. For example, the geolocation module 104 may implement a pre-trained large language model (LLM) via an application programming interface (API) to provide contextual data 116 for a geographic location identified by the online platform 124. When a set of geographic locations associated with a geographic area of ​​interest is received from the online platform 124, the geolocation module 104 may provide instructions to the pre-trained large language model (LLM). The instructions may include certain geolocation data 114 associated with the geographic area of ​​interest. For example, the instruction is "provide a description of Argentina, including its trendy cities, landmarks, places of interest, terrain, popular events, weather, climate, and other interesting information". Then, the pre-trained LLM provides a summary of the contextual data 116 for the identified geographic location.

[0056] At block 204, the prompt generation module 106 generates a plurality of prompts based on the set of geographic locations and the contextual data 116 using a first generative artificial intelligence (AI) model. The plurality of prompts are part of the prompt data 118. The prompt generation module 106 implements a text generative AI model for generating prompts for another generative AI model used at block 206 below to generate an image. An example prompt is "Generate an image of Quebrada de Humahuaca, Argentina". The text generative AI model can be a pre-trained LLM, such as GPT, T5, OPT, BART, or a variation thereof. In some examples, the pre-trained LLM provides an API for the prompt generation module 106 to access the pre-trained LLM.

[0057] In some examples, the prompt generation module 106 weights different types of contextual data 116 associated with a geographic location so that contextual data 116 with heavier weights is emphasized more strongly by the text generation AI model, thereby generating the best prompt. Alternatively, the prompt generation module 106 may weight prompts previously generated by the prompt generation module 106 or prompts specified by the operator with the type of contextual data, and select the best prompt for generating an image. For example, if a composite image 110 generated using prompt A from the prompt data 118 is selected by an operator of the online platform 124, prompt A may be weighted more heavily than other prompts. The prompt generation module 106 may provide prompt A to the image generation module 108 more frequently than other prompts. In addition, if the composite image 110 generated using prompt A and selected by the operator to be used on the online platform 124 attracts more online visitors and converts more online visitors into customers, prompt A may be provided to the image generation module 108 more frequently than other prompts for image generation. The prompt generation module 106 can implement a reinforcement learning model (e.g., a Markov decision process) to learn the weighting of different types of contextual data 116 and the usefulness / success of the generated text prompt based on whether the background or background image 122 generated according to the prompt is included in the synthetic image selected to be deployed on the online platform 124. The weighting can also be based on or alternatively based on user behavior data 138 associated with user interactions with one or more deployed images 134, including user engagement (e.g., whether the deployed image 134 generates higher engagement with the product), user conversion (e.g., whether the used image leads the user to purchase the product), or other performance indicators collected or received by the online platform 124.

[0058] At block 206, the image generation module 108 generates a set of background images 122 based on a plurality of prompts using a second generative AI model. The image generation module 108 implements an image generative AI model for generating one or more background images 122 based on the prompt generated at block 204. In some examples, the image generation module 108 also generates a set of facial images 140 based on the prompt generated at block 204. An example of a facial prompt is "generate Parisian man face". The image generation module 108 may implement an image generative AI model for background generation or a different image generative AI model to generate facial images 140 based on facial prompts. The (multiple) image generative AI models used to generate the background images 122 and the facial images 140 may be pre-trained text-to-image AI models, such as stable diffusion, DALL-E, or variations thereof. The (multiple) pre-trained text-to-image AI models may provide an API for the image generation module 108 to access the pre-trained text-to-image AI models. However, generating a separate background image or a separate face image at block 206 may be optional. The image generation module 108 may directly use the object image and the hint to generate a composite image without generating a separate background image or face image, as will be described below at block 208.

[0059] At block 208, the image generation module 108 generates a plurality of composite images 110. In some examples, each composite image 110 is generated based on a background image 122 and an object image. The image generation module 108 uses image editing techniques to composite the background image 122 and the object image 120. The generative AI model implemented by the image generation module 108 at block 206 also takes the object image 120 as input to generate a composite image 110 that shows a target object on a background provided by the background image 122. The object image 120 can be an image showing a target object on a white background or other flat background. A background mask is also provided to the image generation module 108 as an input to the image generation module 108. The background mask can be generated by an image editing tool (e.g., Adobe Photoshop). ) or accessed via an API of an image editing tool. The background mask identifies and removes the background of the object image 120. Thus, the background image 122 can be added as the background of the target object.

[0060] In some examples, the facial image 140 is also automatically generated or requested by an operator of the online platform 124. The facial image 140 and the facial mask are also used to generate the composite image 110. The facial mask removes the original face of the model in the object image 120. The facial image 140 is added to the face position to provide a face that is different from the original face. Therefore, the composite image 110 can include a background and a person's face corresponding to a specific geographic area of ​​the online platform 124. The online platform 124 can select and display certain composite images 110 as deployed images 134 for users located in a specific geographic area.

[0061] In some examples, the image generation module 108 directly generates a composite image by using an object image and a prompt, without generating a separate background image or face image based on a separate prompt and combining it with the object image. The image generation AI model implemented by the image generation module 108 is configured to interpret the prompt and generate a background or face or both based on the prompt. The prompt can describe the final image that the operation of the online platform wants. The image generation module 108 can automatically generate a background mask for the background in the object image and / or a face mask for the face in the object image. In some examples, the face in the object image can be protected from occlusion, and then the face can not be regenerated. An example prompt is "young people on the streets of Paris". The image generation AI model in this example generates a background depicting the streets of Paris and a face depicting the young people of Paris. If there is a face in the object image and the face in the object image is not protected, in addition to generating a background depicting the streets of Paris in the object image, the image generation AI model regenerates the face of the young people in Paris in the object image. The functionality included in blocks 206 and 208 may be used to implement steps for generating one or more composite images 110 based on the selected hint(s).

[0062] Figure 3 300 is an example workflow for generating a geo-location based object image for a market platform according to certain embodiments of the present disclosure. In this example, an operator of a market platform 302 provides a selection 304 of a geo-location market boundary on a map to a geo-location based catalog API 306 provided by an image generation server (e.g., image generation server 102). At the same time, the operator of the market platform 302 also provides keywords 308 for background generation in the geographic market defined by the selection 304 of the geo-location market boundary. The keywords 308 may be provided manually by the operator of the market platform. Alternatively or additionally, the keywords may be provided by the following Figure 5The keywords 308 are related to places of interest, landmarks, terrain, climate, weather, events, or other information associated with a geographic location within the selection 304 of geographic market boundaries. The keywords 308 act as prompts to the image generation AI model 310 for generating images. The operator of the marketplace platform 302 also provides one or more object images 312, each of which shows a specific product on a flat background. The object images 312 are provided as input to the image generation AI model 310. Image editing tools (e.g., ) to generate a background mask 314 for an object image 312. The background mask 314 is also provided to the image generation AI model 310. Alternatively, the object image 312 can be processed with the background mask 314 to become an object image 312 without a background, which is then provided to the image generation AI model 310. The generative AI model 310 uses the keywords 308, the object image 312, and the background mask 314 to generate a composite image 316. The composite image 316 includes an image of the object in the object image 312, but has a background generated based on the geographic location-related keywords 308. The composite image 316 is then provided to the geographic location-based catalog API 306 for selection and deployment on the market platform. Customers 318 from the geographic market can then view the composite image with the geographic location-based background generated specifically for the geographic market.

[0063] Figure 4 4 is another example workflow 400 for generating a composite image for a market platform based on a geographic location according to certain embodiments of the present disclosure. In this example, an operator of the market platform 402 provides keywords 404 for background generation and / or face generation. The keywords 404 may be provided manually by the operator of the market platform 402. Alternatively or additionally, the keywords 404 may be provided by the following Figure 5 The machine learning model illustrated in is generated. Keywords 404 associated with a geographic location within a predefined geographic market for the market platform may include places of interest, landmarks, terrain, climate, weather, events, or other information associated with a geographic location within a predefined geographic market. Keywords 404 act as prompts to the image generation AI model 406 for generating images. An operator of the market platform 402 also provides the image generation AI model 406 with an original image 408 of one or more items provided by the market platform with a flat background. Image editing tools (e.g., ) to generate a background mask 410 of the original image 408. A facial recognition tool (e.g., The face mask 412 is generated by using a facial recognition tool built on a computer (such as a computer program product). The background mask 410 and the face mask 412 can be combined into a combined mask 414. The combined mask 414 is also provided to the image generation AI model 406. The image generation AI model 406 uses the keywords 404, the original image 408, and the combined mask 414 to generate a geo-location-based image 416. The geo-location-based image 416 includes an image of the merchant in the original image 408, which has a background generated based on the geo-location-related keywords 404 and also has a model whose face is generated based on the geo-location-related keywords 404. The geo-location-based image 416 can be provided to the market platform for use in a search and purchase via a search and purchase transaction. Figure 3 A similar geo-location based catalog API 306 in is selected and deployed in specific geographic markets.

[0064] Figure 5 5 is an example workflow 500 for generating optimal prompts for an image generation AI model 504 for generating composite images for a marketplace platform based on a geographic location, according to certain embodiments of the present disclosure. In some examples, an AI-based prompt generator 502 generates multiple prompts for the image generation AI model 504 using contextual data 508 associated with the geographic location. In some examples, multiple prompts are provided to an operator of the marketplace platform 510. The operator of the marketplace platform can select one or more prompts from the multiple prompts as prompts to the image generation AI model 504. In either case, the image generation AI model 504 receives the prompts and generates an image background based on the prompts. The image generation AI model 504 also generates a composite image using a background generated using the prompts and an image of an object with a flat background, such as Figure 3 or Figure 4 Described in .

[0065] exist Figure 5In the embodiment of the present invention, a reinforcement learning model 506 is provided to learn the importance or weight of different types of contextual data based on customer behavior data associated with images generated using contextual data-based prompts and deployed on a marketplace platform 512. The importance or weight learned by the reinforcement learning model 506 is then provided to the AI-based prompt generator 502 to generate optimal prompts for the image generation AI model 504. For example, the reinforcement learning model 506 uses a Markov decision process to learn the weights of different contextual data and the usefulness / success of the generated text prompts based on the following conditions: (1) whether the marketplace platform ultimately selects a generated image associated with a certain generated prompt, and (2) whether the image used generates higher product engagement, higher conversion rate, or other key performance indicators (KPIs) measured as a function of customer behavior. The reinforcement learning model 506 monitors customer behavior on the marketplace platform 512 and collects customer behavior data to analyze the importance or weight of different contextual data represented by the generated images deployed on the marketplace platform 512.

[0066] Figure 6 An example of a GUI 600 for picking an existing object image for use in generating a new composite image with a different background according to certain embodiments of the present disclosure is depicted. The GUI 600 enables an operator of an online platform 124 to interact with an image generation server 102. In this example, the online platform 124 is a market platform and the object image 120 is a product image 602. The operator of the market platform can pick an existing product image 602 or upload a new image that does not yet exist on the market platform. Once the product image 602 is picked or uploaded, the operator can move to the next step by clicking or pressing a "next" button 604.

[0067] Figure 7 An example of a GUI 700 for providing a descriptive prompt for background generation for a new composite image to be generated according to certain embodiments of the present disclosure is depicted. The GUI 700 enables an operator to type their own prompt or select a prompt from suggested prompts 702 to populate in an input field 704. The suggested prompts 1102 are generated by the prompt generation module 106 on the image generation server 102. The product image 706 at this step shown in the GUI 700 is Figure 6 The product image 602 in , whose background has been removed from the previous step. When the operator clicks or presses the "Generate" button 708, a new composite image can be generated based on the descriptive prompt in the field 704.

[0068] Figure 8 Describes a display based on some embodiments of the present disclosure Figure 7An example of a GUI 800 for a new composite image generated based on the prompts provided in Figure 6 The product image 602 selected from the above is used to generate three new composite images 802, 804 and 806. Each new composite image 802, 804 and 806 includes a plurality of images based on the above Figure 7 The operator of the market platform may select one or more new composite images 802, 804, and 806 to be used on the market platform and displayed to customers in Arizona.

[0069] Figure 9-13 A process for mass-generating composite images for multiple products is shown. Fig. 9 An example of a GUI 900 for selecting a plurality of products using various filters to generate a new composite image 110 with a background based on a geographic location is depicted in accordance with certain embodiments of the present disclosure. An operator of a marketplace platform may select a plurality of products using various filters displayed in the GUI 900. In this example, the operator of the marketplace platform filters products by category and selects "Coats and Jackets" 902 for which a new composite image is to be generated. When the operator clicks or presses an "Apply Filters" button 904, a list of products that satisfy the filters may be displayed in the GUI 1000, as shown in FIG. Fig.10 as shown in .

[0070] Fig.10 Described is a method for providing a method for providing a Fig. 9 1000 for generating a new composite image for a plurality of products selected from the GUI shown in FIG. 1004. A partial list 1002 of filtered products is displayed in the GUI 1000. The operator of the marketplace platform can select one or more products for certain actions 1004. By selecting "generate new image" 1006 from the action 1004, an image generation process is triggered in the image generation server 102 for the product selected from the list of filtered products.

[0071] Fig.11 An example of a GUI 1100 for providing descriptive hints for background generation for a new composite image to be generated according to certain embodiments of the present disclosure is depicted. The GUI 1100 enables an operator to type their own hints or select hints from suggested hints 1102 to populate in field 1104. The suggested hints 1102 are generated by hint generation module 106 on image generation server 102. When the operator clicks or presses a "generate" button 1106, a new composite image may be generated based on the hint in field 1104.

[0072] Fig.12Describes a display based on some embodiments of the present disclosure Fig.11 1200 for generating a new composite image based on the prompts provided in FIG. Fig.12 Only the image generation module 108 for Fig.10 A subset of new composite images is generated based on the corresponding subset of products selected by the operator in . For example, images 1202, 1204, 1206, and 1208 are original product images with a white background. Images 1210, 1212, 1214, and 1216 are based on Fig.11 The operator of the market platform may select one or more new composite images to use on the market platform and display to customers in California.

[0073] Fig.13 An example of a GUI 1300 for displaying a composite image with a background based on a geographic location on a marketplace platform is depicted in accordance with certain embodiments of the present disclosure. The image generation module 108 is based on Fig.11 The prompt "Woman camping in the California wilderness" provided in generates composite images 1302-1316 for corresponding products. The composite images 1302-1316 are deployed on the market platform and displayed to customers in California.

[0074] Fig.14 An example of an object image with a white background and a corresponding composite image generated using a background based on a geographic location according to certain embodiments of the present disclosure is depicted. Object images 1402, 1404, and 1406 show different sofa groups in a white background. Composite images 1408, 1410, and 1412 show corresponding sofa groups with different interior backgrounds. The interior backgrounds in the composite images 1408, 1410, and 1412 can be generated by the image generation server 102 for a market platform for a specific geographic market. The geographic location module 104 of the image generation server 102 can collect context data related to a specific geographic market. The prompt generation module 106 can select specific context data 116 to generate prompts for the image generation module 108 for generating backgrounds for the object images 1402, 1404, and 1406. In the context data 116, demographics, income levels, and urbanization levels can be weighted more heavily than other context data (such as landmarks or events) used to generate the interior background. Based on demographics, occupations, income levels, urbanization levels, and other relevant contextual data related to the culture or demographics of the geographic market, the prompt generation module 106 can generate prompts that describe certain internal characteristics. The operator of the market platform can select from relevant Figure 7 GUI 700 or Fig.111404 and 1406 to generate the composite images 1408 and 1410. In some examples, the image generation module 108 of the image generation server 102 generates certain interior background images 122 based on the prompts selected by the operator, and combines them with the object images 1402, 1404 and 1406 to generate the composite images 1408 and 1410. In some examples, the prompt generation module 106 can automatically generate the best prompt, and the image generation module 108 generates the interior background image 122 based on the best prompt, and generates new composite images 1408, 1410 and 1412 using the generated interior background image 122. Alternatively, the image generation module 108 of the image generation server 102 directly generates the composite images 1408, 1410 and 1412 by generating the background in the object images 1402, 1404 and 1406 based on the prompt selected by the operator or the best prompt automatically generated by the prompt generation module 106.

[0075] Fig.15 and Fig.16 Illustrated are composite images of the same product generated for different geographic markets, where the models' faces represent customers in the corresponding geographic markets. Fig.15 An example of a GUI 1500 displaying a composite image with a model's face based on geographic location deployed on a marketplace platform is depicted in accordance with certain embodiments of the present disclosure. In this example, the model's face in the composite images 1502, 1504, and 1506 is generated specifically for the South American market. The image generation process is generally performed in Figure 2 204 in FIG. 20. For example, the image generation module 108 generates the facial image 140, and further generates the composite images 1502, 1504, and 1506 by applying the facial image 140 to the facial region of the model in the original image. Alternatively, the image generation module 108 generates the composite images 1502, 1504, and 1506 by regenerating the facial region of the model in the original image.

[0076] Fig.16 Another example of a GUI 1600 displaying a composite image with a model's face based on geographic location deployed on a marketplace platform is depicted in accordance with certain embodiments of the present disclosure. In this example, the model's face in composite images 1602, 1604, and 1606 is generated specifically for the Chinese market. The products shown in composite images 1602, 1604, and 1606 are the same as the products shown in composite images 1502, 1504, and 1506. However, the model's face is different depending on the geographic market in which the customer of the marketplace platform is located.

[0077] Fig.17An example of a GUI 1700 for displaying original object images with a white background on a market platform according to certain embodiments of the present disclosure is depicted. The original object images 1702-1716 were taken with human models originally wearing the corresponding products. These original object images can be displayed to all customers on the market platform. However, customers in a certain geographic market who look different from the models may not find these images or products attractive to them.

[0078] Fig.18 An example of a GUI 1800 showing a composite image with a geo-location-based background and a geo-location-based model face deployed on a marketplace platform for a specific geographic market is depicted in accordance with certain embodiments of the present disclosure. Fig.17 Composite images 1802-1816 of the same product in the original object images 1702-1716 in FIG. 1 with a background based on a geographic location and a model face based on a geographic location. In this example, the geographic market is France, and a customer in Paris, France is visiting the marketplace platform. Composite images 1802-1816 are displayed to the customer. The background in composite images 1802-1816 is based on certain landmark features of Paris. The model face is based on the demographic makeup of Paris.

[0079] Any suitable computing system or group of computing systems may be used to perform the operations described herein. For example, Fig.19 An example of a computing system 1900 for implementing certain embodiments of the present disclosure is depicted. An implementation of the computing system 1900 may be used to implement the image generation server 102. In other embodiments, Fig.19 A single computing system 1900 that includes devices similar to those depicted in FIG. 1 (e.g., processor, memory, etc.) may be combined in Figure 1 One or more operations described as a separate system.

[0080] The depicted example of computing system 1900 includes a processor 1902 communicatively coupled to one or more memory devices 1904. Processor 1902 executes computer executable program code stored in memory device 1904, accesses information stored in memory device 1904, or both. Examples of processor 1902 include a microprocessor, an application specific integrated circuit (“ASIC”), a field programmable gate array (“FPGA”), or any other suitable processing device. Processor 1902 may include any number of processing devices, including a single processing device.

[0081] Memory device 1904 includes any suitable non-transient computer-readable medium for storing program code 1905, program data 1907, or both. Computer-readable media may include any electrical, optical, magnetic, or other storage device capable of providing computer-readable instructions or other program codes to the processor. Non-limiting examples of computer-readable media include disks, memory chips, ROMs, RAMs, ASICs, optical storage devices, tapes or other magnetic storage devices, or any other medium from which a processing device can read instructions. Instructions may include processor-specific instructions generated by a compiler or interpreter based on a code written in any suitable computer programming language, including, for example, C, C++, C#, VisualBasic, Java, Python, Perl, JavaScript, and ActionScript.

[0082] The computing system 1900 executes program code 1905 that configures the processor 1902 to perform one or more of the operations described herein. In various embodiments, examples of program code 1905 include applications executed by a geolocation module 104 for receiving contextual data of a geographic location and mapping it to a geographic area of ​​interest, a cue generation module 106 for generating cues based on contextual data associated with certain geographic locations, an image generation module 108 for generating an image based on cues generated by the cue generation module 106, or other suitable applications that perform one or more operations described herein. The program code may reside in the memory device 1904 or any suitable computer-readable medium and may be executed by the processor 1902 or any other suitable processor.

[0083] In some embodiments, one or more memory devices 1904 store program data 1907 including one or more data sets and models described herein. Examples of these data sets include extracted images, feature vectors, aesthetic scores, processed object images, etc. In some embodiments, one or more of the data sets, models, and functions are stored in the same memory device (e.g., one of the memory devices 1904). In additional or alternative embodiments, one or more of the programs, data sets, models, and functions described herein are stored in different memory devices 1904 accessible via a data network. Also included in the computing system 1900 are one or more buses 1906. The bus 1906 communicatively couples one or more components of a respective one of the computing systems 1900.

[0084] In some embodiments, the computing system 1900 also includes a network interface device 1910. The network interface device 1910 includes any device or group of devices suitable for establishing a wired or wireless data connection to one or more data networks. Non-limiting examples of the network interface device 1910 include an Ethernet network adapter, a modem, and the like. The computing system 1900 can use the network interface device 1910 to communicate with one or more other computing devices (e.g., the user computing device 130) via a data network.

[0085] The computing system 1900 may also include a number of external or internal devices, input devices 1920, display devices 1918, or other input or output devices. For example, the computing system 1900 is shown as having one or more input / output ("I / O") interfaces 1908. The I / O interface 1908 may receive input from an input device or provide output to an output device. The input device 1920 may include any device or device group suitable for receiving visual, auditory, or other suitable inputs that control or affect the operation of the processor 1902. Non-limiting examples of input devices 1920 include touch screens, mice, keyboards, microphones, independent mobile computing devices, etc. The display device 1918 may include any device or device group suitable for providing visual, auditory, or other suitable sensory outputs. Non-limiting examples of display devices 1918 include touch screens, monitors, speakers, independent mobile computing devices, etc.

[0086] Although Fig.19 Input device 1920 and display device 1918 are depicted as being local to the computing device executing image generation server 102, but other implementations are possible. For example, in some embodiments, one or more of input device 1920 and display device 1918 may comprise a remote client computing device that communicates with computing system 1900 via network interface device 1910 using one or more data networks described herein.

[0087] Many specific details are set forth herein to provide a thorough understanding of the claimed subject matter. However, those skilled in the art will appreciate that the claimed subject matter may be practiced without these specific details. In other instances, methods, devices, or systems known to those of ordinary skill are not described in detail to avoid obscuring the claimed subject matter.

[0088] Unless otherwise specifically noted, it should be understood that discussions throughout this specification utilize terms such as "processing," "computing," "calculating," "determining," and "identifying," etc. to refer to the actions or processes of a computing device, such as one or more computers or similar electronic computing devices, that manipulate or transform data represented as physical electrical or magnetic quantities within the computing device's memory, registers, or other information storage devices, transmission devices, or display devices.

[0089] The one or more systems discussed herein are not limited to any particular hardware architecture or configuration. A computing device may include any suitable arrangement of components that provide a result conditioned on one or more inputs. Suitable computing devices include a multi-purpose microprocessor-based computer system that accesses stored software that programs or configures the computing system from a general-purpose computing device to a dedicated computing device that implements one or more embodiments of the present subject matter. Any suitable programming, scripting, or other type of language or combination of languages ​​may be used to implement the teachings contained herein in software used when programming or configuring a computing device.

[0090] Embodiments of the methods disclosed herein may be performed in the operation of such a computing device. The order of the blocks presented in the above examples may be changed—for example, the blocks may be reordered, combined, and / or decomposed into sub-blocks. Certain blocks or processes may be performed in parallel.

[0091] The use of "suitable for" or "configured to" herein is meant to be open and inclusive language that does not exclude devices that are suitable for or configured to perform additional tasks or steps. In addition, the use of "based on" is meant to be open and inclusive, as a process, step, calculation, or other action that is "based on" one or more recited conditions or values ​​may, in practice, be based on additional conditions or values ​​beyond those recited. The headings, lists, and numbers included herein are for ease of explanation only and are not meant to be limiting.

[0092] Although the subject matter has been described in detail with respect to its specific embodiments, it should be understood that those skilled in the art, after obtaining an understanding of the foregoing, can easily generate substitutions, modifications and equivalents to such embodiments. Therefore, it should be understood that the present disclosure is proposed for the purpose of illustration rather than limitation, and does not exclude the inclusion of such modifications, changes and / or additions that are obvious to those of ordinary skill in the art.

Claims

1. A method performed by one or more processing devices, comprising: determining, by a geographic location module, a set of geographic locations associated with the geographic area of ​​interest, wherein each geographic location in the set of geographic locations is mapped to contextual data associated with the geographic location; generating, by a prompt generation module, a plurality of prompts based on the set of geographic locations and the contextual data, wherein the prompt generation module comprises a first generative artificial intelligence (AI) model; generating, by an image generation module, a set of background images based on the plurality of prompts, wherein the image generation module comprises a second generative AI model; and A plurality of composite images are generated by the image generation module, wherein each composite image includes a background image from the background image set and an image of a target object. 2 . The method of claim 1 , wherein the target object comprises a product provided by an online market platform. 3 . The method of claim 1 , wherein the contextual data associated with each geographic location in the set of geographic locations includes one or more of places of interest, terrain, climate, weather, or events associated with the geographic location.

4. The method according to claim 1, further comprising: The contextual data is weighted, by the prompt generation module, based on at least one of: (i) a selection of one or more of the composite images for display on the online platform data, or (ii) behavioral data associated with a user interaction with the plurality of composite images; as well as The plurality of prompts are generated by the prompt generation module based on the context data and corresponding weights. The method according to claim 4 , wherein the behavioral data includes user engagement data and user conversion data.

6. The method of claim 4, wherein the prompt generation module uses a reinforcement learning model to weight the contextual data.

7. The method according to claim 1, further comprising: causing the plurality of composite images to be displayed via a graphical user interface GUI; receiving a selection of one or more composite images; as well as In response to the selection, the selected one or more composite images are transmitted to an online platform for deployment.

8. A system comprising: a geographic location module configured to determine a set of geographic locations associated with a geographic area of ​​interest, wherein each geographic location in the set of geographic locations is mapped to contextual data associated with the geographic location; a prompt generation module configured to generate a plurality of prompts based on the set of geographic locations and the contextual data, wherein the prompt generation module comprises a first generative artificial intelligence (AI) model; as well as An image generation module is configured to generate multiple synthetic images based on an image of a target object and an input prompt from the multiple prompts, wherein the image generation module includes a second generative AI model, and wherein each synthetic image depicts the target object in a background generated based on the input prompt.

9. The system of claim 8, wherein the target object comprises a product provided by an online market platform.

10. The system of claim 8, wherein the contextual data associated with each geographic location in the set of geographic locations includes one or more of places of interest, terrain, climate, weather, or events associated with the geographic location.

11. The system of claim 8, wherein the prompt generation module is further configured to: weighting the contextual data based on at least one of: (i) selection of one or more of the composite images for display on the online platform data, or (ii) behavioral data associated with user interactions with the plurality of composite images; and The plurality of prompts are generated based on the contextual data and corresponding weights.

12. The system of claim 11, wherein the behavioral data comprises user engagement data and user conversion data.

13. The system of claim 11, wherein the prompt generation module is configured to use a reinforcement learning model to weight the contextual data.

14. The system according to claim 8, wherein the image generation module is further configured to: causing the plurality of composite images to be displayed via a graphical user interface GUI; receiving a selection of one or more composite images; and In response to the selection, the selected one or more composite images are transmitted to an online platform for deployment.

15. A non-transitory computer readable medium storing executable instructions that, when executed by a processing device, cause the processing device to perform operations comprising: determining a set of geographic locations associated with the geographic area of ​​interest, wherein each geographic location in the set of geographic locations is mapped to contextual data associated with the geographic location; generating, using a first generative artificial intelligence (AI) model, a plurality of prompts based on the set of geographic locations and the contextual data; receiving a selection of a prompt from the plurality of prompts to determine a selected prompt; as well as A step for generating a plurality of composite images based on the selected cue, wherein each composite image comprises an image of a target object and a background image corresponding to the selected cue.

16. The non-transitory computer-readable medium of claim 15, wherein the target object comprises a product provided by an online marketplace platform.

17. The non-transitory computer-readable medium of claim 15, wherein the contextual data associated with each geographic location in the set of geographic locations includes one or more of places of interest, terrain, climate, weather, or events associated with the geographic location.

18. The non-transitory computer readable medium of claim 15, wherein the executable instructions, when executed by a processing device, cause the processing device to perform additional operations, the additional operations comprising: weighting the contextual data using a reinforcement learning model based on at least one of: (i) selection of one or more of the plurality of composite images for display on the online platform data, or (ii) behavioral data associated with user interactions with the one or more of the plurality of composite images; as well as The plurality of prompts are generated based on the contextual data and corresponding weights.

19. The non-transitory computer-readable medium of claim 18, wherein the behavioral data comprises user engagement data and user conversion data.

20. The non-transitory computer readable medium of claim 15, wherein the executable instructions, when executed by a processing device, cause the processing device to perform additional operations, the additional operations comprising: causing the plurality of composite images to be displayed via a graphical user interface GUI; receiving a selection of one or more composite images; as well as In response to the selection, the selected one or more composite images are transmitted to an online platform for deployment.

Images