Method and System for Recommending Dog Clothing Size Using Deep Learning-Based Dog Body Type Analysis

Abstract

The present invention relates to a method and system for recommending dog clothing sizes using deep learning-based dog body shape analysis, and more specifically, to a method for recommending dog clothing sizes using deep learning-based dog body shape analysis that analyzes body proportions and body shape information through a deep learning model using an image of a dog and information on breed and weight entered by a user, precisely calculates the actual dimensions of the dog by matching this with actual body length data in a database, and recommends optimal clothing by reflecting volume errors caused by fur and whether the dog is a mixed breed.

Description

Technology Field

[0001] The present invention relates to a method and system for recommending clothing sizes for dogs using deep learning-based dog body shape analysis, and more specifically, to a method for recommending clothing sizes for dogs using deep learning-based dog body shape analysis in which a deep learning model analyzes skeletal ratios and body shape characteristics from a dog image input by a user, and derives actual body dimensions corrected for measurement errors due to fur volume or posture by performing calculations with a database storing standard actual measurements, thereby deriving and providing optimal clothing products that match the physical characteristics of individual dogs without the need for separate manual measurements. Background Technology

[0003] As the trend of "Pet Humanization," in which companion animals are regarded as members of the family, spreads in modern society, the pet-related market, particularly the pet fashion industry, is growing rapidly. Consequently, the purchase of pet clothing is becoming active not only through offline stores but also through online platforms.

[0004] However, in the case of pets, particularly dogs, even within the same breed, body proportions such as body length, chest circumference, and neck circumference vary greatly from individual to individual, and body shape also changes significantly depending on weight or age. Conventional online pet clothing sales methods often classify and provide sizes (XS, S, M, L, etc.) based primarily on simple indicators such as weight or back length. This method fails to accurately reflect the specific physical characteristics of dogs, resulting in a problem where the rate of consumers requesting exchanges or refunds after purchase due to ill-fitting sizes is very high.

[0005] In particular, one of the biggest technical challenges in determining clothing sizes for dogs is the presence of fur. Due to thick fur, dogs frequently appear much larger than their actual body volume when viewed with the naked eye or in simple images, leading to significant discrepancies between the measurements and the dimensions of clothing that must fit snugly against the skin. Furthermore, for mixed-breed dogs—a group of two or more breeds that has been increasing recently—it is difficult to directly apply standard databases for specific breeds, placing limitations on accurately estimating body dimensions using only existing statistical methods.

[0006] Therefore, there is an urgent need to develop a new dimension of dog clothing size recommendation technology that can calculate precise body measurements suitable for actual wear by utilizing basic information and images entered by the user without the need for separate complex tools, correcting visual errors caused by fur, and even taking into account the non-standardized body shape of mixed-breed dogs. Prior art literature

[0008] Korean Patent Publication KR 10-2025-0076995 A “Method and device for providing clothing size information for companion dogs” (May 30, 2025) The problem to be solved

[0009] The present invention relates to a method and system for recommending dog clothing sizes using deep learning-based dog body shape analysis. More specifically, the invention aims to provide a method and system for recommending dog clothing sizes using deep learning-based dog body shape analysis that analyzes body proportions and body shape information through a deep learning model using an image of a dog and information on breed and weight entered by a user, precisely calculates the actual dimensions of the dog by matching this with actual body length data in a database, and recommends optimal clothing by reflecting volume errors caused by fur and whether the dog is a mixed breed. means of solving the problem

[0011] To solve the above problem, a method for recommending dog clothing sizes using deep learning-based dog body shape analysis, performed on a server system comprising one or more processors and one or more memories, wherein the server system includes a database storing clothing information including the body length, body circumference, and neck circumference of a corresponding clothing product entered by a seller for each of a plurality of clothing products to be sold; and the actual total body length of the corresponding dog according to combinations of dog breed, weight, and body shape; and the dog clothing size recommendation method comprises a body shape information derivation step of receiving input information including one or more of dog breed information, weight information, and image information from a user, and deriving body shape information that classifies the dog's body shape into overweight, standard weight, and underweight based on the input information; A length derivation step of inputting the above image information into a pre-trained length derivation model to derive the body circumference, body length, neck circumference, and total body length of the pet dog identified in the above image information; a ratio information calculation step of calculating a first ratio information regarding the ratio of the body circumference to the total body length, a second ratio information regarding the ratio of the body length to the total body length, and a third ratio information regarding the ratio of the neck circumference to the total body length; and an actual length calculation step of deriving a matching total body length that matches the dog breed information, weight information, and body shape information of the user from among a plurality of actual total body lengths stored in the database, and calculating the actual body circumference, actual body length, and actual neck circumference of the pet dog based on the matching total body length, the first ratio information, the second ratio information, and the third ratio information. The present invention provides a method for recommending dog clothing sizes, comprising: a clothing recommendation step of comparing the actual body circumference, actual body length, and actual neck circumference with the clothing body length, clothing body circumference, and clothing neck circumference, respectively, and selecting and providing clothing products that correspond to the dog’s body circumference, body length, and neck circumference.

[0012] In one embodiment of the present invention, the total body length corresponds to the length from a first feature point referring to the tip of the dog's head to a second feature point referring to the starting point of the tail, the body circumference corresponds to the circumference at a third feature point corresponding to the thickest part of the dog's chest, the body length corresponds to the length measured along the back line from a fourth feature point referring to the starting point where the dog's neck and back meet to a fifth feature point referring to the starting point of the tail, and the neck circumference corresponds to the circumference measured based on a sixth feature point corresponding to the lowest part of the neck where the dog's neck and shoulder are connected.

[0013] In one embodiment of the present invention, the length derivation model is a deep learning model trained based on training data in which at least one of the total body length, torso circumference length, torso length, and neck circumference length of a corresponding dog is labeled on a plurality of dog images according to the combination of dog breed information, weight information, and body shape information, and can identify a first feature point, a second feature point, a third feature point, a fourth feature point, a fifth feature point, and a sixth feature point that serve as reference points for deriving the total body length, torso circumference length, torso length, and neck circumference length from the image information, and can derive the total body length, torso circumference length, torso length, and neck circumference length based on the derived plurality of feature points.

[0014] In one embodiment of the present invention, the length derivation step may further include: a step of extracting from the image information a first layer corresponding to an outer silhouette including the dog's fur and a second layer corresponding to an inner silhouette excluding the dog's fur; and a step of calculating a fur length compressible by the dog's fur based on the distance difference between the first layer and the second layer.

[0015] In one embodiment of the present invention, the actual length calculation step includes a step of calculating the body circumference excluding hair and the neck circumference excluding hair by subtracting the hair length from the actual body circumference and the actual neck circumference, respectively, when the hair length exceeds a preset hair length threshold; and the clothing recommendation step can select clothing products based on the body circumference excluding hair and the neck circumference excluding hair.

[0016] In one embodiment of the present invention, when the breed information corresponds to a mixed dog comprising two or more breeds, the ratio information calculation step further comprises: a step of deriving the distance between the eyes with minimal change in skeletal size according to the breed of the pet dog from the image information; and a step of deriving the ratio of the distance between the eyes to the ratio of the total body length; and the actual length calculation step may include a body length correction step of calculating a correction value based on the ratio of the eyes to the average value of a plurality of actual total body lengths derived from the database based only on the weight information and body shape information, and calculating the actual torso circumference, actual torso length, and actual neck circumference of the mixed dog, respectively, based on the correction value.

[0017] In one embodiment of the present invention, the clothing recommendation step may further include the step of inputting the image information for the selected clothing product into a pre-trained generative artificial intelligence model to convert the pet dog into a 3D model, and generating a virtual fitting model in which the clothing product is combined on the 3D model and providing it to the user.

[0018] In one embodiment of the present invention, a dog clothing size recommendation system using deep learning-based dog body shape analysis, comprising one or more processors and one or more memories, comprises: clothing information including the body length, body circumference, and neck circumference of a corresponding clothing product entered by a seller for each of a plurality of clothing products to be sold; a database storing the actual total body length of a corresponding dog according to combinations of dog breed, weight, and body shape; a body shape information derivation unit that receives input information including one or more of dog breed information, weight information, and image information from a user, and derives body shape information classifying the body shape of the dog into overweight, standard weight, and underweight based on the input information; and a length derivation unit that inputs the image information into a pre-trained length derivation model to derive the body circumference, body length, neck circumference, and total body length of the dog identified in the image information, respectively. A ratio information calculation unit that calculates a first ratio information regarding the ratio of the torso circumference to the total body length, a second ratio information regarding the ratio of the torso length to the total body length, and a third ratio information regarding the ratio of the neck circumference to the total body length; an actual length calculation unit that derives a matching total body length that matches the dog breed information, weight information, and body shape information of the user from among a plurality of actual total body lengths stored in the database, and calculates the actual torso circumference, actual torso length, and actual neck circumference of the pet dog, respectively, based on the matching total body length, the first ratio information, the second ratio information, and the third ratio information; The present invention provides a dog clothing size recommendation system comprising: a clothing recommendation unit that compares the actual body circumference, actual body length, and actual neck circumference with the clothing body length, clothing body circumference, and clothing neck circumference, respectively, and selects and provides clothing products that correspond to the dog’s body circumference, body length, and neck circumference. Effects of the invention

[0020] According to one embodiment of the present invention, the inconvenience of a user having to directly measure a moving dog using a measuring tape is eliminated, and by inputting only the captured image and basic dog breed and weight information, a deep learning model automatically calculates precise body dimensions by combining body proportions and actual measurement information in a database, thereby achieving the effect of dramatically improving both user convenience and measurement accuracy.

[0021] According to one embodiment of the present invention, a deep learning model clearly identifies anatomical feature points such as the tip of a dog's head, the starting point of its tail, and the connection between its neck and shoulder to derive dimensions, thereby preventing errors that occurred when a non-expert user subjectively determined the measurement location, and enabling highly reliable clothing recommendations through the standardization of measurement standards.

[0022] According to one embodiment of the present invention, when analyzing an image, the outer silhouette including the dog's fur and the inner silhouette excluding the fur are distinguished through deep learning layer analysis, and the length inflated by the fur is calculated and subtracted from the actual body dimensions, thereby solving the problem of clothing being loose or ill-fitting even for dog breeds with thick fur (such as Pomeranians and Bichons), and thus providing the effect of recommending an optimized size for the dog.

[0023] According to one embodiment of the present invention, in the case of mixed-breed dogs for which it is difficult to apply standard body shape information from a breed database as is, the length between the eyes, which undergoes less change due to skeletal growth and is less affected by hair, is used as an absolute reference point to correct the total body length, thereby enabling the effect of estimating dimensions with high accuracy even for various dog breeds for which standardized data is unavailable.

[0024] According to one embodiment of the present invention, by providing a virtual fitting service that allows users to try on recommended clothing based on calculated size information onto a 3D model of a pet dog using a generative artificial intelligence model, users can visually check how the clothing looks when worn before purchasing, thereby reducing the probability of purchase failure and increasing shopping satisfaction. Brief explanation of the drawing

[0026] FIG. 1 illustrates a server system that performs a method for recommending dog clothing sizes according to an embodiment of the present invention. FIG. 2 illustrates the steps of performing a method for recommending dog clothing sizes according to an embodiment of the present invention. FIG. 3 illustrates the definitions of total body length, torso circumference, torso length, and neck circumference according to one embodiment of the present invention. FIG. 4 illustrates the learning process of a length derivation model according to one embodiment of the present invention and the principle of length derivation using the same. FIG. 5 illustrates an example of performing a body shape information collection step, a length derivation step, and a ratio information calculation step according to an embodiment of the present invention. FIG. 6 illustrates an example of performing an actual length calculation step and a clothing recommendation step according to an embodiment of the present invention. FIG. 7 illustrates an example of deriving a first layer and a second layer from image information according to an embodiment of the present invention and calculating hair length based thereon. FIG. 8 illustrates the process of correcting the body circumference and neck circumference based on the calculated hair length according to one embodiment of the present invention. FIG. 9 illustrates an example of deriving the distance between the two eyes from an image and calculating the ratio of the two eyes based on the distance when the breed information of a pet dog according to one embodiment of the present invention corresponds to a mixed breed dog. FIG. 10 illustrates an example of calculating the total body length based on both eyes and the average total body length, respectively, by referring to database information according to an embodiment of the present invention. FIG. 11 illustrates an example of calculating the final corrected total body length using the difference between the total body length based on both eyes and the average total body length according to one embodiment of the present invention. FIG. 12 illustrates an example of generating a virtual fitting model using a generative artificial intelligence model in a clothing recommendation step according to an embodiment of the present invention. FIG. 13 schematically illustrates the internal configuration of a computing device according to one embodiment of the present invention. Specific details for implementing the invention

[0027] Hereinafter, various embodiments and / or aspects are disclosed with reference to the drawings. For illustrative purposes, numerous specific details are disclosed in the following description to aid in a general understanding of one or more aspects. However, it will also be recognized by those skilled in the art that these aspects may be practiced without such specific details. The following description and the accompanying drawings describe specific exemplary aspects of one or more aspects in detail. However, these aspects are exemplary, and some of the various methods in the principles of the various aspects may be used, and the description is intended to include all such aspects and their equivalents.

[0029] In addition, various aspects and features will be presented by a system that may include multiple devices, components and / or modules, etc. It should also be understood and recognized that various systems may include additional devices, components and / or modules, etc., and / or may not include all of the devices, components, modules, etc. discussed in relation to the drawings.

[0030] Terms such as “embodiment,” “example,” “aspect,” “example,” etc. as used herein may not be interpreted as implying that any aspect or design described is superior or more advantageous than other aspects or designs. Terms used below, such as “part,” “component,” “module,” “system,” “interface,” etc., generally refer to computer-related entities and may, for example, refer to hardware, a combination of hardware and software, or software.

[0031] Additionally, the terms “comprising” and / or “comprising” should be understood to mean that the relevant feature and / or component is present, but not to exclude the presence or addition of one or more other features, components and / or groups thereof.

[0032] Additionally, terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but said components are not limited by said terms. Such terms are used solely for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be named the second component, and similarly, the second component may be named the first component. The term "and / or" includes a combination of a plurality of related described items or any of a plurality of related described items.

[0033] Furthermore, in the embodiments of the present invention, all terms used herein, including technical or scientific terms, unless otherwise defined, have the same meaning as generally understood by those skilled in the art to which the present invention 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 in the embodiments of the present invention.

[0035] FIG. 1 illustrates a server system that performs a method for recommending dog clothing sizes according to an embodiment of the present invention.

[0036] According to one embodiment of the present invention, the server system comprises a database in which clothing information including the torso length, torso circumference, and neck circumference of the corresponding clothing product entered by a seller for each of a plurality of clothing products to be sold; and the actual total body length of the corresponding dog according to combinations of the dog's breed, weight, and body type are stored; and the dog clothing size recommendation method comprises: a body type information derivation step of receiving input information including one or more of the dog's breed information, weight information, and image information from a user, and deriving body type information that classifies the dog's body type into overweight, standard weight, and underweight based on the input information; and a length derivation step of inputting the image information into a pre-trained length derivation model to derive the torso circumference, torso length, neck circumference, and total body length of the dog identified in the image information, respectively. A ratio information calculation step for calculating a first ratio information regarding the ratio of the torso circumference to the total body length, a second ratio information regarding the ratio of the torso length to the total body length, and a third ratio information regarding the ratio of the neck circumference to the total body length; an actual length calculation step for deriving a matching total body length that matches the dog breed information, weight information, and body shape information of the user from among a plurality of actual total body lengths stored in the database, and calculating the actual torso circumference, actual torso length, and actual neck circumference of the pet dog, respectively, based on the matching total body length, the first ratio information, the second ratio information, and the third ratio information; and may include a clothing recommendation step of comparing the actual body circumference, actual body length, and actual neck circumference with the clothing body length, clothing body circumference, and clothing neck circumference, respectively, to select and provide clothing products that correspond to the body circumference, body length, and neck circumference of the pet dog.

[0038] A server system (1) according to one embodiment of the present invention may correspond to a computing readable recording medium comprising one or more processors and one or more memories. The server system (1) may correspond to a computing device that transmits and receives data and performs communication with an external user terminal through a network.

[0039] In addition, the user can input input information including one or more of the following: a photo of the dog (image information), breed (poodle, Maltese, etc.), and weight information, to the server system (1) through the user terminal.

[0041] As illustrated in FIG. 1, the server system may be configured to include a body shape information extraction unit (10), a length extraction unit (11), a ratio information calculation unit (12), an actual length calculation unit (13), a clothing recommendation unit (14), and a database (15).

[0043] First, the database (15) can store clothing information including clothing body length, clothing body circumference, and clothing neck circumference for each clothing product received from a seller. Additionally, the database (15) can store actual total body length data corresponding to various combinations of dog breed, weight, and body shape in the form of a mapping table.

[0045] The above body shape information derivation unit (10) can receive input information including one or more of the dog breed information, weight information, and image information from a user (via a user terminal) and perform the function of deriving body shape information of the dog based on the received input information. Specifically, it can perform a body shape information derivation step that derives body shape information including whether the dog is overweight, has a standard weight, or is underweight by analyzing the dog breed information, weight information, and image information entered by the user using a deep learning-based multi-modal model. A detailed explanation of the above multi-modal model will be provided later in the description of FIG. 2.

[0047] The length derivation unit (11) can recognize each part of a dog's body and measure its length by utilizing a pre-trained deep learning-based length derivation model. Specifically, the length derivation unit (11) can input the image information into the length derivation model to identify key feature points such as the dog's head, neck, back, and tail on the input image, and perform a length derivation step to derive the total body length, body circumference, body length, and neck circumference (preferably in units of image pixels) based on this. A detailed description of the length derivation model will be provided later in the description of FIGS. 2 and FIGS. 4.

[0049] The above ratio information calculation unit (12) can perform the function of calculating the relative size ratio between each body part measured by the above length derivation unit (11). Specifically, in order to secure a unique body ratio that is not affected by the measurement distance or camera zoom magnification, it can perform a ratio information calculation step of calculating the first ratio information, the second ratio information, and the third ratio information, which are the ratios that the torso circumference, torso length, and neck circumference each occupy based on the total body length.

[0051] The actual length calculation unit (13) can derive data corresponding to the user's dog's total body length among the actual total body length data stored in the database (15), and calculate actual physical dimensions by applying the previously calculated ratio information. Specifically, it can perform an actual length calculation step by finding a matching total body length that corresponds to the user's dog's breed information, weight information, and body shape information among the actual total body lengths stored in the database (15), and applying the matching total body length to the first ratio information, second ratio information, and third ratio information, respectively, to calculate the dog's actual torso circumference, actual torso length, and actual neck circumference.

[0053] The above clothing recommendation unit (14) can select and provide the most suitable clothing product to the user based on the calculated actual body measurements of the pet dog. Specifically, it can perform a clothing recommendation step by comparing the clothing body circumference, clothing body length, and clothing neck circumference for each clothing product, which have been pre-entered by the seller, with the pet dog's actual body circumference, actual body length, and actual neck circumference, and providing the user with a clothing product of a size that can be worn comfortably within an error range.

[0055] FIG. 2 illustrates the steps of performing a method for recommending dog clothing sizes according to an embodiment of the present invention.

[0057] As illustrated in FIG. 2, first, in the body shape information derivation step (S10), input information including breed information, weight information, and image information of the user's pet dog can be received from the user. In the body shape information derivation step (S10), the degree of obesity of the pet dog can be determined by analyzing visual image information together with the weight numerical value, rather than simply looking at the weight numerical value.

[0058] To this end, the present invention may utilize a deep learning-based multimodal model capable of analyzing text data and image data by combining them. The multimodal model may correspond to a model that has been pre-trained based on a vast dataset of images of dogs of various breeds and body types, and actual body measurements and obesity labeling data corresponding to each image.

[0059] Specifically, the multimodal model can derive actual external characteristics of a pet dog that are difficult to identify solely by weight figures by cross-referencing and analyzing text embedding vectors based on input dog breed information and weight information, and visual feature vectors based on image information. Through this, the multimodal model can probabilistically calculate whether the pet dog is relatively underweight, normal weight, or overweight within the same breed and weight group, and derive body shape information that includes one of underweight, normal weight, or overweight, corresponding to the range with the highest probability.

[0061] Next, in the length derivation step (S11), the previously received image information can be input into the length derivation model. The length derivation model can detect a dog object within the image, extract multiple feature points based on joints or boundary points of major body parts, and derive the body length of the dog based on the multiple feature points.

[0062] Specifically, the above length derivation model derives multiple feature points on image pixels and, based on these multiple feature points, can quantify and derive the total body length, which is the distance from the tip of the dog's head to the starting point of its tail; the torso circumference, which is the width of the thickest part of the chest; the torso length, which is the length of the back line from the point where the neck meets the back to the tail; and the neck circumference, which is the circumference of the neck. The values ​​derived in this case may correspond to lengths in pixel units dependent on image resolution or relative size values, rather than actual physical units.

[0064] Additionally, in the length derivation step (S11), if the reliability of the result value derived by the length derivation model is less than a preset reliability threshold, or if the dog's posture in the image information is determined to be unsuitable for measuring body length, such as when the dog is crouching or a part of its body is obscured, the server system (1) may further include a step of sending a message requesting re-shooting to the user terminal.

[0065] In this case, for precise analysis, a request may be made to include specific guidelines asking for a reference object of standardized size, such as a credit card or A4 paper, to be placed alongside the dog for filming.

[0067] In the above ratio information calculation step (S12), the body ratio can be calculated using the lengths of each part obtained in the above length derivation step (S11).

[0068] Since the dog appears small when the distance between the camera and the dog is far and large when the distance is close, the absolute pixel length measured in the image cannot be used directly as an actual dimension. Therefore, the ratio information calculation step (S12) can generate ratio information that is not affected by the shooting distance.

[0069] Specifically, the ratio information calculation step (S12) can calculate a first ratio information which is the ratio of the torso circumference to the total body length, a second ratio information which is the ratio of the torso length to the total body length, and a third ratio information which is the ratio of the neck circumference to the total body length.

[0071] In the actual length calculation step (S13), among the actual total body lengths stored in the database (15), a matching total body length that matches the input information and body shape information is derived, and the actual physical dimensions can be calculated by combining the matching total body length with the previously calculated ratio information.

[0072] Specifically, the server system (1) can query the database (15) to extract a matching total body length corresponding to the actual total body length that matches the dog breed information and weight information entered by the user and the body shape information derived in the body shape information derivation step (S10) among the actual total body lengths stored in the database (15). By multiplying the matching total body length by the first ratio information, second ratio information, and third ratio information calculated in the ratio information calculation step (S12), respectively, the actual torso circumference, actual torso length, and actual neck circumference of the corresponding pet dog can be finally calculated.

[0074] Finally, in the clothing recommendation step (S14), the calculated actual dimensions of the dog can be compared with the dimensions of the clothing products registered by the seller to make a match.

[0075] Specifically, the actual torso circumference, actual torso length, and actual neck circumference calculated in the actual length calculation step (S13) can be compared with the clothing torso length, clothing torso circumference, and clothing neck circumference included in the clothing information stored in the database (15). At this time, clothing items within a preset appropriate error range can be filtered by considering the elasticity of the clothing, the thickness of the material, and the allowance required when wearing. Subsequently, the system can organize the clothing items selected through the filtering process into a recommendation list and provide it to the user.

[0076] Additionally, the clothing information stored in the database (15) may include breed information suitable for the corresponding clothing product entered in advance by the seller. Even if the actual body circumference, actual body length, and actual neck circumference are suitable for the clothing product's body length, body circumference, and neck circumference, the clothing product may not be suitable for the dog due to the inherent body shape characteristics of the breed. Therefore, in the clothing recommendation step (S14), the breed information entered by the user is compared with the breed information included in the clothing information, and weights are assigned to clothing products that are specialized in designing or determined to be suitable for the corresponding breed, thereby selecting them by priority and providing them to the user as a recommendation list.

[0078] According to one embodiment of the present invention, the inconvenience of a user having to directly measure a moving dog using a measuring tape is eliminated, and by inputting only the captured image and basic dog breed and weight information, a deep learning model automatically calculates precise body dimensions by combining body proportions and actual measurement information in a database, thereby achieving the effect of dramatically improving both user convenience and measurement accuracy.

[0080] FIG. 3 illustrates the definitions of total body length, torso circumference, torso length, and neck circumference according to one embodiment of the present invention.

[0081] In a method for recommending clothing sizes for pet dogs according to an embodiment of the present invention, the total body length corresponds to the length from a first feature point referring to the tip of the pet dog's head to a second feature point referring to the starting point of the tail, the torso circumference corresponds to the circumference at a third feature point corresponding to the thickest part of the pet dog's chest, the torso length corresponds to the length measured along the back line from a fourth feature point referring to the starting point where the pet dog's neck and back meet to a fifth feature point referring to the starting point of the tail, and the neck circumference corresponds to the circumference measured based on a sixth feature point corresponding to the lowest part of the neck where the pet dog's neck and shoulder are connected.

[0083] The deep learning-based length derivation model used in the present invention can improve the accuracy of the analysis by setting anatomically clearly identifiable feature points rather than ambiguous regions when analyzing the body structure of a pet dog.

[0084] First, as a reference point for measuring the total body length, the tip of the nose, which is clearly distinguished from the background and located at the foremost point of the dog's body, can be set as the first feature point, and the point at the base of the tail, where the spine ends and the tailbone begins, can be set as the second feature point. As illustrated in FIG. 3 (a), the straight distance between the first feature point and the second feature point, or the curved distance along the spine line, can be defined as the total body length that serves as the reference scale of the system.

[0085] The body circumference (chest circumference) is one of the most important measurements when wearing clothing, and the point where the rib cage is most expanded, located just behind the dog's front legs, can be set as the third feature point. When analyzing images, it can be calculated based on the circumference of the cross-section passing through the corresponding third feature point.

[0086] Torso length is a factor determining the length of the garment and can be defined as the length measured along the spinal line from the fourth feature point, where the neck and shoulders meet, to the fifth feature point, where the tail begins. Here, the fifth feature point may be at the same location as the previously defined second feature point.

[0087] Finally, the neck circumference is the point where the neckline of the clothing item is located. The lowest point of the neck where the dog's neck and shoulder connect is set as the 6th feature point, and the circumference when wrapping the entire neck 360 degrees around this point can be measured. This corresponds to a position slightly below where the leash is worn, where the clothing naturally rests.

[0089] According to one embodiment of the present invention, a deep learning model clearly identifies anatomical feature points such as the tip of a dog's head, the starting point of its tail, and the connection between its neck and shoulder to derive dimensions, thereby preventing errors that occurred when a non-expert user subjectively determined the measurement location, and enabling highly reliable clothing recommendations through the standardization of measurement standards.

[0091] FIG. 4 illustrates the learning process of a length derivation model according to one embodiment of the present invention and the principle of length derivation using the same.

[0092] According to one embodiment of the present invention, the length derivation model is a deep learning model trained based on training data in which at least one of the total body length, torso circumference length, torso length, and neck circumference length of a corresponding dog is labeled on a plurality of dog images according to the combination of dog breed information, weight information, and body shape information, and can identify a first feature point, a second feature point, a third feature point, a fourth feature point, a fifth feature point, and a sixth feature point that serve as reference points for deriving the total body length, torso circumference length, torso length, and neck circumference length from the image information, and can derive the total body length, torso circumference length, torso length, and neck circumference length based on the derived plurality of feature points.

[0094] The above length derivation model may correspond to a model that has been pre-trained based on a dataset of tens of thousands of images of dogs of various breeds and body types. In each image of the training data used at this time, not only numerical information such as the actual total body length, body circumference, body length, and neck circumference of the dog included in the image, but also pixel coordinates from the first feature point to the sixth feature point defined in Fig. 3 above may be stored together as metadata.

[0095] The above length derivation model can be designed to automatically detect multiple feature points within an input image that serve as criteria for measuring body length by learning the correlation between the external characteristics and skeletal structure of a pet dog based on the above training data. Specifically, the above length derivation model can be trained to predict the coordinates of multiple feature points, such as the tip of the nose, the starting point of the tail, and the connection point between the neck and shoulders, with high accuracy by analyzing visual patterns such as changes in pixel values, brightness, and texture within the image.

[0097] As illustrated in FIG. 4, the learned length derivation model can identify a first feature point, a second feature point, a third feature point, a fourth feature point, a fifth feature point, and a sixth feature point from the input image, respectively. Then, by calculating the distance between these identified feature points, the total body length, torso circumference, torso length, and neck circumference on the image can be finally derived.

[0099] FIG. 5 illustrates an example of performing a body shape information collection step, a length derivation step, and a ratio information calculation step according to an embodiment of the present invention.

[0101] First, in the above body shape information derivation step (S10), the user can transmit to the server system (1) via their terminal dog breed information that the dog's breed is 'Poodle' and weight information that the dog's weight is '10kg', along with image information of the dog (preferably a side view of the dog's entire body). The body shape information derivation unit (10) of the server system (1) can input the received text data containing the dog breed information 'Poodle' and weight information '10kg', and the image information, into the multimodal model to derive body shape information of 'normal' for the dog.

[0103] In the above length derivation step (S11), the deep learning-based length derivation model can analyze the image information to quantify the length of each body part. The value derived at this time may not be an actual physical unit of cm, but may correspond to the number of pixels within the image resolution or a relative distance value. For example, the model may recognize that the total body length of the dog in the image is 1,000 pixels, the torso circumference is 300 pixels, the torso length is 600 pixels, and the neck circumference is 200 pixels.

[0105] In the above ratio information calculation step (S12), ratio information that is not affected by shooting distance or angle can be calculated using the previously derived total body length, torso circumference length, torso length, and neck circumference length.

[0106] As shown in the example illustrated in FIG. 5, the ratio of the torso circumference (300 pixels) to the total body length (1000 pixels) can be calculated to obtain a first ratio information of 1:0.3, the ratio of the torso length (600 pixels) to the total body length (1000 pixels) can be calculated to obtain a second ratio information of 1:0.6, and the ratio of the neck circumference (200 pixels) to the total body length (1000 pixels) can be calculated to obtain a third ratio information of 1:0.2. The ratios of 0.3, 0.6, and 0.2 calculated in this way can later be combined with the actual total body length in the database (15) and used as parameters to restore the actual dimensions.

[0108] FIG. 6 illustrates an example of performing an actual length calculation step and a clothing recommendation step according to an embodiment of the present invention.

[0109] FIG. 6 (a) illustrates the process of calculating actual body measurements by deriving the matching total body length from the database (15), and FIG. 6 (b) illustrates the process of selecting a suitable product by comparing the calculated actual measurements with clothing information stored in the database.

[0111] After performing the ratio information calculation step (S12) of the example in Fig. 5, in the actual length calculation step (S13), the server system (1) can first query the database (15) using the dog breed information (Poodle), weight information (10kg) entered by the user, and the derived body shape information (normal) as key values.

[0112] Looking at the database example shown in FIG. 6(a), it can be seen that the actual total body length of a Poodle with a weight of more than 8 kg and less than or equal to 10 kg (8~10 kg) and a normal body shape is set to 15 cm. The server system (1) can determine this 15 cm as the matching total body length of the corresponding pet dog.

[0113] The actual torso circumference, actual torso length, and actual neck circumference can be calculated by multiplying the above-mentioned total matching body length by the first ratio information, second ratio information, and third ratio information calculated in the preceding ratio information calculation step (S12), respectively. In the example of FIG. 6 (a), the actual torso circumference is calculated as 4.5 cm by multiplying 15 cm by 0.3, the actual torso length is calculated as 9 cm by multiplying 15 cm by 0.6, and the actual neck circumference is calculated as 3 cm by multiplying 15 cm by 0.2.

[0115] Subsequently, in the clothing recommendation step (S14), the server system (1) can search for clothing information registered by the seller in the database (15) based on the calculated actual torso circumference, actual torso length, and actual neck circumference.

[0116] In the example of FIG. 6 (b), the server system (1) can search for clothing products in the database (15) that match within a preset error range based on the calculated actual body dimensions (body circumference 4.5 cm, body length 9 cm, neck circumference 3 cm).

[0117] The pre-set margin of error here refers to an appropriate allowance that is larger than the actual body size but not too loose to ensure the dog's activity and comfort of wearing, and can be set, for example, as '1 cm or more and 3 cm or less compared to the actual size' or '105% to 120% of the actual size'.

[0118] In the example of FIG. 6(b), the preset error range Assuming the size is 2cm, product A is filtered out because it has a size that is more than 10cm larger than the actual size (4.5cm, 9cm, 3cm) and is significantly outside the margin of error.

[0119] On the other hand, Product B is confirmed to have a torso circumference of 5cm (+0.5cm), a torso length of 10cm (+1cm), and a neck circumference of 5cm (+2cm), so it is determined that all parts satisfy the preset error range (±2cm). In addition, Product C has a torso circumference of 3cm (-1.5cm), a torso length of 8cm (-1cm), and a neck circumference of 3cm (0cm), so although it is manufactured smaller than the actual dimensions, it is included within the preset error range (±2cm) and is therefore not filtered. Accordingly, the server system (1) can select both Product B and Product C as suitable products and generate a recommended clothing list including them and provide it to the user.

[0121] FIG. 7 illustrates an example of deriving a first layer and a second layer from image information according to an embodiment of the present invention and calculating hair length based thereon.

[0122] The length derivation step according to one embodiment of the present invention may further include: a step of extracting from the image information a first layer corresponding to an outer silhouette including the fur of the dog and a second layer corresponding to an inner silhouette excluding the fur of the dog; and a step of calculating a fur length compressible by the fur of the dog based on the distance difference between the first layer and the second layer.

[0124] For long-haired or double-coated breeds with long, thick fur, such as Pomeranians, Bichon Frises, and Poodles, there can be a significant discrepancy between their apparent visual volume and their actual body size.

[0125] When the length is derived based on feature points on the image in the above length derivation step (S11), since this is basically a value measured based on the appearance of the body covered with fur, it is highly likely to be over-measured compared to the actual body.

[0126] Since clothing is worn close to the skin with the fur compressed, calculating measurements without considering the volume of the fur can lead to the recommendation of clothing that is excessively large compared to the actual size required. Therefore, to calculate precise measurements, it may be necessary to determine and correct for the thickness of the fur separately.

[0128] To solve this, the length derivation model used in the present invention can integrate the function of extracting feature points as described in FIG. 4, as well as the function of precisely separating objects within an image at the pixel level.

[0129] To perform this function, the length derivation model may be pre-trained based on vast training data consisting of pairs of original image data of companion dogs of various breeds and hair lengths, and information on the entire external shape area including the hair and the actual skin area inside the hair, labeled for each of the images by experts or precision measuring equipment.

[0131] Specifically, the above length derivation model can comprehensively analyze the subtle texture differences and light and shadow of hair that are difficult to distinguish with the human eye, as well as learned body structure data, to extract, for the input image information, a first layer corresponding to the external silhouette, which is the boundary line of the outwardly visible thick hair, and a second layer corresponding to the internal silhouette, which is the boundary line of the actual skin hidden within the hair, as independent regions.

[0132] At this time, the second layer may correspond to a virtual skin line where the actual body is predicted to be located by back-calculating the sagging or density of the hair due to gravity by the length derivation model, or it may correspond to a skeletal structure generated based on the first to sixth feature points and other major joint feature points derived in Fig. 6.

[0134] As illustrated in FIG. 7, the length derivation model can calculate the difference in distance between the boundary line of the first layer extracted from the input image information and the boundary line of the second layer. This difference in distance represents the thickness of the hair covering the area and can be defined as the compressible hair length that can be compressed and reduced when wearing clothing. The hair length may be calculated as an average value over the entire torso, or it may be calculated locally for the neck or chest area, which is sensitive to the fit of the clothing.

[0136] FIG. 8 illustrates the process of correcting the body circumference and neck circumference based on the calculated hair length according to one embodiment of the present invention.

[0137] According to one embodiment of the present invention, the actual length calculation step comprises the step of calculating the hair-excluded torso circumference and hair-excluded neck circumference by subtracting the hair length from the actual torso circumference and actual neck circumference, respectively, when the hair length exceeds a preset hair length threshold; and the clothing recommendation step can select clothing products based on the hair-excluded torso circumference and hair-excluded neck circumference.

[0139] The present invention may perform a process of comparing the hair length calculated through the process described above in the description of FIG. 7 with a preset hair length threshold to determine whether the hair length is a meaningful subject for correction. In this case, even if the hair is abundant, the torso length measured along the back line is a body part where volume distortion, such as length increase or decrease due to the hair, hardly occurs, so it can be excluded from the hair length subtraction target and the existing value can be maintained.

[0141] As shown in FIG. 8, if the hair length does not exceed the hair length threshold (NO), the server system (1) considers the dog as a breed that does not require hair correction (short-haired breed, etc.), and can maintain the previously calculated actual body circumference and actual neck circumference and transmit them to the clothing recommendation step (S14).

[0143] On the other hand, if the hair length exceeds the hair length threshold (YES), the server system (1) can determine that the dog is a long-haired breed with significant volume distortion due to hair and perform a correction operation. To reflect the physical phenomenon where clothing presses down on the hair and adheres to the skin, the value obtained by subtracting the hair length from the previously calculated actual body circumference length can be newly defined as the body circumference excluding hair, and the value obtained by subtracting the hair length from the actual neck circumference length can be newly defined as the neck circumference excluding hair.

[0144] Specifically, the body circumference excluding hair can be calculated by reflecting the volume reduction corresponding to the hair length in the actual body circumference, and the neck circumference excluding hair can be calculated by reflecting the volume reduction corresponding to the hair length in the actual neck circumference.

[0145] Finally, the above server system (1) can determine the calculated body circumference excluding fur and neck circumference excluding fur as the effective body measurements of the corresponding dog and transmit them as input values ​​to the clothing recommendation step (S14).

[0147] To summarize, if the hair length is less than or equal to the hair length threshold (NO), the actual torso circumference, actual torso length, and actual neck circumference calculated in the actual length calculation step (S13) are transmitted to the clothing recommendation step (S14), and if the hair length is greater than the hair length threshold (YES), the hair-excluded torso circumference and hair-excluded neck circumference, which are calculated by subtracting the hair length from the actual torso length, actual torso circumference, and actual neck circumference calculated in the actual length calculation step (S13), can be transmitted to the clothing recommendation step (S14).

[0149] According to one embodiment of the present invention, when analyzing an image, the outer silhouette including the dog's fur and the inner silhouette excluding the fur are distinguished through layer analysis, and the length inflated by the fur is calculated and subtracted from the actual body measurements, thereby resolving the problem of clothing being loose or ill-fitting even for dog breeds with thick fur, and thus providing the effect of recommending an optimized size for the dog.

[0151] FIG. 9 illustrates an example of deriving the distance between the two eyes from an image and calculating the ratio of the two eyes based on the distance when the breed information of a pet dog according to one embodiment of the present invention corresponds to a mixed breed dog.

[0152] In a method for recommending dog clothing sizes according to an embodiment of the present invention, when the dog breed information corresponds to a mixed dog including two or more dog breeds, the ratio information calculation step further comprises: a step of deriving the distance between the eyes with minimal change in skeletal size according to the dog breed from the image information; and a step of deriving the ratio of the distance between the eyes to the ratio of the total body length; and the actual length calculation step may include a body length correction step of calculating a correction value based on the ratio of the eyes to the average value of a plurality of actual total body lengths derived based only on the weight information and body shape information in the database (15), and calculating the actual torso circumference, actual torso length, and actual neck circumference of the mixed dog, respectively, based on the correction value.

[0154] With the increase in households with pet dogs, the proportion of mixed-breed dogs—which combine the characteristics of two or more breeds rather than being purebreds—is on the rise. In the case of purebreds, standard body shape data by breed is relatively well-established, allowing for high accuracy to be expected through database matching alone. However, for mixed-breed dogs, body proportions can differ significantly from the standard body shape depending on the parent breeds or factors such as atavism. For example, even among mixed-breed dogs weighing the same 10kg, some may have long legs and a short body, while others may have a very long body like a Dachshund. Therefore, for mixed-breed dogs, it is difficult to derive accurate standard lengths from databases based solely on weight and body shape information.

[0156] To solve this, the present invention can utilize the distance between the eyes as a new correction reference point for mixed-breed dogs, as this part of the body has the least variation in skeletal structure and hardly changes in length even when gaining or losing weight.

[0157] As illustrated in FIG. 9, in the ratio information calculation step (S12), if the dog breed information input in the preceding body shape information derivation step (S10) corresponds to a mixed dog, the left eye and the right eye on the image can be identified as additional feature points through the length derivation model and the pixel distance between them can be measured.

[0158] Next, in the ratio information calculation step (S12), the ratio of the total body length derived earlier to the length between the two eyes can be calculated. In the example of FIG. 9, the ratio of the two eyes is derived as 1:0.2.

[0159] Since the above-mentioned two-eyed standard ratio is a value directly derived from the individual's unique traits without relying on a breed database, it can function as a key standard indicator for estimating the actual dimensions of mixed-breed dogs for which data is scarce.

[0161] The above-mentioned two-eyed standard ratio can be used as a standard measure to determine the specificity of the body proportions of the mixed dog and to accurately calculate the actual body length in the subsequent body length correction step.

[0162] The above body length correction step is a process for correcting the atypical body proportions of a mixed-breed dog, and may correspond to a step of determining the final actual body circumference, actual body length, and actual neck circumference by cross-referencing the average of the total body length in the database with the individual's unique bilateral eye ratio derived earlier. In other words, it corresponds to a process that improves precision in deriving the body dimensions of a pet dog by correcting the length by reflecting the skeletal ratio characteristics of the individual rather than simply relying on average values, and a detailed explanation will be provided later in FIGS. 10 and 11.

[0164] FIG. 10 illustrates an example of calculating the total body length based on both eyes and the average total body length, respectively, by referring to database information according to an embodiment of the present invention.

[0165] According to one embodiment of the present invention, the database stores an actual distance between the eyes, which is the actual length of the distance between the eyes corresponding to the combination of the weight information and the body shape information, and the body length correction step may include: a step of deriving the actual distance between the eyes that matches the weight information and the body shape information by referring to the database, and deriving the total body length based on the eyes by applying the ratio of the two eyes to the actual distance between the eyes; and a step of calculating an average total body length, which is the average of the actual total body lengths of all dog breeds corresponding to the weight information and the body shape information by referring to the database.

[0167] In order to obtain accurate body measurements of a mixed dog for which standard data is lacking, the body length correction step may perform a process of comparing the two-eye standard total body length based on the two-eye standard ratio with the average total body length, which is the average value of the actual total body length stored in the database (15).

[0168] This comparison process serves as a criterion for determining whether the mixed dog falls within the range of normal body proportions or has a unique body shape with an unusually long or short waist, and by adjusting the weights accordingly, it is possible to make precise body measurement corrections even for mixed dogs for which standard data is unavailable.

[0170] As illustrated in FIG. 10 (a), in the body length correction step, an actual distance between the eyes value of 5 cm corresponding to "10 kg, normal body type" can be derived from the database (15). At this time, the derived actual distance between the eyes is skeletal information that hardly changes even if weight is gained or lost, so it can serve as a reliable reference point with little variation according to weight class.

[0171] The above body length correction step takes the actual distance between the two eyes, 5cm, as a reference value and applies the two-eye standard ratio of 1:0.2 derived in the preceding ratio information derivation step (S12) illustrated in FIG. 9 to the two-eye standard total body length It can produce.

[0173] Additionally, as illustrated in Fig. 10 (b), the body length correction step can calculate an average total body length of 16 cm, which is the average of the actual total body length data of all dog breeds corresponding to "10 kg, normal body type" in the database (15). This is length information that is universally possessed by dogs of the weight class to which the mixed dog belongs, and may correspond to a statistical standard value in which individual specificity is excluded.

[0175] FIG. 11 illustrates an example of calculating the final corrected total body length using the difference between the total body length based on both eyes and the average total body length according to one embodiment of the present invention.

[0176] The body length correction step according to one embodiment of the present invention may include the step of, when the difference value between the average total body length and the total body length based on both eyes exceeds a preset correction threshold value, calculating a corrected total body length by applying a weight based on the difference value to the total body length based on both eyes, and calculating the actual torso circumference, actual torso length, and actual neck circumference based on the corrected total body length.

[0178] In the above body length correction step, the difference between the previously calculated total body length based on both eyes and the average total body length is calculated, and if the difference is smaller than a preset correction threshold, it is determined that the mixed dog falls within the general body type category, and the above average total body length can be determined as the corrected total body length.

[0179] On the other hand, if the difference value exceeds a preset correction threshold, the individual is determined to have a unique body structure, and the final corrected total body length can be derived by assigning a higher weight to the value derived from the individual's inherent proportion.

[0181] As illustrated in FIG. 11, in the body length correction step, the total body length based on both eyes (25 cm) and the average total body length (16 cm) are calculated, and the difference value can be compared with a preset correction threshold value.

[0182] For example, if the above correction threshold is 3cm, the difference value of 9cm exceeds the correction threshold, so in the above body length correction step, this dog can be identified as having a unique body structure with an unusually long waist compared to a typical 8-10kg dog (e.g., Welsh Corgi or Dachshund mix, etc.).

[0183] Accordingly, in the above body length correction step, the corrected total body length can be calculated by assigning a higher weight than the average total body length of 16 cm to the total body length based on both eyes of 25 cm, which reflects the unique characteristics of the dog. Subsequently, the above body length correction step can calculate the actual torso circumference, actual torso length, and actual neck circumference by multiplying ratio information based on this corrected total body length.

[0185] For example, the corrected total body length by applying a weight of 0.8 to the above dual-eye-based total body length of 25cm and a weight of 0.2 to the average total body length of 16cm It can be calculated as.

[0186] And the actual torso circumference length ( ), actual torso length( ) and actual neck circumference( ) can be produced.

[0187] If the average total body length (16cm) was used without correction, the actual torso length is Although short clothes would have been recommended for the mixed dog with a long waist based on the calculation, by undergoing the correction process of the present invention, a measurement of 13.92 cm can be obtained.

[0189] In summary, if the dog breed information included in the input information corresponds to a mixed dog, the server system (1) can perform a process of deriving the distance between eyes and the ratio of the eyes as the dog’s unique skeletal traits through the length derivation model in the ratio information calculation step (S12) (Fig. 9), and in the body length correction step, perform a process of calculating the total body length based on the eyes and the average total body length, which is the average in the database, respectively (Fig. 10). Subsequently, in the body length correction step, by analyzing the difference between the two values ​​and applying different weights depending on whether they exceed a preset correction threshold value to determine the final corrected total body length (Fig. 11), specific body dimensions can be derived even for mixed dogs that deviate from standard data.

[0190] The actual body circumference, actual body length, and actual neck circumference of the mixed breed dog, which are finally derived through the process of FIGS. 9 to 11, are then transmitted to the clothing recommendation step (S14), and the server system (1) can select clothing products suitable for the mixed breed dog by matching them with the seller's clothing products based on the actual body circumference, actual body length, and actual neck circumference.

[0192] According to one embodiment of the present invention, in the case of mixed-breed dogs for which it is difficult to apply standard body length information from a database as is, the total body length is corrected by utilizing the distance between the eyes, which has less change due to skeletal growth and is less affected by hair, as a reference point, thereby enabling high-accuracy estimation of dimensions even for various dog breeds for which standardized data is unavailable.

[0194] FIG. 12 illustrates an example of generating a virtual fitting model using a generative artificial intelligence model in a clothing recommendation step according to an embodiment of the present invention.

[0195] The clothing recommendation step according to one embodiment of the present invention may further include the step of inputting the image information for the selected clothing product into a pre-trained generative artificial intelligence model to convert the pet dog into a 3D model, and generating a virtual fitting model in which the clothing product is combined on the 3D model and providing it to the user.

[0197] The present invention goes beyond merely recommending numerically suitable clothing and can provide virtual fitting modeling utilizing generative artificial intelligence technology so that the user can visually check how the clothing looks when worn.

[0198] As illustrated in FIG. 12, the server system (1) can generate a 3D model of a pet dog by inputting 2D image information included in the input information into a generative artificial intelligence model. In this process, the previously calculated actual body circumference, actual body length, and actual neck circumference are input together as parameters, so that a 3D avatar with a body shape almost identical to that of a real pet dog can be realized.

[0200] Next, the server system (1) can load 3D texture data of the clothing product selected in the clothing recommendation step (S14) and map it onto the 3D model of the pet dog. At this time, by simulating the material characteristics of the clothing with a physics engine, a virtual fitting model can be completed that realistically expresses phenomena such as the clothes wrapping around the body or wrinkling. Furthermore, the user can rotate or zoom in 360 degrees to view the pet dog moving while wearing the clothes through a user terminal.

[0202] FIG. 13 schematically illustrates the internal configuration of a computing device according to one embodiment of the present invention.

[0203] The server system illustrated in FIG. 1 described above may include the components of the computing device (11000) illustrated in FIG. 13.

[0204] As illustrated in FIG. 13, the computing device (11000) may include at least one processor (11100), memory (11200), peripheral interface (11300), input / output subsystem (I / O subsystem) (11400), power circuit (11500), and communication circuit (11600). In this case, the computing device (11000) may correspond to the server system illustrated in FIG. 1.

[0205] The memory (11200) may include, for example, high-speed random access memory, a magnetic disk, SRAM, DRAM, ROM, flash memory, or non-volatile memory. The memory (11200) may include software modules, instruction sets, or various other data required for the operation of the computing device (11000).

[0206] At this time, access to memory (11200) from other components, such as the processor (11100) or peripheral device interface (11300), can be controlled by the processor (11100).

[0207] The peripheral device interface (11300) can connect input and / or output peripheral devices of the computing device (11000) to the processor (11100) and memory (11200). The processor (11100) can perform various functions for the computing device (11000) and process data by executing software modules or instruction sets stored in the memory (11200).

[0208] The input / output subsystem can connect various input / output peripherals to the peripheral interface (11300). For example, the input / output subsystem may include a controller for connecting peripherals such as a monitor, keyboard, mouse, printer, or, if necessary, a touchscreen or sensor to the peripheral interface (11300). According to another aspect, input / output peripherals may be connected to the peripheral interface (11300) without passing through the input / output subsystem.

[0209] The power circuit (11500) can supply power to all or part of the components of the terminal. For example, the power circuit (11500) may include one or more power sources such as a power management system, a battery or alternating current (AC), a charging system, a power failure detection circuit, a power converter or inverter, a power status indicator, or any other components for power generation, management, and distribution.

[0210] The communication circuit (11600) can enable communication with another computing device using at least one external port.

[0211] Alternatively, as described above, the communication circuit (11600) may enable communication with other computing devices by including an RF circuit and transmitting and receiving an RF signal, also known as an electromagnetic signal.

[0212] The embodiment of FIG. 13 is merely an example of a computing device (11000), and the computing device (11000) may have some components shown in FIG. 13 omitted, additional components not shown in FIG. 13 added, or a configuration or arrangement that combines two or more components. For example, a computing device for a communication terminal in a mobile environment may include a touchscreen or sensors, etc., in addition to the components shown in FIG. 13, and the communication circuit (11600) may include a circuit for RF communication of various communication methods (WiFi, 3G, LTE, Bluetooth, NFC, Zigbee, etc.). The components that can be included in the computing device (11000) may be implemented as hardware, software, or a combination of both hardware and software, including one or more integrated circuits specialized for signal processing or applications.

[0213] Methods according to embodiments of the present invention may be implemented in the form of program instructions that can be executed through various computing devices and recorded on a computer-readable medium. In particular, the program according to the present embodiment may be configured as a PC-based program or an application dedicated to a mobile terminal. An application to which the present invention is applied may be installed on a computing device (11000) through a file provided by a file distribution system. For example, the file distribution system may include a file transmission unit (not shown) that transmits the file in response to a request from the computing device (11000).

[0215] The device described above may be implemented as a hardware component, a software component, and / or a combination of a hardware component and a software component. For example, the device and components described in the embodiments may be implemented using one or more general-purpose or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as parallel processors, are also possible.

[0216] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or command the processing unit independently or collectively. Software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave so as to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computing devices and stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0217] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiment, or they may be those known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0219] Although the embodiments have been described above with reference to limited examples and drawings, those skilled in the art can make various modifications and variations from the description above. For example, suitable results may be achieved even if the described techniques are performed in a different order than described, and / or if the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents. Therefore, other implementations, other embodiments, and equivalents to the claims below are also within the scope of the claims.

Claims

Claim 1 A method for recommending dog clothing sizes using deep learning-based dog body shape analysis, performed on a server system comprising one or more processors and one or more memories, wherein the server system comprises clothing information including the torso length, torso circumference, and neck circumference of the corresponding clothing product entered by a seller for each of a plurality of clothing products to be sold; and includes a database in which the actual total body length of the corresponding dog is stored according to combinations of the dog's breed, weight, and body type; and the dog clothing size recommendation method comprises: a body type information derivation step of receiving input information from a user including one or more of the dog's breed information, weight information, and image information, and deriving body type information that classifies the dog's body type into overweight, standard weight, and underweight based on the input information; a length derivation step of inputting the image information into a pre-trained length derivation model to derive the torso circumference, torso length, neck circumference, and total body length of the dog identified in the image information, respectively; a ratio information calculation step of calculating a first ratio information regarding the ratio of the torso circumference to the total body length, a second ratio information regarding the ratio of the torso length to the total body length, and a third ratio information regarding the ratio of the neck circumference to the total body length; and among a plurality of actual total body lengths stored in the database, the A step of calculating actual lengths, which derives a matching total body length that matches the dog breed information, weight information, and body shape information of the user, and calculates the actual torso circumference, actual torso length, and actual neck circumference of the pet dog based on the matching total body length, first ratio information, second ratio information, and third ratio information; and a step of recommending clothing products that correspond to the torso circumference, actual torso length, and actual neck circumference of the pet dog by comparing the clothing torso length, clothing torso circumference, and clothing neck circumference with the clothing torso length, clothing torso circumference, and clothing neck circumference, respectively, and selecting and providing clothing products that correspond to the torso circumference, torso length, and neck circumference of the pet dog.A method for recommending dog clothing sizes, comprising: a step of deriving the distance between the eyes corresponding to the length of a part that reflects the unique traits of the dog and has minimal change in skeletal size according to the dog's breed in the image information, and a step of deriving the ratio of the distance between the eyes to the ratio of the total body length; and further comprising the step of deriving the ratio of the distance between the eyes to the ratio of the total body length; wherein the actual length calculation step comprises a body length correction step of calculating a correction value based on the ratio of the ratio of the eyes to the average value of a plurality of actual total body lengths derived from the database based only on the weight information and body shape information, and calculating the actual torso circumference, actual torso length, and actual neck circumference of the mixed dog, respectively, based on the correction value. Claim 2 A method for recommending dog clothing sizes according to claim 1, wherein the total body length corresponds to the length from a first feature point referring to the top of the dog's head to a second feature point referring to the starting point of the tail, the torso circumference corresponds to the circumference at a third feature point corresponding to the thickest part of the dog's chest, the torso length corresponds to the length measured along the back line from a fourth feature point referring to the starting point where the dog's neck and back meet to a fifth feature point referring to the starting point of the tail, and the neck circumference corresponds to the circumference measured based on a sixth feature point corresponding to the lowest part of the neck where the dog's neck and shoulder are connected. Claim 3 A method for recommending dog clothing sizes according to claim 1, wherein the length derivation model is a deep learning model trained based on training data in which at least one of the total body length, torso circumference length, torso length, and neck circumference length of a corresponding dog is labeled on a plurality of dog images according to the combination of dog breed information, weight information, and body shape information, and the method identifies a first feature point, a second feature point, a third feature point, a fourth feature point, a fifth feature point, and a sixth feature point that serve as reference points for deriving the total body length, torso circumference length, torso length, and neck circumference length from the image information, and derives the total body length, torso circumference length, torso length, and neck circumference length based on the derived plurality of feature points. Claim 4 A method for recommending dog clothing sizes according to claim 1, wherein the length derivation step further comprises: a step of extracting from the image information a first layer corresponding to an outer silhouette including the dog's fur and a second layer corresponding to an inner silhouette excluding the dog's fur; and a step of calculating a fur length compressible by the dog's fur based on the distance difference between the first layer and the second layer. Claim 5 In claim 4, the actual length calculation step comprises the step of calculating the body circumference excluding hair and the neck circumference excluding hair by subtracting the hair length from the actual body circumference length and the actual neck circumference length, respectively, when the hair length exceeds a preset hair length threshold; and the clothing recommendation step comprises selecting clothing products based on the body circumference excluding hair and the neck circumference excluding hair, a method for recommending dog clothing sizes. Claim 6 delete Claim 7 A method for recommending dog clothing sizes according to claim 1, wherein the clothing recommendation step further comprises the step of inputting image information for the selected clothing product into a pre-trained generative artificial intelligence model to convert the dog into a 3D model, and generating a virtual fitting model in which the clothing product is combined on the 3D model and providing it to a user. Claim 8 A dog clothing size recommendation system using deep learning-based dog body shape analysis, comprising one or more processors and one or more memories, wherein for each of a plurality of clothing products to be sold, clothing information including clothing body length, clothing body circumference length, and clothing neck circumference length of the corresponding clothing product entered by the seller; and a database in which the actual total body length of the corresponding dog is stored according to combinations of the dog's breed, weight, and body type; a body type information derivation unit that receives input information from a user including one or more of the dog's breed information, weight information, and image information, and derives body type information classifying the dog's body type into overweight, standard weight, and underweight based on the input information; a length derivation unit that inputs the image information into a pre-trained length derivation model to derive the torso circumference, torso length, neck circumference, and total body length of the dog identified in the image information, respectively; a ratio information calculation unit that calculates a first ratio information regarding the ratio of the torso circumference to the total body length, a second ratio information regarding the ratio of the torso length to the total body length, and a third ratio information regarding the ratio of the neck circumference to the total body length; and among a plurality of actual total body lengths stored in the database, the dog breed information, weight information, and for the user A real length calculation unit that derives a matching total body length that matches body shape information, and calculates the actual torso circumference, actual torso length, and actual neck circumference of the pet dog based on the matching total body length, first ratio information, second ratio information, and third ratio information; and a clothing recommendation unit that compares the actual torso circumference, actual torso length, and actual neck circumference with the clothing torso length, clothing torso circumference, and clothing neck circumference, respectively, and selects and provides clothing products that correspond to the torso circumference, torso length, and neck circumference of the pet dog; wherein, if the breed information corresponds to a mixed dog comprising two or more breeds, the ratio information calculation unit derives the distance between the two eyes corresponding to the length of a part that reflects the unique traits of the pet dog and has minimal change in skeletal size according to the breed of the pet dog in the image information;A pet dog clothing size recommendation system, comprising: a step of deriving a ratio of the two eyes to the ratio of the distance between the two eyes to the total body length; and further performing the step of the actual length calculation unit, which calculates a correction value based on the ratio of the two eyes to the average value of a plurality of actual total body lengths derived from the database based only on the weight information and body shape information, and calculates the actual torso circumference, actual torso length, and actual neck circumference of the mixed dog, respectively, based on the correction value.

Images