DESC:FORM 2
THE PATENTS ACT, 1970
(39 of 1970)
&
THE PATENTS RULES, 2003
COMPLETE SPECIFICATION
(See sections 10; rule 13)

1. TITLE OF THE INVENTION
“A SYSTEM FOR CLASSIFYING BODY SIZE AND A METHOD THEREOF”

2. APPLICANT (S)
(a) NAME: IMERSIVE.IO PRIVATE LIMITED
(b) NATIONALITY: a corporation organized and existing under the laws of India
(c) ADDRESS: NEW NO. 9 (OLD NO. 4/16), GROUND FLOOR CASUARINA DRIVE, KAPLEESWARAR NAGAR, NEELANKARAI, CHENNAI 600115 TAMIL NADU, IN

3. PREAMBLE TO THE DESCRIPTION

The following specification particularly describes the invention and the manner in which it is to be performed:

A SYSTEM FOR CLASSIFYING BODY SIZE AND A METHOD THEREOF

TECHNICAL FIELD
[0001] The embodiments herein generally relate to computer vision, and more particularly, to a system and a method for classifying body sizes.
BACKGROUND
[0002] Accurate body measurements are critical for a variety of applications, particularly in the apparel and fashion industries. However, obtaining precise 3D body measurements using a camera poses significant technical challenges. One of the primary issues is the sensitivity of body measurement outputs to the pose of the user in front of the camera. Traditional methods often require users to take multiple photos from different angles. This variability undermines the reliability of the measurements, making it difficult to achieve consistent and accurate results.
[0003] Further, existing solutions often rely on custom 3D scanning setups. These setups typically require dedicated hardware, such as specialized cameras and scanning devices, which are not only expensive but also complex to operate. The need for such equipment limits the accessibility of accurate 3D body measurement technologies to other web-applications over the internet. Additionally, the high cost and technical complexity associated with custom 3D scanning setups make them impractical for widespread use, particularly where ease of use and cost-effectiveness are essential.
[0004] In addition to the challenges of obtaining accurate measurements, there is the technical problem of mapping these measurements to custom sizing classifications used by different brands. Clothing brands have unique size charts and measurement standards, which vary widely. This lack of standardization means that the same body measurements can correspond to different sizes across various brands. The task of translating precise body measurements into brand-specific size labels requires complex data processing and the application of sophisticated algorithms to ensure accuracy. This problem is compounded by the need to handle a diverse range of sizing charts and measurement terminologies, further complicating the process of providing reliable size recommendations. As a result, there is a need for a method that not only captures accurate 3D body measurements but also effectively maps these measurements to the customized sizing classifications.
[0005] Accordingly, there remains a need for a more efficient method for mitigating and/or overcoming drawbacks associated with current methods.
SUMMARY
[0006] This section is provided to introduce certain objects and aspects of the present disclosure in a simplified form that are further described below in the detailed description. This summary is not intended to identify the key features or the scope of the claimed subject matter.
[0007] It is an object of the present disclosure to extract anthropometric measurements from a single video clip and classify the extracted measurements.
[0008] It is another object of the present disclosure to classify body sizes based on customized sizing system.
[0009] It is another object of the present disclosure to provide an automated measurement process tailored specifically for the creation of bespoke or made-to-measure garments.
[00010] It is another object of the invention to provide an enhanced body size classification system that categorizes body sizes for both ready-to-wear garments as well as made-to-measure custom clothing.
[00011] It is another object of the present disclosure to provide a system that accommodates diverse body shapes, improves sizing accuracy and enables personalization.
[00012] It is yet another object of the present disclosure to provide a method for defining any type of body measurement necessary for garment manufacturing, particularly for clothing brands that create innovative product designs and new garment types, where custom or made-to-measure clothing is offered as a premium solution.
[00013] It is yet another object of the present disclosure to provide an enhanced input data control system that integrates real-time pose detection to ensure that the user submits input such as a video clip in the correct and intended manner.
[00014] Conventional methods either require multiple poses taken from different angles or custom 3D scanning setups, which are complex to operate and often provide inaccurate results. In addition to the challenges of obtaining accurate measurements, there is the technical problem of mapping these measurements to custom sizing classifications used by different brands. Further, these methods often fail to accommodate diverse body shapes, leading to poor fitness and high return rates.
[00015] To address these challenges, the present disclosure provides an automated body size classification system that extracts any custom body measurement from a single video clip and categorizes body sizes for both ready-to-wear garments as well as made-to-measure custom clothing. A custom body measurement refers to any specific measurement taken from an individual’s body for purposes such as tailoring or garment design aimed at creating a garment that fits them perfectly. Further, the present disclosure classifies body sizes based on customized sizing system. The present disclosure introduces an innovative system designed to accommodate a wide range of body shapes, enhance sizing precision and support personalized garment fitting. Additionally, the present disclosure utilizes an input data control system that checks the camera image feed in real-time using off-the-shelf pose detection software and ensures accurate capture of the input video. Furthermore, the present disclosure provides a system and method for defining any anthropometric measurement needed for garment manufacturing, particularly supporting brands offering innovative or made-to-measure clothing.
[00016] In an aspect of the present disclosure, a system for classifying body size may be provided. The system may include one or more user devices configured to receive/record input data provided by a user, a server configured to receive the one or more input data provided by the one or more user devices, extract anthropometric measurements from the one or input data and generate a multi-class classification model based on the extracted anthropometric measurements and classify body size by one or more classification machine learning model.
[00017] In another embodiment of the present disclosure, the server may be configured to receive the atleast one input video and one or more physical dimensions of the user to the one or more user devices, extract a plurality of frames from the input video, down-sample and normalize each frame of the input video to a predetermined resolution, extract estimated anthropometric measurements from each of the plurality of normalized frames, validate the plurality of normalized frames by filtering-out one or more outlier frames, wherein the outlier frames includes frames with significant deviations from estimated anthropometric measurements and extract aggregated anthropometric measurements from the one or more validated frames and classify body size by the one or more classification machine learning model.
[00018] In another embodiment of the present disclosure, the server may be configured to filter the outlier frames from the plurality of normalized frames using the multi-class classification model to obtain the one or more validated frames.
[00019] In another embodiment of the present disclosure, the server is configured to extract estimated anthropometric measurements from each of the plurality of normalized frames using a skinned multi-person linear (SMPL) model.
[00020] In another embodiment of the present disclosure, the one or more user devices may include a computer vision algorithm for capturing the video, wherein the computer vision algorithm may be refined based on the discrepancies between the estimated anthropometric measurements and the aggregated anthropometric measurements.
[00021] In another embodiment of the present disclosure, the server may be configured to generate the multi-class classification model from the anthropometric measurements of each of the plurality of normalized frames, train the one or more classification machine learning model by a first dataset of standardized size information, a second dataset of synthetic data and a sizing classification, wherein the synthetic data includes simulated/generated anthropometric measurements.
[00022] In another embodiment of the present disclosure, the multi-class classification model regmay be a multi-dimensional multi-class classification model.
[00023] In another embodiment of the present disclosure, the server may further be configured to prompt a user at one or more user devices to receive/record brand specific input data provided by the user, receive new rules provided by the user, update a standardized brand specific sizing chart in a database based on the new rules, generate synthetic data based on a standardized brand specific sizing chart, training the classification machine learning model based on the generated synthetic data and classifying based on the classification machine learning model.
[00024] In another aspect of the present disclosure. a method for classifying body size may include prompting a user at one or more user devices to input receive/record input data provided by a user, receiving the one or more input data provided by the one or more user devices, extracting anthropometric measurements from the one or more input data, generating a multi-class classification model based on the extracted anthropometric measurements and classifying body size by a classification machine learning model.
[00025] In another aspect of the present disclosure, the server may be configured to receive the atleast one input video/photo and one or more physical dimensions, the server may be configured to extract a plurality of frames from the input video, down-sample and normalize each frame of the input video to a predetermined resolution, extract the estimated anthropometric measurements from each of the plurality of normalized frames, validate the plurality of normalized frames by filtering-out one or more outlier frames, wherein the outlier frames includes frames with significant deviations from estimated anthropometric measurements and extract aggregated anthropometric measurements from the one or more validated frames and classify body size by one or more classification machine learning model.
[00026] In another aspect of the present disclosure, the server may be configured to filter the outlier frames from the plurality of normalized frames using the frame classification model to obtain the one or more validated frames.
[00027] In another aspect of the present disclosure, the server may be configured to extract estimated anthropometric measurements from each of the plurality of normalized frames using a 3D body modeling technique.
[00028] In another aspect of the present disclosure, the server may be configured to extract estimated anthropometric measurements from each of the plurality of normalized frames using a skinned multi-person linear (SMPL) model.
[00029] In another aspect of the present disclosure, the server may be configured to generate the frame classification model from the anthropometric measurements of each of the plurality of normalized frames and train the classification machine learning model by a first dataset of standardized size information, a second dataset of synthetic data and a sizing classification, wherein the synthetic data includes simulated/generated anthropometric measurements.
[00030] In another aspect of the present disclosure, the one or more user devices includes a computer vision algorithm for capturing the video, wherein the computer vision algorithm is refined based on the discrepancies between the estimated anthropometric measurements and the aggregated anthropometric measurements.
[00031] In another aspect of the present disclosure, the multi-class classification model is a multi-dimensional multi-class classification model.
[00032] In another aspect of the present disclosure, the classification machine learning model is a multi-class classification machine learning model.
[00033] In another aspect of the present disclosure, the server is further configured to prompt a user, at one or more user devices 102a-n to receive/record brand specific input data provided by the user, receive new rules provided by the user, update a standardized brand specific sizing chart in a database based on the new rules, generate synthetic data based on a standardized brand specific sizing chart, training the classification machine learning model based on the generated synthetic data and classifying based on the classification machine learning model.

BRIEF DESCRIPTION OF THE DRAWINGS
[00034] The embodiments herein will be better understood from the following detailed description with reference to the drawings, in which:
[00035] FIG. 1 is a block diagram that illustrates a system for automatically extracting anthropometric measurements from a single video-clip for classifying body size within customized size classification according to some embodiments herein;
[00036] FIG. 2 is an exploded view of a server of FIG. 1 according to some embodiments herein;
[00037] FIG. 3 is a schematic diagram that illustrates a process for automatically extracting anthropometric measurements from single video-clip according to some embodiments herein;
[00038] FIG. 4 is a flow diagram that illustrates a method for classifying body size according to some embodiments herein; and
[00039] FIG. 5 is a schematic diagram of a computer architecture in accordance with the embodiments herein.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS
[00040] The embodiments herein and the various features and advantageous details thereof are explained more fully with reference to the non-limiting embodiments that are illustrated in the accompanying drawings and detailed in the following description. Descriptions of well-known components and processing techniques are omitted so as to not unnecessarily obscure the embodiments herein. The examples used herein are intended merely to facilitate an understanding of ways in which the embodiments herein may be practiced and to further enable those of skill in the art to practice the embodiments herein. Accordingly, the examples should not be construed as limiting the scope of the embodiments herein.
[00041] The term “customized size classification” refers to a customized categorization of body sizes categorizes based on unique size charts and measurement standards of an entity. For example, a first customized size classification may use class names such as "Small," "Medium," and "Large," while a second customized size classification may use class names such as "1," "2," and "3.". Even though the class names may be same across two customized size classifications, the measurement standards specific to each brand may be different. For example, a class name “XL” for a first customized size classification may stand for 41.5 cm to 43.5 cm and for a second customized size classification may stand for 42 cm to 44 cm.
[00042] The term “anthropometric measurements” refers to quantitative assessments or dimensional data of the human body's size, proportions, and composition, used to evaluate health, nutritional status, and growth patterns.
[00043] The term “normalized frames” refers to frames obtained by downsampling each frame in the video clip to a standard resolution.
[00044] The term “validated frames” refers to frames that remain after automatically filtering out outlier frames at a dimension level using a multi-dimensional multi-class classification model.
[00045] The term “synthetic population anthropometric measurements” refers to simulated body measurements representing a diverse population, used for training the multi-class classification machine learning model.
[00046] The term “aggregated anthropometric measurements” refers to combined anthropometric measurements from the validated frames, used as input for the multi-class classification machine learning model to classify body size.
[00047] As mentioned, there remains a need for automatically extracting anthropometric measurements from single video-clip for classifying body size within customized size classification. Referring now to the drawings, and more particularly to FIGS. 1 through 5, where similar reference characters denote corresponding features consistently throughout the figures, there are shown preferred embodiments.
[00048] FIG. 1 is a block diagram that illustrates a system 100 for automatically extracting anthropometric measurements from single video-clip for classifying body size within customized size classification according to some embodiments herein. The system 100 includes one or more user devices 102a-n, a server 104 that are communicatively connected to each other. According to an embodiment, the one or more user devices 102a-n and a server 104 may be communicatively connected to each other using a data communication network 106. In some embodiments, the one or more user devices 102a-n, without limitation, is a camera enabled internet device and may include a mobile phone, a special purpose device having a camera and a display screen, a tablet, a desktop computer, a laptop computer, and the like. The server 104 is configured to prompt a user at a user device 102a-n to provide input data.
[00049] According to an embodiment, the input data includes (a) a video clip of the user turning once clockwise and once counterclockwise, (b) gender and (b) a height of the user.
[00050] According to an embodiment, the user device 102a-n includes an input data control system that checks the video clip in real-time using an off-the-shelf pose detection software to detect if the entire body is visible within the frame and if the video clip has been captured accurately. It intimates the user by providing visual, text and audio prompts to adjust the position of the camera, posture and distance of the user from the camera after analyzing the video clip to ensure that the entire body is visible within the frame and the video clip is captured accurately. The pose detection software provides real time pose detection on the video feed and prompts the user with text, audio feedback to correct their position relative to the camera (e.g., text/voice prompts "MOVE BACK", "PERFECT", "TURN AROUND SLOWLY" etc.,) that guides the user towards an accurate input submission.
[00051] Advantageously, the present disclosure overcomes the drawbacks of conventional systems by providing an accurate and reliable body measurement process that integrated real-time pose detection that ensures that the user submits input such as a video clip in the correct and intended manner. Moreover, this real-time monitoring helps prevent incorrect posture or positioning during capture, which could otherwise lead to measurement errors.
[00052] The server 104 is configured to extract multiple frames from the input video or the video clip. Each frame is then downsampled and normalized to a predetermined resolution.
[00053] The server 104 extracts estimated anthropometric measurements from the validated frames. According to an embodiment, the server 104 is configured to extract estimated anthropometric measurements from each of a plurality of normalized frames using a skinned multi-person linear (SMPL) model.
[00054] The server 104 is configured to validate the normalized frames by filtering-out outlier frames. Outlier frames are frames with significant deviations or unusual outcomes from the estimated anthropometric measurements. Outlier frames are removed as they can introduce errors and inconsistencies into the actual body measurement or anthropometry measurement process. Estimated anthropometric measurements are used to detect and eliminate outlier frames, ensuring that only validated frames are retained for accurate body measurement analysis.
[00055] The server 104 extracts aggregated anthropometric measurements from the validated frames and classifies body size by the one or more classification machine learning model.
[00056] The user devices 102a-n capture the video clip and refines it with a computer vision algorithm for capturing the video, wherein the computer vision algorithm is refined based on the discrepancies between the estimated anthropometric measurements and the aggregated anthropometric measurements.
[00057] The multi-class classification model is generated from the estimated anthropometric measurements extracted from the normalized frames.
[00058] According to an embodiment, the server 104 is configured to generate a multi-dimensional multi-class classification model using a machine learning (ML) model which learns from the each of the estimated anthropometric measurements and the aggregated anthropometric measurements.
[00059] The server 104 is configured to automatically filter at a dimension level, outlier frames from the plurality of normalized frames using the multi-dimensional multi-class classification model to obtain validated frames.
[00060] The server 104 is configured to train a multi-class classification ML model using a first dataset of standardized size information, a second dataset of synthetic population anthropometric measurements, and the customized sizing classification. The server 104 is configured to apply the multi-class classification ML model to aggregated anthropometric measurements of the validated frames for automatically classifying body size of a user within a customized size classification.
[00061] The system 100 is of advantage that the system 100 enables accurate non-contact 3D body measurements by utilizing computer vision and machine learning. The generation of a multi-dimensional multi-class classification model using ML further optimizes the process by ensuring that only validated frames are used. The extraction of anthropometric measurements from normalized frames enables a fast and real-time response, significantly improving the overall computational efficiency and reducing latency.
[00062] Additionally, the system 100 requires minimal input from a user, requiring only a single video clip, gender and height, making it simple to use and applicable to all genders without the need for specific pose requirements. The gender and dimensions enable determination of accurate anthropometric measurements. The system 100 accommodates diverse ethnic demographic groups by using synthetic population anthropometric measurements. Further, the system 100 accurately mapping body measurements to custom sizing classifications.
[00063] The system 100 is device-agnostic, functioning seamlessly on any internet-enabled, camera-equipped device via a standard web browser. Further, the system 100 integrates effortlessly with other web-applications using an application programming interface (API), enabling classification of body size within customized size classifications based on accurate 3D body measurements.
[00064] In some embodiments, a convolutional neural network is trained and utilized to recognize human body landmarks.
[00065] In some embodiments, the anthropometric data used for refining the 3D body model is derived from a reference database of body measurements.
[00066] In some embodiments, the dimensions include height, arm length, chest circumference, waist circumference, hip circumference, and leg length. The method of claim 1, further comprising the step of integrating the aggregated anthropometric measurements with a database of clothing sizes to provide size recommendations specific to a clothing brand.
[00067] In some embodiments, the system 100 automatically extracts anthropometric measurements from single video-clip and classifies body size within customized size classification in near real-time, within a total processing time of 3 seconds from the time when the last frame of the video clip is captured.
[00068] In some embodiments, the step of capturing the video clip is performed using a standard web browser interface. The video frames may be transmitted to a cloud-based server for scalable and parallel computation.
[00069] In some embodiments, the video clip is captured using a front-facing camera of a mobile device.
[00070] In some embodiments, the system 100 includes a feedback loop that refines a computer vision algorithm based on discrepancies between estimated and aggregated anthropometric measurements observed over multiple uses.
[00071] FIG. 2 is an exploded view of the server 104 of FIG. 1 according to some embodiments herein. The server104 includes a video and height input module 202, an SMPL measurements extraction module 204, a multi-class classification model generation module 206, an outlier frames filtering module 208, a custom size classifier training module 210 and a body size classification module 212.
[00072] The user enters the input data through the user interface of the display device. As the video is captured or uploaded, the input data control system uses pose detection software to monitor the input in real-time, ensuring the correct posture and alignment. The user devices 102a-n includes a computer vision algorithm for capturing the video which is refined based on the discrepancies between the estimated anthropometric measurements and the aggregated anthropometric measurements.
[00073] The video and height input module 202 is configured to prompt a user, at a user device 102a-n, to input (a) a video clip of the user turning once clockwise and once counterclockwise, (b) a height and (c) gender of the user. In an example, the video clip is recorded at 30 frames per second for 10 seconds, that provides an estimated 300 frames of the user.
[00074] The SMPL measurements extraction module 204 is configured to extract anthropometric measurements from each of a plurality of normalized frames using a skinned multi-person linear (SMPL) model, wherein the plurality of normalized frames is obtained by downsampling each frame in the video clip to a standard resolution.
[00075] The multi-class classification model generation module 206 is configured to generate a multi-dimensional multi-class classification model using machine learning (ML) with the anthropometric measurements of each of the plurality of normalized frames and reference output measurements associated with each anthropometric measurement.
[00076] The outlier frames filtering module 208 is configured to automatically filter at a dimension level, outlier frames from the plurality of normalized frames using the multi-dimensional multi-class classification model to obtain validated frames. The frames with significant deviations from the estimated anthropometric measurements or expected body shape are filtered.
[00077] The custom size classifier training module 210 is configured to train a multi-class classification ML model using a first dataset of standardized size information, a second dataset of synthetic population anthropometric measurements, and the customized sizing classification. The body size classification module 212 is configured to apply the multi-class classification ML model to aggregated anthropometric measurements of the validated frames for automatically classifying body size of a user within a customized size classification.
[00078] FIG. 3 illustrates a process for automatically extracting anthropometric measurements from single video-clip according to some embodiments herein. To obtain a comprehensive set of frames that capture the body of the user from all angles, the process begins by extracting frames from the video clip, where the user turns once clockwise and once counterclockwise. This process generates a 360-degree view of the body of the user. Each frame is then downsampled to a standard resolution to create normalized frames, facilitating consistent processing.
[00079] Next, a skinned multi-person linear (SMPL) model is applied to each of the normalized frames to extract a detailed 3D body model that captures various anthropometric measurements, such as limb lengths and body circumferences, for each frame. The estimated anthropometric measurements provide a comprehensive dataset representing the user's body dimensions from multiple angles.
[00080] Further, the process includes utilizing an estimator machine learning (ML) model that analyzes estimated anthropometric measurements for each frame at a dimension level, considering factors such as deviations from expected measurement ranges and inconsistencies with other frames. Frames that are determined to contribute to measurement errors are automatically filtered out, thereby determining accepted frames or validated frames for further processing.
[00081] Custom body measurements are specific to a person and used to build a custom-made suit, dress, or other clothing. These measurements can be categorized into three types. The first is circumference measurements, which involve measuring around a body part for example, calf circumference, which is important for designing fitted pants or compression garments. The second is linear or straight-line measurements, which are taken from one point to another in a direct line, for instance, the full-sleeve length of a shirt, measured from the shoulder to the wrist. Lastly, there are geodesic or contour-based measurements, which follow the natural curves of the body rather than a straight path. An example is the upper bust circumference which is taken along the contour of the chest above the bust. These different types of measurements allow for a more accurate and personalized fit, which is particularly important in custom clothing and precise garment construction.
[00082] Additionally, the system 100 classifies body size based on a customized classification system. The sizing chart varies according to the brand. Brand-specific sizing for ready-to-wear clothing may be determined by receiving inputs of attributes including brand name, product category, and a sizing chart from the user. The server 104 updates/stores a standardized brand specific sizing chart in a database based on the new rules provided by the user. The attributes undergo standardization using a rule-based method to ensure consistency across measurement nomenclature and units. Subsequently, the sizing chart is utilized to generate synthetic data records representing population anthropometric measurements, integrating regional, ethnic, and gender-wise distributions. The classification machine learning model is trained by synthetic data generated based on a standardized brand specific sizing chart. According to an embodiment, the dataset is then employed to train a multi-class classification machine learning model, associating 3D body measurement vectors with brand-specific sizing labels.
[00083] As brands create their own unique size charts and standards, the system 100 determines appropriate clothing sizes by mapping the user's precise body measurements to each brand’s specific size criteria. The aggregated anthropometric measurements obtained from the one or more validated frames are converted into the correct size for each specific brand. The body size is classified based on aggregated anthropometric measurements by the one or more classification machine learning models.
[00084] In some embodiments, an apparatus for determining brand-specific sizing for ready-to-wear clothing may include (a) a database comprising size charts provided by brands, code modules implementing configurable data standardization rules to adapt to brand-specific requirements, (b) an extensible system capable of accommodating new rules as brands are onboarded, (c) code modules for synthesizing records, scalable based on configuration parameters, (d) proprietary configuration parameters ensuring synthesized training data aligns with logical, plausible human population values, (e) a machine learning program constructing a multi-class classification ML model, with saved neural network weights generating brand-specific ML models, (f) deployment module for brand-specific model data files on a cloud service, accessible via a secure web API.
[00085] FIG. 4 is a flow diagram that illustrates a method for classifying body size according to some embodiments herein. At step 402, the method begins with prompting a user, at a user device receive/record input data provided by a user. The input data includes(a) a video clip of the user turning once clockwise and once counterclockwise, (b) gender and (b) a height of the user. The user devices 102a-n includes a computer vision algorithm for capturing the video and a pose detection algorithm for obtaining accurate input from the user.
[00086] At step 404, the input data provided by the user devices 102a-n is received by the server 104.
[00087] At step 406, aggregated anthropometric measurements are extracted from the input data. This step involves extracting a plurality of frames from the input video, down-sampling and normalizing each of the extracted frames using a skinned multi-person linear (SMPL) model to a standard resolution.
[00088] At a step 408, a multi-dimensional multi-class classification model using machine learning (ML) is generated with the anthropometric measurements of each of the plurality of normalized frames and reference output measurements associated with each anthropometric measurement. The outlier frames are filtered from the plurality of normalized frames using the multi-dimensional multi-class classification model to obtain validated frames. According to an embodiment, the outlier frames are filtered from the plurality of normalized frames using the multi-class classification model is a multi-dimensional multi-class classification model. This is followed by training a multi-class classification ML model using a first dataset of standardized size information, a second dataset of synthetic population anthropometric measurements, and the customized sizing classification.
[00089] At a step 410, the method includes applying the classification machine learning model (ML) model to aggregated anthropometric measurements obtained from the validated frames for automatically classifying body size of a user within a customized size classification.
[00090] According to an embodiment, the classification machine learning model is a multi-class classification machine learning model. The body size is classified based on a customized classification system. As the sizing chart varies according to the brand, the server 104 updates/stores a standardized brand specific sizing chart in a database based on the new rules provided by the user. The attributes undergo standardization using a rule-based method to ensure consistency across measurement nomenclature and units. Subsequently, the sizing chart is utilized to generate synthetic data records representing population anthropometric measurements, integrating regional, ethnic, and gender-wise distributions. The classification machine learning model is trained by synthetic data generated based on a standardized brand specific sizing chart. According to an embodiment, the dataset is then employed to train a multi-class classification machine learning model, associating 3D body measurement vectors with brand-specific sizing labels.
[00091] Advantageously, accurate non-contact 3D body measurements by utilizing computer vision and machine learning. The generation of a multi-dimensional multi-class classification model using ML further optimizes the process by ensuring that only validated frames are used. The extraction of anthropometric measurements from normalized frames enables a fast and real-time response, significantly improving the overall computational efficiency and reducing latency.
[00092] Additionally, the method requires minimal input from a user, requiring only a single video clip and height, making it simple to use and applicable to all genders without the need for specific pose requirements. The method accommodates diverse ethnic demographic groups by using synthetic population anthropometric measurements. Further, the method accurately mapping body measurements to custom sizing classifications.
[00093] The method is device-agnostic, functioning seamlessly on any internet-enabled, camera-equipped device via a standard web browser. Further, the method integrates effortlessly with other web-applications using an application programming interface (API), enabling classification of body size within customized size classifications based on accurate 3D body measurements.
[00094] In some embodiments, a convolutional neural network is trained and utilized to recognize human body landmarks.
[00095] In some embodiments, the anthropometric data used for refining the 3D body model is derived from a reference database of body measurements.
[00096] In some embodiments, the dimensions include height, arm length, chest circumference, waist circumference, hip circumference, and leg length. The method of claim 1, further comprising the step of integrating the compiled body measurements with a database of clothing sizes to provide size recommendations specific to a clothing brand.
[00097] In some embodiments, the method automatically extracts anthropometric measurements from single video-clip and classifies body size within customized size classification in near real-time, within a total processing time of 3 seconds from the time when last frame of the video clip is captured.
[00098] In some embodiments, the step of capturing the video clip is performed using a standard web browser interface. The video frames may be transmitted to a cloud-based server for scalable and parallel computation.
[00099] In some embodiments, the video clip is captured using a front-facing camera of a mobile device.
[000100] In some embodiments, the method includes a feedback loop that refines a computer vision algorithm based on discrepancies between estimated and actual measurements observed over multiple uses.
[000101] The embodiments herein may include a computer program product configured to include a pre-configured set of instructions, which when performed, can result in actions as stated in conjunction with the methods described above. In an example, the pre-configured set of instructions can be stored on a tangible non-transitory computer readable medium or a program storage device. In an example, the tangible non-transitory computer readable medium can be configured to include the set of instructions, which when performed by a device, can cause the device to perform acts similar to the ones described here. Embodiments herein may also include tangible and/or non-transitory computer-readable storage media for carrying or having computer executable instructions or data structures stored thereon.
[000102] Generally, program modules utilized herein include routines, programs, components, data structures, objects, and the functions inherent in the design of special-purpose processors, etc. that perform particular tasks or implement particular abstract data types. Computer executable instructions, associated data structures, and program modules represent examples of the program code means for executing steps of the methods disclosed herein. The particular sequence of such executable instructions or associated data structures represents examples of corresponding acts for implementing the functions described in such steps.
[000103] The embodiments herein can include both hardware and software elements. The embodiments that are implemented in software include but are not limited to, firmware, resident software, microcode, etc.
[000104] A data processing system suitable for storing and/or executing program code will include at least one processor coupled directly or indirectly to memory elements through a system bus. The memory elements can include local memory employed during actual execution of the program code, bulk storage, and cache memories which provide temporary storage of at least some program code in order to reduce the number of times code must be retrieved from bulk storage during execution.
[000105] Input/output (I/O) devices (including but not limited to keyboards, displays, pointing devices, etc.) can be coupled to the system either directly or through intervening I/O controllers. Network adapters may also be coupled to the system to enable the data processing system to become coupled to other data processing systems or remote printers or storage devices through intervening private or public networks. Modems, cable modem and Ethernet cards are just a few of the currently available types of network adapters.
[000106] A representative hardware environment for practicing the embodiments herein is depicted in FIG. 5, with reference to FIGS. 1 through 4. This schematic drawing illustrates a hardware configuration of a server/computer system/user device in accordance with the embodiments herein. The viewer device 104 includes at least one processing device 10. The special-purpose CPUs 10 are interconnected via system bus 12 to various devices such as a random-access memory (RAM) 14, read-only memory (ROM) 16, and an input/output (I/O) adapter 18. The I/O adapter 18 can connect to peripheral devices, such as disk units 11 and tape drives 13, or other program storage devices that are readable by the system. The viewer device 104 can read the inventive instructions on the program storage devices and follow these instructions to execute the methodology of the embodiments herein. The viewer device 104 further includes a user interface adapter 19 that connects a keyboard 15, mouse 17, speaker 24, microphone 22, and/or other user interface devices such as a touch screen device (not shown) to the bus 12 to gather user input. Additionally, a communication adapter 20 connects the bus 12 to a data processing network 25, and a display adapter 21 connects the bus 12 to a display device 23, which provides a graphical user interface (GUI) 29 of the output data in accordance with the embodiments herein, or which may be embodied as an output device such as a monitor, printer, or transmitter, for example. Further, a transceiver 26, a signal comparator 27, and a signal converter 28 may be connected with the bus 12 for processing, transmission, receipt, comparison, and conversion of electric or electronic signals.
[000107] The present disclosure offers numerous advantages related to a system and method for extracting anthropometric measurements from a single video clip and classifying the extracted measurements. A few of the advantages achieved using the features of the present disclosure are provided below:
[000108] The present disclosure extracts body measurements anthropometric measurements from a single video clip and classifies body sizes based on customized sizing system.
[000109] The present disclosure provides a system and method to extract any custom body measurement from a single video clip.
[000110] The present disclosure provides an automated measurement process tailored specifically for the creation of bespoke or made-to-measure garments.
[000111] The present disclosure provides an enhanced body size classification system that categorizes body sizes for both ready-to-wear garments as well as made-to-measure custom clothing.
[000112] The present disclosure provides a system that accommodates diverse body shapes, improves sizing accuracy and enables personalization.
[000113] The present disclosure provides a method for defining any type of body measurement necessary for garment manufacturing, particularly for clothing brands that create innovative product designs and new garment types, where custom or made-to-measure clothing is offered as a premium solution.
[000114] The present disclosure provides an enhanced input data control system that integrates real-time pose detection to ensure that the user submits input such as a video clip in the correct and intended manner.
[000115] The present disclosure utilizes an input data control system that checks the camera image feed in real-time using off-the-shelf pose detection software and ensures accurate capture of the input video.
[000116] Furthermore, the present disclosure provides a system and method for defining any anthropometric measurement needed for garment manufacturing, particularly supporting brands offering innovative or made-to-measure clothing.
[000117] The foregoing description of the specific embodiments will so fully reveal the general nature of the embodiments herein that others can, by applying current knowledge, readily modify and/or adapt for various applications such specific embodiments without departing from the generic concept, and, therefore, such adaptations and modifications should and are intended to be comprehended within the meaning and range of equivalents of the disclosed embodiments. It is to be understood that the phraseology or terminology employed herein is for the purpose of description and not of limitation. Therefore, while the embodiments herein have been described in terms of preferred embodiments, those skilled in the art will recognize that the embodiments herein can be practiced with modification within the spirit and scope.

Dated this 20th day of May 2025

Signature:
Name: Mahalakshmi S [IN/PA-4670]
Of KRIA Law
Agent for Applicant
,CLAIMS:We Claim
1. A system (100) for classifying body size, the system comprising:
one or more user devices (102a-n) configured to receive/record input data provided by a user;
a server 104 configured to:
receive the one or more input data provided by the one or more user devices (102a-n);
extract aggregated anthropometric measurements from the one or input data;
generate a multi-class classification model based on the extracted aggregated anthropometric measurements; and
classify body size by one or more classification machine learning model.

2. The system (100) for classifying body size as claimed in claim 1, wherein the server (104) is configured to:
receive the atleast one input video, gender and one or more physical dimensions of the user from the one or more user devices (102a-n);
extract a plurality of frames from the input video;
down-sample and normalize each frame of the input video to a predetermined resolution;
extract estimated anthropometric measurements from each of the plurality of normalized frames;
validate the plurality of normalized frames by filtering-out one or more outlier frames, wherein the outlier frames includes frames with significant deviations from estimated anthropometric measurements;
extract aggregated anthropometric measurements from the one or more validated frames; and
classify body size based on aggregated anthropometric measurements by the one or more classification machine learning model.

3. The system (100) for classifying body size as claimed in claim 2, wherein the server (104) is configured to filter the outlier frames from the plurality of normalized frames using the multi-class classification model to obtain the one or more validated frames.

4. The system (100) for classifying body size as claimed in claim 2, wherein the server (104) is configured to:
extract estimated anthropometric measurements from each of the plurality of normalized frames using a skinned multi-person linear (SMPL) model.

5. The system (100) for classifying body size as claimed in claim 2, wherein the one or more user devices (102a-n) includes a computer vision algorithm for capturing the video, wherein the computer vision algorithm is refined based on the discrepancies between the estimated anthropometric measurements and the aggregated anthropometric measurements.

6. The system (100) for classifying body size as claimed in claim 2, wherein the server (104) is configured to:
generate the multi-class classification model from the estimated anthropometric measurements of each of the plurality of normalized frames;
train the one or more classification machine learning model by a first dataset of standardized size information, a second dataset of synthetic data and a sizing classification, wherein the synthetic data includes simulated/generated anthropometric measurements.

7. The system (100) for classifying body size as claimed in claim 1, wherein the multi-class classification model is a multi-dimensional multi-class classification model.

8. The system (100) for classifying body size as claimed in claim 5, wherein the server (104) is further configured to:
prompt a user, at one or more user devices (102a-n) to receive/record brand specific input data provided by the user;
receive new rules provided by the user;
update a standardized brand specific sizing chart in a database based on the new rules;
generate synthetic data based on a standardized brand specific sizing chart;
training the classification machine learning model based on the generated synthetic data; and
classifying based on the classification machine learning model.

9. A method for classifying body size, the method comprising:
prompting a user, at one or more user devices (102a-n) to input receive/record input data;
receiving the one or more input data provided by the one or more user devices (102a-n) by the server (104);
extracting anthropometric measurements from the one or more input data;
generating a frame classification model based on the extracted anthropometric measurements; and
classifying body size by a classification machine learning model.

10. The method for classifying body size, as claimed in claim 9, wherein the server (104) is configured to receive the atleast one input video/photo and one or more physical dimensions, wherein the server (104) is configured to:
extract a plurality of frames from the input video;
down-sample and normalize each frame of the input video to a predetermined resolution;
extract estimated anthropometric measurements from each of the plurality of normalized frames;
validate the plurality of normalized frames by filtering-out one or more outlier frames, wherein the outlier frames includes frames with significant deviations from estimated anthropometric measurements; and
extract aggregated anthropometric measurements from the one or more validated frames; and
classify body size by one or more classification machine learning model.

11. The method for classifying body size, as claimed in claim 10, wherein the server (104) is configured to filter the outlier frames from the plurality of normalized frames using the multi-class classification model to obtain the one or more validated frames.

12. The method for classifying body size, as claimed in claim 10, wherein the server (104) is configured to:
extract estimated anthropometric measurements from each of the plurality of normalized frames using a skinned multi-person linear (SMPL) model.

13. The method for classifying body size, as claimed in claim 10, wherein the server (104) is configured to:
generate the frame classification model from the anthropometric measurements of each of the plurality of normalized frames; and
train the classification machine learning model by a first dataset of standardized size information, a second dataset of synthetic data and a sizing classification, wherein the synthetic data includes simulated/generated anthropometric measurements.

14. The method for classifying body size as claimed in claim 9, wherein the one or more user devices (102a-n) includes a computer vision algorithm for capturing the video, wherein the computer vision algorithm is refined based on the discrepancies between the estimated anthropometric measurements and the aggregated anthropometric measurements

15. The method for classifying body size as claimed in claim 9,
wherein the multi-class classification model is a multi-dimensional multi-class classification model;
wherein the classification machine learning model is a multi-class classification machine learning model.

16. The method for classifying body size as claimed in claim 13, wherein the server (104) is further configured to:
prompt a user, at one or more user devices (102a-n) to input receive/record brand specific input data provided by the user;
receive new rules provided by the user;
update a standardized brand specific sizing chart in a database based on the new rules;
generate synthetic data based on a standardized brand specific sizing chart;
training the classification machine learning model based on the generated synthetic data; and
classifying based on the classification machine learning model.

Dated this 20th day of May 2025

Signature:
Name: Mahalakshmi S [IN/PA-4670]
Of KRIA Law
Agent for Applicant