System and method for fashion design of garments

Abstract

The present application belongs to the technical field of clothing intelligent manufacturing, and discloses a clothing fashion design system and method, which comprises a fashion collection unit, a fashion processing unit, a fashion design unit and a fashion generation unit; the fashion collection unit is used for collecting various kinds of fashion elements; the fashion processing unit is used for statistically processing and learning the fashion elements by using a deep learning algorithm; the fashion design unit is used for separating and reclassifying clothing attributes representing the fashion elements; and the fashion generation unit is used for generating clothes by referring to semantics or environmental conditions. According to the present application, clothing styles can be automatically designed according to artificial intelligence algorithms by inputting constraint conditions, thereby saving a large amount of time and manpower. In the fashion field, clothes can be freely customized according to the needs of users, the design cost is greatly reduced, and the design efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of intelligent clothing, and more particularly, to a clothing fashion design system and method. BACKGROUND

[0002] Currently, in the field of fashion design, a designer usually designs new clothes through his own knowledge and experience, and a large amount of time and effort is consumed each time a clothing is designed, and it is impossible for a designer to design every style of clothing required. Therefore, in the fashion field, there is a potential and huge application scenario for the design of clothes according to requirements.

[0003] A private custom fashion clothing design system disclosed in Chinese Patent No. CN108606384A has a sliding block slidingly installed on the outside of the left and right vertical plates, which quickly measures the limb width of the customer at different heights to design the clothing. This method only proposes a design device without a fashion design method. Chinese Patent No. CN201910495710.9 "Cognitive Automation and Interactive Personalized Fashion Design" uses a computer device to train a computer model using computer vision based on deep learning, uses a cognitively determined fashion score (F-score) to identify, uses a computer model and identification to create a new fashion design. This method does not integrate various style features, and the designed clothing has limitations. SUMMARY

[0004] The technical problem to be solved by the present application is to provide a clothing fashion design system and method, which can realize clothing fashion design based on artificial intelligence, greatly improve design efficiency, and reduce clothing design difficulty.

[0005] To solve the above technical problems, the technical solution adopted by the present application is: a fashion design system, comprising a fashion collection unit, a fashion processing unit, a fashion design unit and a fashion generation unit;

[0006] The fashion collection unit is used to collect various types of fashion elements;

[0007] The fashion processing unit is used to statistically and learn the fashion elements using a deep learning algorithm;

[0008] The fashion design unit is used to separate and reclassify the clothing attributes representing the fashion elements;

[0009] The fashion generation unit is used to generate clothing by referring to semantics or environmental conditions.

[0010] In a preferred solution, the fashion processing unit comprises a training module, in which the collected fashion elements are established into a three-dimensional model, and the three-dimensional model is manually labeled, the manually labeled model is parameter adapted according to input parameters to form a sample set, the sample set is sent to a deep learning algorithm for training to obtain an artificial intelligence model.

[0011] In a preferred solution, the fashion design unit comprises an artificial intelligence model for operating the input three-dimensional model and input parameters, the artificial intelligence model gives a fashion evaluation, and the data is transmitted to the fashion generation unit.

[0012] In a preferred solution, the fashion design unit comprises an artificial intelligence model for operating the input three-dimensional model and input parameters, the artificial intelligence model gives a fashion evaluation, and the data is transmitted to the fashion generation unit.

[0013] In a preferred solution, the element decomposition includes element disassembly and plane mapping, the disassembled elements and the mapped plane graphics are combined and processed in a combiner after one or more operations of shape, pattern, color, position, scaling, flipping, twisting, and array, and then three-dimensional reconstruction is performed to realize fusion and reorganization of design elements.

[0014] In a preferred solution, the artificial intelligence model comprises a classifier, a semantic recognizer, an adapter, and a clusterer.

[0015] The classifier is used to classify the three-dimensional model, and the classification is associated with the semantic keywords identified by the semantic recognizer.

[0016] The semantic recognizer is used to convert input parameters into semantic keywords, and the semantic keywords are associated with the classification of the classifier.

[0017] The adapter is used to adapt the samples of the three-dimensional model according to the semantic keywords to adapt the three-dimensional model that conforms to the semantic keywords.

[0018] The clusterer is used to aggregate the adapted three-dimensional model according to fashion classification to generate a fashion evaluation.

[0019] In a preferred solution, the fashion collection unit is provided with a three-dimensional model database.

[0020] A design method using the above-mentioned fashion design system, comprising the following steps:

[0021] S1, input a limited parameter, and an artificial intelligence model reads stereoscopic model data;

[0022] The artificial intelligence model is obtained after artificial marking, parameter adaptation and model training;

[0023] S2, fashion evaluation is performed on the stereoscopic model according to the limited parameter;

[0024] S3, the design with a higher fashion vote evaluation is sent to a fashion generation unit;

[0025] Through the above steps, fashion design based on artificial intelligence is realized.

[0026] In the preferred scheme, the step S2 further includes the steps of element decomposition and fusion recombination, thereby generating new samples, and a part of the new samples is used for iteration of the artificial intelligence model after artificial marking, parameter adaptation and model training, and the other part is sent to the artificial intelligence model for processing;

[0027] The steps of element decomposition and fusion recombination include:

[0028] S21, design elements are elementally decomposed, including shape, pattern, color and position;

[0029] The stereoscopic model is mapped to a plane;

[0030] S22, the decomposed elements are combined or combined plane structures through one or more of scaling, flipping, twisting and arranging;

[0031] S23, the combined plane structures are stereoscopically reconstructed to realize fusion recombination.

[0032] In the preferred scheme, the artificial intelligence model includes a classifier, a semantic recognizer, an adapter and a clusterer;

[0033] The classifier is used for classifying the stereoscopic model according to keywords;

[0034] The semantic recognizer is used for converting the input parameter into keywords;

[0035] The adapter is used for associating the classified stereoscopic model according to the keywords;

[0036] The clusterer is used for clustering the associated stereoscopic model according to the fashion evaluation.

[0037] The clothing fashion design system and method provided by the application have the following beneficial effects compared with the prior art:

[0038] (1) The novel fashion design system and method provided by the application can automatically design clothing styles according to artificial intelligence algorithms through constraint condition input, thereby saving a large amount of time and manpower. In the fashion field, clothing can be freely customized according to the needs of users, thereby greatly reducing design costs and improving design efficiency.

[0039] (2) The application adopts three-dimensional model data, which is more intuitive and convenient for customers to select. Moreover, the adoption of three-dimensional model data is also more convenient for subsequent modification and pattern design, thereby facilitating the design and production of high-quality fashion clothing.

[0040] (3) The novel fashion design system and method provided by the application can be applied not only in online applications but also in embedded devices, thereby greatly improving practicability. BRIEF DESCRIPTION OF DRAWINGS

[0041] The application will be further described below in combination with the drawings and embodiments:

[0042] Figure 1 It is a flowchart of the fashion design system and method of the application.

[0043] Figure 2 It is a preferred flowchart of the fashion design system and method of the application.

[0044] Figure 3 It is a flowchart of the application for adding new fashion design samples of clothing.

[0045] Figure 4 It is a functional module diagram of the artificial intelligence model of the application. DETAILED DESCRIPTION

[0046] Embodiment 1:

[0047] In order to make the purpose, technical scheme and advantages of the application more clear and explicit, the application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the application and do not limit the application. In addition, the technical features involved in each embodiment of the application described below can be combined with each other as long as they do not conflict with each other.

[0048] Referring to Figure 1 , a fashion design system includes a fashion collection unit, a fashion processing unit, a fashion design unit and a fashion generation unit;

[0049] The fashion collection unit is used to collect various types of fashion elements;

[0050] The fashion processing unit is used to use a deep learning algorithm to statistically and learn fashion elements;

[0051] The fashion design unit is used to separate and reclassify the clothing attributes representing fashion elements;

[0052] The fashion generation unit is used to generate clothing according to semantic or environmental conditions.

[0053] The preferred scheme is as follows Figure 2 In the preferred scheme, the fashion processing unit includes a training module, in which a three-dimensional model of the collected fashion elements is established. The three-dimensional model can be closer to the real design, and can avoid the defects that the fashion design is good but the pattern design is insufficient in the post-production process. Preferably, the established three-dimensional model is implemented by cutting and three-dimensionally splicing. Although the initial three-dimensional model increases the workload, the subsequent workload becomes very light, and the free combination of various different design fashion elements and the evaluation of the final effect are facilitated. The three-dimensional model is manually marked. The purpose of manual marking is to define the three-dimensional model according to parameters, so as to facilitate the training of the subsequent artificial intelligence model, such as distinguishing between spring and autumn clothes and summer clothes, distinguishing between underwear and outerwear, distinguishing between upper and lower clothes, distinguishing between accessories, distinguishing between body shape fitting degree, and giving fashion evaluation for different combinations.

[0054] The manually marked model is adapted according to the input parameters to form a sample set, which is sent to a deep learning algorithm for training to obtain an artificial intelligence model. In the artificial intelligence model, the fashion evaluation value of the three-dimensional model with different inputs is obtained by adaptation. In the artificial intelligence model, not only is the selection made according to the input parameters, but also the fashion evaluation value is obtained according to the evaluation of the three-dimensional model by the manually marked sample.

[0055] The preferred scheme is as follows Figure 2 In the preferred scheme, the fashion design unit includes an artificial intelligence model that operates on the input three-dimensional model and input parameters, and gives a fashion evaluation, and sends the data to the fashion generation unit. Generally, the fashion generation unit generates a plane or a three-dimensional figure for display according to the order of the fashion evaluation, and adds elements that set off the atmosphere during the display process, such as a human model, accessories outside the three-dimensional model, ambient lighting, flashing light and shadow, etc.

[0056] The preferred scheme is as follows Figure 2In the specific embodiment, the fashion design unit comprises an artificial intelligence model for operating the input three-dimensional model and input parameters, the artificial intelligence model gives a fashion evaluation, decomposes the elements, and then fuses and recombines, the sample generated after the fusion and recombination is divided into two parts, one part is sent to artificial marking, the artificial marking model is adapted according to the input parameters to form a sample set, and the sample set is sent to a deep learning algorithm for training to iterate the artificial intelligence model; the other part is directly sent to the artificial intelligence model for fashion evaluation. Thus, more fashion design samples can be obtained, or more three-dimensional model data can be obtained to obtain more selectable results after subsequent input of corresponding parameters.

[0057] In the preferred embodiment, the fashion design unit comprises an artificial intelligence model for operating the input three-dimensional model and input parameters, the artificial intelligence model gives a fashion evaluation, decomposes the elements, and then fuses and recombines, the sample generated after the fusion and recombination is divided into two parts, one part is sent to artificial marking, the artificial marking model is adapted according to the input parameters to form a sample set, and the sample set is sent to a deep learning algorithm for training to iterate the artificial intelligence model; the other part is directly sent to the artificial intelligence model for fashion evaluation. Thus, more fashion design samples can be obtained, or more three-dimensional model data can be obtained to obtain more selectable results after subsequent input of corresponding parameters. Figure 3 In the specific embodiment, the element decomposition comprises element disassembly and plane mapping, and the disassembled elements and the plane graphics obtained by mapping are combined in the combiner after one or more operations of shape, pattern, color, position, scaling, flipping, twisting, and arraying, and then are subjected to three-dimensional reconstruction to realize fusion and recombination of the design elements. The three-dimensional reconstruction refers to a process of remapping the combined graphics to a three-dimensional model according to the coordinates in the element decomposition. In the preferred embodiment, screening is performed after combination, a plurality of preset rules are set in the set screen, and most of the results that do not conform to the design principle are deleted to reduce the operation amount of the subsequent three-dimensional reconstruction and the operation amount of the artificial intelligence model operation step. For example, fuzzy images, obviously misaligned images, and images with broken graphics are deleted.

[0058] In the preferred embodiment, the fashion design unit comprises an artificial intelligence model for operating the input three-dimensional model and input parameters, the artificial intelligence model gives a fashion evaluation, decomposes the elements, and then fuses and recombines, the sample generated after the fusion and recombination is divided into two parts, one part is sent to artificial marking, the artificial marking model is adapted according to the input parameters to form a sample set, and the sample set is sent to a deep learning algorithm for training to iterate the artificial intelligence model; the other part is directly sent to the artificial intelligence model for fashion evaluation. Thus, more fashion design samples can be obtained, or more three-dimensional model data can be obtained to obtain more selectable results after subsequent input of corresponding parameters. Figure 4 In the specific embodiment, the artificial intelligence model comprises a classifier, a semantic recognizer, an adapter, and a clusterer; thus, the operation amount of each module can be reduced, the complexity of each module can be reduced, and the processing efficiency can be improved.

[0059] The classifier is used for classifying the three-dimensional model, and the classification is associated with the semantic keywords recognized by the semantic recognizer; the classifier can reduce the subsequent operation amount. For example, a certain three-dimensional model belongs to summer clothes, upper garment, white color, silk fabric, and no pattern. After classification, it is convenient to adapt according to the demand of the keywords. The classifier adopts a deep neural network based on CNN or fast-CNN.

[0060] The semantic recognizer is used for converting the input parameters into semantic keywords, and the semantic keywords are associated with the classification of the classifier; that is, the semantic keywords are the basis for classification in the classifier. The semantic recognizer adopts RNNs, that is, a recurrent neural network.

[0061] The adapter is used to adapt the samples of the 3D model according to the semantic keywords, and adapt the 3D model that matches the semantic keywords. The adapter can be regarded as a keyword-based filter. The 3D model data with the keyword match can pass through the adapter, while the 3D model data without the keyword match is removed by the adapter.

[0062] The clusterer is used to aggregate the adapted 3D models according to fashion categories, thereby generating different fashion ratings. This approach yields results that meet the requirements.

[0063] In a preferred embodiment, the fashion acquisition unit is equipped with a 3D model database. The fashion acquisition unit is used to collect images of popular and classic clothing elements. Images can be acquired online using tools such as web crawlers, replacing time-consuming manual downloads. In special locations, such as dances, and in certain regions like Wuhan and Shanghai, cameras are used to capture elements, and a data server stores them in real time. The data server stores standard 3D human body models. Preferably, it can also store typical 3D human body models, such as child models, thin or overweight models, and obese models. Adjustable panels are set on the surface of the 3D human body model. Clothing is constructed using these adjustable panel structures, and clothing images are mapped onto the adjustable panel structures of the clothing according to the acquisition location. Parameters of the panel structures are set, such as shape, pattern, color, texture, and gloss. The advantage of this approach is that the completed fashion clothing design can directly obtain pattern data that can be used for cutting, and this pattern data can be directly applied to corresponding different types of human body models. This significantly reduces subsequent workload and has great commercial value.

[0064] Example 2:

[0065] Based on Example 1, such as Figures 2-3 In this context, a design method employing the aforementioned fashion design system includes the following steps:

[0066] S1. Input limited parameters, including but not limited to body shape, skin color, hairstyle, season, applicable scenario, etc., and the artificial intelligence model reads the 3D model data;

[0067] like Figure 2 In this context, the artificial intelligence model is obtained after manual labeling, parameter adaptation, and model training. Through manual labeling, the artificial intelligence model can obtain more accurate training samples, which helps improve the accuracy of the artificial intelligence model. Since fashion standards involve relatively subjective judgments, manual labeling plays a very important role in the training process of the artificial intelligence model.

[0068] Preferred solutions include Figure 4In some embodiments, the artificial intelligence model comprises a classifier, a semantic recognizer, an adapter, and a clusterer.

[0069] The classifier is configured to classify the three-dimensional model according to the keywords; the classifier is configured to define the three-dimensional model in a classified manner, for example, adding corresponding labels of body type, applicable season, applicable scene, pattern category, color category, texture category, gloss category, whether having elasticity, applicable specific position, for example, position of scarf, position of chest ornament, position of arm ornament, position of waist ornament, fashion evaluation score, and weighted relationship data between the score and corresponding body type, skin color, and hairstyle.

[0070] The semantic recognizer is configured to convert the input parameters into keywords.

[0071] The adapter is configured to associate the classified three-dimensional model according to the keywords; the adapter is configured to select the three-dimensional model that meets the input parameters.

[0072] The clusterer is configured to cluster the associated three-dimensional model according to the fashion evaluation.

[0073] In the clusterer, the planar mapping image data of the three-dimensional model is convolutionally encoded, and the convolutional neural network features of each image are extracted, and the algorithm is learned according to the training data set to have fine-grained classification ability for fashion elements. The decoder composed of the convolutional neural network performs clustering operation on the fashion unit, and guides the generation mode of the fashion garment according to the training data and the standard of the fashion evaluation.

[0074] S2, fashion evaluation of the three-dimensional model according to the limited parameters;

[0075] In a preferred embodiment, the steps of element decomposition and fusion recombination are further included, thereby generating new samples, a part of which is used for iteration of the artificial intelligence model after artificial marking, parameter adaptation, and model training, and another part is sent to the artificial intelligence model for processing.

[0076] The steps of element decomposition and fusion recombination include:

[0077] S21, element decomposition of the design elements, including but not limited to shape, pattern, color, relative position of pattern and garment, material, texture, and gloss;

[0078] Planar mapping of the three-dimensional model; it refers to unfolding the three-dimensional model with cutting planes to obtain planar images, and performing subsequent operations on the planar images.

[0079] S22, the disassembled elements are combined by one or more of scaling, flipping, twisting, mirroring, sharpening, texturing, blurring, and arranging; the planar structure refers to a multi-layer structure composed of multiple planar images stacked together, for example, the bottom layer is a planar image of a garment, the upper layer is a pattern, the upper layer is a texture, and the upper layer is a jewelry, so that the planar images of each layer can be mapped to different structures of the three-dimensional structure.

[0080] S23, the combined planar structure is reconstructed in three dimensions to realize fusion and recombination.

[0081] The surface of the three-dimensional human model is provided with adjustable sheet structures, which means that the position and shape of the sheet structure and the connection structure between the sheets can be adjusted. The adjustable sheet structure is used to construct a garment, and the combined planar structure of the clothing image with clothing elements is mapped to the adjustable sheet structure of the garment according to the collected position. The sheet structure is parameterized, such as shape, pattern, color, texture, and luster, to realize fusion and recombination operation. The completed fashion design can directly obtain pattern data that can be used for cutting.

[0082] S3, the design with a high fashion ticket evaluation is sent to a fashion generation unit; the fashion generation unit generates a planar or three-dimensional graph for display according to the ranking of the fashion evaluation, and adds elements that set off the atmosphere during the display process, such as a human model, accessories outside the three-dimensional model, atmosphere lighting, and flickering light and shadow.

[0083] Through the above steps, the fashion design based on artificial intelligence is realized.

[0084] The above embodiments are only preferred technical solutions of the present application, and should not be regarded as a limitation of the present application. The embodiments in the application and the features in the embodiments can be combined with each other without conflict. The protection scope of the present application should be based on the technical solutions claimed in the claims, including the equivalent replacement solutions of the technical features claimed in the claims. That is, within this range, equivalent replacement improvements are also within the protection scope of the present application.

Claims

1. A fashion design system characterized by: The fashion collection unit, the fashion processing unit, the fashion design unit and the fashion generation unit are included. The fashion collection unit is used to collect various kinds of fashion elements. The fashion processing unit is used to use a deep learning algorithm to statistically analyze and learn the fashion elements. The fashion processing unit includes a training module, in which a three-dimensional model of the collected fashion elements is established, and the three-dimensional model is artificially marked. The fashion design unit is used to separate and reclassify the clothing attributes representing the fashion elements. The fashion design unit includes an artificial intelligence model that operates on the input three-dimensional model and input parameters. The artificial intelligence model gives a fashion evaluation, decomposes the elements, and then recombines them. The recombined sample is divided into two parts. One part is sent to an artificial marking model, which is parameter-adapted according to input parameters to form a sample set. The sample set is sent to a deep learning algorithm for training, and the artificial intelligence model is iterated. The other part is directly sent to the artificial intelligence model for fashion evaluation. Element decomposition includes element disassembly and plane mapping. The design elements are disassembled, including shape, pattern, color, and position. The three-dimensional model is mapped to a plane.

2. The fashion design system of claim 1, wherein: The disassembled elements and the mapped plane graphics are combined through scaling, flipping, twisting, and arranging.

3. The fashion design system of claim 1, wherein: The combined structure is processed in a combiner, and then reconstructed in three dimensions to achieve the fusion and recombination of design elements.

4. A design method using the fashion design system according to any one of claims 1 to 3, characterized by The artificial intelligence model includes a classifier, a semantic recognizer, an adapter, and a clusterer. The classifier is used to classify the three-dimensional model. The semantic keywords identified by the semantic recognizer are associated with the classification of the classifier. The semantic recognizer is used to convert input parameters into semantic keywords. The semantic keywords are associated with the classification of the classifier. The adapter is used to adapt the three-dimensional model samples according to the semantic keywords. The clusterer is used to aggregate the adapted three-dimensional models according to fashion classification to generate fashion evaluation. The fashion generation unit is used to generate clothing based on semantics. The fashion design unit includes an artificial intelligence model that operates on the input three-dimensional model and input parameters. The fashion collection unit has a three-dimensional model database. The method includes the following steps: S1, input the defined parameters, and the artificial intelligence model reads the three-dimensional model data. The artificial intelligence model is obtained after artificial marking, parameter adaptation, and model training. S2, evaluate the three-dimensional model based on the defined parameters. S3, designs with high fashion vote evaluation are sent to the fashion generation unit. The above steps realize fashion design based on artificial intelligence.

5. The design method of claim 4, characterized in that: In step S2, element decomposition and fusion recombination are further included, thereby generating new samples, a part of which is used for iteration of the artificial intelligence model after artificial marking, parameter adaptation and model training, and another part is sent to the artificial intelligence model for processing; The step of element decomposition and fusion recombination includes: S21, element decomposition of design elements, including shape, pattern, color and position; Plane mapping of the three-dimensional model is performed; S22, the decomposed elements are combined through one or more of scaling, flipping, twisting, arranging or combined plane structure; S23, three-dimensional reconstruction of the combined plane structure is performed to realize fusion recombination.

6. The method of claim 4, wherein: The artificial intelligence model includes a classifier, a semantic recognizer, an adapter and a clusterer; The classifier is used to classify the three-dimensional model according to keywords; The semantic recognizer is used to convert input parameters into keywords; The adapter is used to associate the classified three-dimensional model according to the keywords; The clusterer is used to cluster the associated three-dimensional model according to fashion evaluation.

Images