CN114357544A - Fashion design system and method for clothes - Google Patents
Fashion design system and method for clothes Download PDFInfo
- Publication number
- CN114357544A CN114357544A CN202111455629.1A CN202111455629A CN114357544A CN 114357544 A CN114357544 A CN 114357544A CN 202111455629 A CN202111455629 A CN 202111455629A CN 114357544 A CN114357544 A CN 114357544A
- Authority
- CN
- China
- Prior art keywords
- fashion
- model
- design
- artificial intelligence
- elements
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000013461 design Methods 0.000 title claims abstract description 83
- 238000000034 method Methods 0.000 title claims abstract description 23
- 238000013473 artificial intelligence Methods 0.000 claims abstract description 53
- 238000012545 processing Methods 0.000 claims abstract description 15
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 13
- 238000013135 deep learning Methods 0.000 claims abstract description 11
- 230000007613 environmental effect Effects 0.000 claims abstract description 4
- 238000011156 evaluation Methods 0.000 claims description 30
- 238000012549 training Methods 0.000 claims description 23
- 230000004927 fusion Effects 0.000 claims description 16
- 238000002372 labelling Methods 0.000 claims description 11
- 230000006978 adaptation Effects 0.000 claims description 9
- 238000000354 decomposition reaction Methods 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 9
- 238000005215 recombination Methods 0.000 claims description 7
- 230000006798 recombination Effects 0.000 claims description 7
- 230000002596 correlated effect Effects 0.000 claims description 3
- 239000003086 colorant Substances 0.000 claims description 2
- 230000004931 aggregating effect Effects 0.000 claims 1
- 238000004519 manufacturing process Methods 0.000 abstract description 2
- 230000008521 reorganization Effects 0.000 description 9
- 230000008569 process Effects 0.000 description 5
- 238000013527 convolutional neural network Methods 0.000 description 4
- 230000000875 corresponding effect Effects 0.000 description 4
- 239000007787 solid Substances 0.000 description 4
- 230000037237 body shape Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000005520 cutting process Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000005094 computer simulation Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
Description
技术领域technical field
本发明属于智能服装技术领域，更具体地，涉及一种服装时尚设计系统与方法。The invention belongs to the technical field of intelligent clothing, and more particularly, relates to a clothing fashion design system and method.
背景技术Background technique
目前，在时尚设计领域，通常会由设计师通过自己的学识经验来设计新的服饰，而且每次设计一款服饰要消耗大量的时间和精力，设计师也不可能面面俱到的设计出所需的每一种风格的服饰。因此，在时尚领域，对服饰按需求的设计拥有潜在且巨大的应用场景。At present, in the field of fashion design, designers usually design new clothing through their own knowledge and experience, and it takes a lot of time and energy to design a piece of clothing each time, and it is impossible for designers to design everything they need. Every style of clothing. Therefore, in the field of fashion, the design of clothing according to needs has a potential and huge application scenario.
公开号为CN108606384A的中国专利一种私人定制时尚服饰设计系统通过左右两侧所述竖板的外部均滑动安装有滑块，快速的对客户进行不同高度的肢体宽度的测量，来对服装进行设计，该方法只提出了设计设备而没有设计时尚设计方法。申请号为CN201910495710.9的中国专利“认知自动化和交互式个性化时尚设计”采用由计算机设备使用基于深度学习的计算机视觉来训练计算机模型，使用经认知地确定的时尚性得分(F得分)来标识，使用计算机模型和经标识来创建新时尚设计，此方法没有融合各种风格特征，设计出的服饰具有局限性。The Chinese Patent Publication No. CN108606384A is a private custom fashion clothing design system. The outer sides of the vertical plates on the left and right sides are slidably installed with sliders to quickly measure the width of limbs of different heights for customers to design clothing. , this method only proposes design equipment but not fashion design method. The Chinese patent "Cognitive Automation and Interactive Personalized Fashion Design" with application number CN201910495710.9 employs deep learning-based computer vision by computer equipment to train a computer model, using a cognitively determined fashion score (F-score). ) to identify, using computer models and logos to create new fashion designs, this method does not incorporate various stylistic characteristics, and the designed clothing has limitations.
发明内容SUMMARY OF THE INVENTION
本发明所要解决的技术问题是提供一种服装时尚设计系统与方法，能够实现基于人工智能的服装时尚设计，大幅提高设计效率，降低服装设计难度。The technical problem to be solved by the present invention is to provide a clothing fashion design system and method, which can realize the clothing fashion design based on artificial intelligence, greatly improve the design efficiency, and reduce the difficulty of clothing design.
为解决上述技术问题，本发明所采用的技术方案是：一种时尚设计系统，包括时尚采集单元、时尚处理单元、时尚设计单元和时尚生成单元；In order to solve the above-mentioned technical problems, the technical scheme adopted in the present invention is: a fashion design system, including a fashion collection unit, a fashion processing unit, a fashion design unit and a fashion generation unit;
所述时尚采集单元用于采集各个种类的时尚元素；The fashion collection unit is used for collecting various kinds of fashion elements;
所述时尚处理单元用于使用深度学习算法对时尚元素进行统计与学习；The fashion processing unit is used for statistics and learning of fashion elements using a deep learning algorithm;
所述时尚设计单元用于将代表时尚元素的服装属性进行分离与再归类；The fashion design unit is used to separate and reclassify clothing attributes representing fashion elements;
所述时尚生成单元用于参考语义或环境条件来生成服饰。The fashion generation unit is used to generate apparel with reference to semantic or environmental conditions.
优选的方案中，所述的时尚处理单元包括训练模块，在训练模块内，将采集的时尚元素建立立体模型，并将立体模型进行人工标记，对人工标记模型根据输入参数进行参数适配，形成样本集，将样本集送入到深度学习算法进行训练，获得人工智能模型。In a preferred solution, the fashion processing unit includes a training module. In the training module, a three-dimensional model of the collected fashion elements is established, and the three-dimensional model is manually marked, and the manual marked model is parameterized according to the input parameters to form a The sample set is sent to the deep learning algorithm for training to obtain an artificial intelligence model.
优选的方案中，时尚设计单元包括将输入的立体模型和输入参数进行运算的人工智能模型，人工智能模型给出时尚评价，并将数据输送至时尚生成单元。In a preferred solution, 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 fashion evaluation, and sends the data to the fashion generation unit.
优选的方案中，时尚设计单元包括将输入的立体模型和输入参数进行运算的人工智能模型，人工智能模型给出时尚评价，将元素分解，然后融合重组，融合重组后生成的样本分成两个部分，一部分送入人工标记，对人工标记模型根据输入参数进行参数适配，形成样本集，将样本集送入到深度学习算法进行训练，对人工智能模型进行迭代，另一部分直接送入人工智能模型进行时尚评价。In a preferred solution, the fashion design unit includes an artificial intelligence model that operates on the input three-dimensional model and the input parameters, the artificial intelligence model gives fashion evaluation, decomposes the elements, then fuses and reorganizes, and the generated sample after fusion and reorganization is divided into two parts , part of it is sent to manual labeling, the parameters of the artificial labeling model are adapted according to the input parameters to form a sample set, the sample set is sent to the deep learning algorithm for training, the artificial intelligence model is iterated, and the other part is directly sent to the artificial intelligence model Do fashion reviews.
优选的方案中，元素分解包括元素拆解和平面映射，将拆解后的元素与映射得到的平面图形，经过形状、图案、颜色、位置、缩放、翻转、扭曲、阵列中一种或多种操作在组合器进行组合处理后，进行立体重构，实现设计元素的融合重组。In a preferred solution, element decomposition includes element disassembly and plane mapping. The disassembled elements and the plane graphics obtained by mapping are processed by one or more of shape, pattern, color, position, scaling, flipping, twisting, and array. After the operation is performed by the combiner, three-dimensional reconstruction is performed to realize the fusion and reorganization of design elements.
优选的方案中，所述的人工智能模型包括分类器、语义识别器、适配器和聚类器；In a preferred solution, the artificial intelligence model includes a classifier, a semantic recognizer, an adapter and a clusterer;
分类器用于将立体模型进行分类，所述的分类与语义识别器识别出的语义关键词互相关联；The classifier is used to classify the three-dimensional model, and the classification is associated with the semantic keywords recognized by the semantic recognizer;
语义识别器用于将输入参数转换为语义关键词，所述的语义关键词与分类器的分类互相关联；The semantic recognizer is used to convert the input parameters into semantic keywords, and the semantic keywords are correlated with the classification of the classifier;
适配器用于根据语义关键词对立体模型的样本进行适配，适配出符合语义关键词的立体模型；The adapter is used to adapt the samples of the three-dimensional model according to the semantic keywords, and adapt the three-dimensional model conforming to the semantic keywords;
聚类器用于对适配出的立体模型根据时尚归类进行聚合，用于生成时尚评价。The clusterer is used to aggregate the fitted three-dimensional models according to fashion classification, and is used to generate fashion evaluations.
优选的方案中，所述的时尚采集单元设有立体模型数据库。In a preferred solution, the fashion collection unit is provided with a three-dimensional model database.
一种采用上述的时尚设计系统的设计方法，包括以下步骤：A design method using the above-mentioned fashion design system, comprising the following steps:
S1、输入限定参数，人工智能模型读取立体模型数据；S1. Input the limited parameters, and the artificial intelligence model reads the three-dimensional model data;
所述的人工智能模型经过人工标记、参数适配和模型训练后获得；The artificial intelligence model is obtained after manual marking, parameter adaptation and model training;
S2、根据限定参数对立体模型进行时尚评价；S2. Perform fashion evaluation on the three-dimensional model according to the limited parameters;
S3、具有较高时尚票评价的设计送入到时尚生成单元；S3. The design with higher fashion ticket evaluation is sent to the fashion generation unit;
通过以上步骤，实现基于人工智能的时尚设计。Through the above steps, the fashion design based on artificial intelligence is realized.
优选的方案中，步骤S2中，还包括元素分解，融合重组的步骤，由此步骤生成新的样本，新的样本其中一部分经过人工标记、参数适配和模型训练后用于对人工智能模型进行迭代，另一部分送入到人工智能模型进行处理；In a preferred solution, step S2 also includes the steps of element decomposition, fusion and reorganization, and new samples are generated in this step, and a part of the new samples is used for artificial intelligence model after manual labeling, parameter adaptation and model training. Iterate, and the other part is sent to the artificial intelligence model for processing;
元素分解，融合重组的步骤包括：The steps of element decomposition, fusion and recombination include:
S21、将设计元素进行元素拆解，包括形状、图案、颜色、位置；S21. Disassemble the design elements, including shapes, patterns, colors, and positions;
将立体模型进行平面映射；Plan the solid model for mapping;
S22、将拆解后的元素经过缩放、翻转、扭曲、陈列中一种或多种的组合或者组合平面结构；S22. The disassembled elements are scaled, flipped, distorted, and displayed in one or more combinations or combined plane structures;
S23、将组合平面结构进行立体重构，实现融合重组。S23, performing three-dimensional reconstruction of the combined plane structure to realize fusion and reorganization.
优选的方案中，所述的人工智能模型包括分类器、语义识别器、适配器和聚类器；In a preferred solution, the artificial intelligence model includes a classifier, a semantic recognizer, an adapter and a clusterer;
分类器用于根据关键词对立体模型进行分类；The classifier is used to classify the solid model according to the keywords;
语义识别器用于将输入参数转换为关键词；Semantic recognizers are used to convert input parameters into keywords;
适配器用于将分类后的立体模型根据关键词进行关联；The adapter is used to associate the classified three-dimensional models according to the keywords;
聚类器用于将关联后的立体模型根据时尚评价进行聚类。The clusterer is used to cluster the associated three-dimensional models according to the fashion evaluation.
本发明提供的一种服装时尚设计系统与方法，与现有技术相比，具有以下的有益效果：Compared with the prior art, a clothing fashion design system and method provided by the present invention has the following beneficial effects:
（1）本发明提供的新型的时尚设计系统与方法通过约束性条件输入，可以根据人工智能算法自动设计服装款式，从而节约大量的时间和人力。在时尚领域，可以全面的按照用户的需求对服饰进行自由定制，大幅减少了设计成本，提高了设计效率。(1) The novel fashion design system and method provided by the present invention can automatically design clothing styles according to artificial intelligence algorithms through the input of constrained conditions, thereby saving a lot of time and manpower. In the field of fashion, clothing can be freely customized according to the needs of users, which greatly reduces the design cost and improves the design efficiency.
（2）本发明采用立体模型数据，更为直观，便于客户选择。而且采用立体模型数据，也更便于后继的修改和版型设计，便于设计和生产出高质量的时尚服装。(2) The present invention adopts three-dimensional model data, which is more intuitive and convenient for customers to choose. Moreover, the use of three-dimensional model data is also more convenient for subsequent modification and layout design, and it is convenient to design and produce high-quality fashion clothing.
（3）本发明提供的这种新型的时尚设计系统与方法，不仅可以应用在线上应用中，在嵌入式设备中同样适用，大幅提高了实用性。(3) The novel fashion design system and method provided by the present invention can be applied not only in online applications, but also in embedded devices, greatly improving the practicability.
附图说明Description of drawings
下面结合附图和实施例对本发明作进一步说明：Below in conjunction with accompanying drawing and embodiment, the present invention will be further described:
图1为本发明的服装时尚设计系统与方法的流程示意图。FIG. 1 is a schematic flow chart of the clothing fashion design system and method of the present invention.
图2为本发明的服装时尚设计系统与方法的优选的流程示意图。FIG. 2 is a schematic flow chart of the preferred method of the clothing fashion design system and method of the present invention.
图3为本发明的增加新的服装时尚设计样本的流程示意图。FIG. 3 is a schematic flowchart of adding a new fashion design sample of clothing according to the present invention.
图4为本发明的人工智能模型的功能模块示意图。FIG. 4 is a schematic diagram of functional modules of the artificial intelligence model of the present invention.
具体实施方式Detailed ways
实施例1：Example 1:
为了使本发明的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用于限定本发明。此外，下面所描述的本发明各个实施方式中所涉及到的技术特征只要彼此之间未构成冲突就可以相互组合。In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.
参照图1所示，一种时尚设计系统，包括时尚采集单元、时尚处理单元、时尚设计单元和时尚生成单元；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;
所述时尚采集单元用于采集各个种类的时尚元素；The fashion collection unit is used for collecting various kinds of fashion elements;
所述时尚处理单元用于使用深度学习算法对时尚元素进行统计与学习；The fashion processing unit is used for statistics and learning of fashion elements using a deep learning algorithm;
所述时尚设计单元用于将代表时尚元素的服装属性进行分离与再归类；The fashion design unit is used to separate and reclassify clothing attributes representing fashion elements;
所述时尚生成单元用于参考语义或环境条件来生成服饰。The fashion generation unit is used to generate apparel with reference to semantic or environmental conditions.
优选的方案如图2中，所述的时尚处理单元包括训练模块，在训练模块内，将采集的时尚元素建立立体模型，建立立体模型能够更加接近真实的设计，尤其是能够避免出现在后期制作过程中，时尚设计较佳，但是版型设计不足的缺陷，优选的，建立的立体模型按照裁剪切片并立体拼接的方式实现，虽然初始的立体模型增加了工作量，但是后继的工作量则变得非常轻松，便于各种不同设计时尚要素的自由组合和最终效果的评估。将立体模型进行人工标记，人工标记的目的是根据参数对立体模型进行定义，以便于后继人工智能模型的训练，例如区分春秋装和夏装、冬装，区分内衣和外套、区分上装和下装，区分配饰等，区分体型适配度，并对不同的搭配给出时尚评价。The preferred solution is as shown in Figure 2. The fashion processing unit includes a training module. In the training module, a three-dimensional model of the collected fashion elements is established. The establishment of the three-dimensional model can be closer to the real design, especially to avoid appearing in post-production. In the process, the fashion design is better, but the layout design is insufficient. Preferably, the established three-dimensional model is realized by cutting slices and three-dimensional splicing. Although the initial three-dimensional model increases the workload, the subsequent workload changes. It is very easy, which is convenient for the free combination of various design fashion elements and the evaluation of the final effect. The three-dimensional model is manually marked. The purpose of manual marking is to define the three-dimensional model according to the parameters, so as to facilitate the training of subsequent artificial intelligence models, such as distinguishing spring and autumn clothes from summer clothes, winter clothes, distinguishing underwear and coats, distinguishing tops and bottoms, distinguishing between Accessories, etc., distinguish the fitness of body shape, and give fashion evaluations to different collocations.
对人工标记模型根据输入参数进行参数适配，形成样本集，将样本集送入到深度学习算法进行训练，获得人工智能模型。在人工智能模型中，通过适配得出不同输入的立体模型的时尚评估值。在人工智能模型中，不仅根据输入参数进行选择，并根据人工标记的样本对立体模型的评价得出时尚评估值。The parameters of the artificial labeling model are adapted according to the input parameters to form a sample set, and the sample set is sent to the deep learning algorithm for training to obtain an artificial intelligence model. In the artificial intelligence model, the fashion evaluation value of different input stereo models is obtained through adaptation. In the artificial intelligence model, the selection is not only based on the input parameters, but also the fashion evaluation value is obtained based on the evaluation of the three-dimensional model by the manually labeled samples.
优选的方案如图2中，时尚设计单元包括将输入的立体模型和输入参数进行运算的人工智能模型，人工智能模型给出时尚评价，并将数据输送至时尚生成单元。通常时尚生成单元，根据时尚评价的排序，生成用于展示的平面或立体图形，并在展示过程中增加衬托氛围的要素，例如人体模特，立体模型之外的配饰，氛围灯光，闪烁光影等。A preferred solution is as shown in Figure 2, 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 fashion evaluation, and sends the data to the fashion generation unit. Usually, the fashion generation unit generates flat or three-dimensional graphics for display according to the order of fashion evaluation, and adds elements to set off the atmosphere during the display process, such as mannequins, accessories other than three-dimensional models, ambient lighting, flickering light and shadow, etc.
优选的方案如图2中，时尚设计单元包括将输入的立体模型和输入参数进行运算的人工智能模型，人工智能模型给出时尚评价，将元素分解，然后融合重组，融合重组后生成的样本分成两个部分，一部分送入人工标记，对人工标记模型根据输入参数进行参数适配，形成样本集，将样本集送入到深度学习算法进行训练，对人工智能模型进行迭代；送入人工标记有助于通过人工干涉对人工智能模型进行迭代，以提高评价的准确性。另一部分直接送入人工智能模型进行时尚评价。由此方案，能够获得更多的时尚设计样本，或者获得更多的立体模型数据以在后继输入相应参数后获得更多的可选择结果。The preferred solution is shown in Figure 2. 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 fashion evaluation, decomposes the elements, and then fuses and reorganizes them. The samples generated after the fusion and reorganization are divided into Two parts, one part is sent to manual labeling, the parameters of the artificial labeling model are adapted according to the input parameters to form a sample set, the sample set is sent to the deep learning algorithm for training, and the artificial intelligence model is iterated; Helps to iterate the artificial intelligence model through manual intervention to improve the accuracy of the evaluation. The other part is fed directly into the AI model for fashion evaluation. With this solution, 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.
优选的方案如图3中，元素分解包括元素拆解和平面映射，将拆解后的元素与映射得到的平面图形，由此方案，能够大幅减少运算量，经过形状、图案、颜色、位置、缩放、翻转、扭曲、阵列中一种或多种操作在组合器进行组合处理后，进行立体重构，实现设计元素的融合重组。立体重构是指根据元素分解时的坐标，将组合后的图形重新映射到立体模型上的过程。优选的方案中，在组合后先进行筛选，设置的筛选器设定多个预设规则，将不符合设计原理的结果删除大部分，以减少后继立体重构的运算量，以及人工智能模型运算步骤的运算量。例如删除模糊的图像，明显错位的图像，存在破碎图形的图像。The preferred solution is shown in Figure 3. Element decomposition includes element disassembly and plane mapping, and the disassembled elements are mapped to the plane graphics obtained by mapping. This scheme can greatly reduce the amount of calculation. After the shape, pattern, color, position, After one or more operations of scaling, flipping, twisting, and array are combined and processed by the combiner, three-dimensional reconstruction is performed to realize the fusion and reorganization of design elements. Stereoscopic reconstruction refers to the process of remapping the combined graphics to the three-dimensional model according to the coordinates when the elements are decomposed. In the preferred solution, screening is performed first after the combination, and the set filter sets multiple preset rules to delete most of the results that do not conform to the design principles, so as to reduce the amount of calculation for subsequent stereo reconstruction and the calculation of artificial intelligence models. The amount of computation for the step. For example, remove blurred images, images that are clearly misaligned, images with broken graphics.
优选的方案如图4中，所述的人工智能模型包括分类器、语义识别器、适配器和聚类器；由此方案，能够减少各个模块的运算量，减少各个模块的复杂程度，从而提高处理效率。The preferred solution is as shown in Figure 4, the artificial intelligence model includes a classifier, a semantic recognizer, an adapter and a clusterer; this solution can reduce the computational load of each module and reduce the complexity of each module, thereby improving processing efficiency.
分类器用于将立体模型进行分类，所述的分类与语义识别器识别出的语义关键词互相关联；分类器能够减少后继的运算量。例如某立体模型，属于夏装，上装，颜色为白色，面料为丝绸，无图案。进行分类后，便于根据关键词的需求进行适配。分类器采用基于CNN或者fast-CNN深度神经网络。The classifier is used for classifying the three-dimensional model, and the classification is correlated with the semantic keywords recognized by the semantic recognizer; the classifier can reduce the amount of subsequent computation. For example, a three-dimensional model belongs to summer clothes, tops, the color is white, the fabric is silk, and there is no pattern. After classification, it is convenient to adapt according to the needs of keywords. The classifier uses a deep neural network based on CNN or fast-CNN.
语义识别器用于将输入参数转换为语义关键词，所述的语义关键词与分类器的分类互相关联；即语义关键词是分类器中进行分类的依据。语义识别器采用RNNs，即循环神经网络。The semantic recognizer is used to convert 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. Semantic recognizers use RNNs, or Recurrent Neural Networks.
适配器用于根据语义关键词对立体模型的样本进行适配，适配出符合语义关键词的立体模型；适配器可以看作是一个基于关键词的筛选器，具有关键词命中的立体模型数据能够通过适配器，而未命中的立体模型数据则被适配器去除。The adapter is used to adapt the samples of the stereo model according to the semantic keywords, and adapt the stereo model that conforms to the semantic keywords; the adapter can be regarded as a keyword-based filter, and the stereo model data with keyword hits can be passed through adapter, and the missing stereo model data is removed by the adapter.
聚类器用于对适配出的立体模型根据时尚归类进行聚合，用于生成不同的时尚评价。由此方案获得符合需求的结果。The clusterer is used to aggregate the fitted three-dimensional models according to fashion classification to generate different fashion evaluations. This solution achieves the desired results.
优选的方案中，所述时尚采集单元设有立体模型数据库。时尚采集单元用于采集存在的流行以及经典服饰元素的服饰图像，可以在网站上利用“爬虫”等工具进行图像获取来替代耗时的人工下载，在特殊的场所，如舞会和某些地区，如武汉，上海使用摄像机来采集元素，并使用一个数据服务器进行实时存储。在数据服务器中，被存储有标准的立体人体模型。优选的，也可以存储典型的立体人体模型，例如儿童型、偏瘦型或偏胖型、肥胖型等人体模型，在立体人体模型体表设置可调的片，利用可调的片结构构建服装，并将服饰元素的服饰图像根据采集的位置映射到服装的可调片结构上，对片结构进行参数设置，例如形状、图案、颜色、纹理和光泽的表达。采用该方案的优点在于，完成后的时尚服装设计，能够直接获得可用于裁剪的版型数据，而且该版型数据能够被直接的适用到对应的不同类型的人体模型。大幅降低后继的工作量，也具有极大的商业价值。In a preferred solution, the fashion collection unit is provided with a three-dimensional model database. The fashion collection unit is used to collect clothing images of existing popular and classic clothing elements. It can use tools such as "crawlers" to obtain images on the website to replace time-consuming manual downloads. In special places, such as dance parties and certain areas, For example, Wuhan and Shanghai use cameras to capture elements and use a data server for real-time storage. In the data server, standard three-dimensional mannequins are stored. Preferably, it is also possible to store typical three-dimensional mannequins, such as children's, thin or obese, obese mannequins, etc., and set adjustable sheets on the body surface of the three-dimensional mannequin, and use the adjustable sheet structure to construct clothing , and map the clothing image of the clothing element to the adjustable sheet structure of the clothing according to the collected position, and set the parameters of the sheet structure, such as the expression of shape, pattern, color, texture and gloss. The advantage of adopting this solution is that the finished fashion clothing design can directly obtain the pattern data that can be used for cutting, and the pattern data can be directly applied to corresponding different types of human body models. Significantly reducing the workload of the follow-up also has great commercial value.
实施例2：Example 2:
在实施例1的基础上，如图2~3中，一种采用上述的时尚设计系统的设计方法，包括以下步骤：On the basis of Embodiment 1, as shown in Figures 2 to 3, a design method using the above-mentioned fashion design system, comprising the following steps:
S1、输入限定参数，包括但不限于体型、肤色、发型、季节、适用场景等，人工智能模型读取立体模型数据；S1. Input limited parameters, including but not limited to body shape, skin color, hairstyle, season, applicable scene, etc., and the artificial intelligence model reads the three-dimensional model data;
如图2中，所述的人工智能模型经过人工标记、参数适配和模型训练后获得；通过人工标记，人工智能模型能够获得较为准确的训练样本，以帮助人工智能模型提高准确率，由于时尚标准涉及较为主观的判断，因此人工标记在人工智能模型的训练过程中起到非常重要的作用。As shown in Figure 2, the artificial intelligence model is obtained after manual labeling, parameter adaptation and model training; through manual labeling, the artificial intelligence model can obtain relatively accurate training samples to help the artificial intelligence model improve the accuracy rate. Criteria involve more subjective judgments, so manual labeling plays a very important role in the training process of artificial intelligence models.
优选的方案如图4中，所述的人工智能模型包括分类器、语义识别器、适配器和聚类器；The preferred solution is as shown in Figure 4, the artificial intelligence model includes a classifier, a semantic recognizer, an adapter and a clusterer;
分类器用于根据关键词对立体模型进行分类；分类器的作用是将立体模型以分类的方式进行定义，例如对立体模型加入体型的对应标记，适用的季节，适用的场景，图案类别，颜色类别，纹理类别，光泽类别，是否具有弹性，适用的具体位置，例如围巾的位置，胸饰的位置，臂饰的位置，腰饰的位置等，时尚评价分值，以及该分值与对应的体型、肤色、发型之间的加权关系数据。The classifier is used to classify the three-dimensional model according to the keywords; the function of the classifier is to define the three-dimensional model in a classified way, such as adding the corresponding mark of the body type to the three-dimensional model, applicable season, applicable scene, pattern category, color category , texture category, gloss category, elastic or not, applicable specific positions, such as the position of the scarf, the position of the chest ornament, the position of the arm ornament, the position of the waist ornament, etc., the fashion evaluation score, and the score and the corresponding body shape , skin color, hair style weighted relationship data.
语义识别器用于将输入参数转换为关键词；Semantic recognizers are used to convert input parameters into keywords;
适配器用于将分类后的立体模型根据关键词进行关联；通过适配器将符合输入参数的立体模型选取。The adapter is used to associate the classified solid models according to the keywords; the solid models that conform to the input parameters are selected through the adapter.
聚类器用于将关联后的立体模型根据时尚评价进行聚类。The clusterer is used to cluster the associated three-dimensional models according to the fashion evaluation.
聚类器中，将立体模型的平面映射图像数据进行卷积编码，并提取每张图像的卷积神经网络特征，根据训练数据集来学习算法对于时尚元素的细粒度归类能力。由卷积神经网络组成的解码器进行时尚单元的聚类操作，根据训练数据依据时尚评价的标准来引导时尚服装的生成方式。In the clusterer, the plane mapping image data of the stereo model is convolutionally encoded, and the convolutional neural network features of each image are extracted, and the fine-grained classification ability of the algorithm for fashion elements is learned according to the training data set. The decoder composed of the convolutional neural network performs the clustering operation of fashion units, and guides the generation of fashion clothes according to the criteria of fashion evaluation according to the training data.
S2、根据限定参数对立体模型进行时尚评价；S2. Perform fashion evaluation on the three-dimensional model according to the limited parameters;
优选的方案中，还包括元素分解，融合重组的步骤，由此步骤生成新的样本，新的样本其中一部分经过人工标记、参数适配和模型训练后用于对人工智能模型进行迭代，另一部分送入到人工智能模型进行处理；In the preferred solution, it also includes the steps of element decomposition, fusion and reorganization, and new samples are generated in this step. One part of the new samples is used to iterate the artificial intelligence model after manual labeling, parameter adaptation and model training, and the other part is used to iterate the artificial intelligence model. sent to the artificial intelligence model for processing;
元素分解，融合重组的步骤包括：The steps of element decomposition, fusion and recombination include:
S21、将设计元素进行元素拆解，包括但不限于形状、图案、颜色、图案与服装的相对位置、材质、纹理和光泽；S21. Elemental disassembly of design elements, including but not limited to shape, pattern, color, relative position of pattern and clothing, material, texture and gloss;
将立体模型进行平面映射；是指将立体模型以裁剪平面进行展开，得到平面图像，以平面图像进行后继的运算。Plane mapping of the three-dimensional model refers to expanding the three-dimensional model with the clipping plane to obtain a plane image, and performing subsequent operations with the plane image.
S22、将拆解后的元素经过缩放、翻转、扭曲、镜像、锐化、纹理化、模糊、陈列排列中一种或多种的组合或者组合平面结构；平面结构是指构成由多个平面图像堆叠的多层结构，例如底层为服装的平面图像、上层为图案、再上层为纹理、再上层为饰品，由此方案，使各层的平面图像能够分别映射到立体结构的不同结构上。S22. The disassembled elements are scaled, flipped, distorted, mirrored, sharpened, textured, blurred, and arranged in one or more combinations or combined plane structures; plane structures refer to the composition consisting of multiple plane images The stacked multi-layer structure, for example, the bottom layer is the plane image of the clothing, the top layer is the pattern, the top layer is the texture, and the top layer is the ornament. This scheme enables the plane images of each layer to be mapped to different structures of the three-dimensional structure.
S23、将组合平面结构进行立体重构，实现融合重组。S23, performing three-dimensional reconstruction of the combined plane structure to realize fusion and reorganization.
立体人体模型体表设置可调的片结构，可调是指能够调节片结构的位置和形状，以及片之间的连接结构，利用可调的片结构构建服装，并将具有服饰元素的服饰图像的组合平面结构，根据采集的位置映射到服装的可调片结构上，对片结构进行参数设置，例如形状、图案、颜色、纹理和光泽的表达，实现融合重组操作。完成后的时尚服装设计，能够直接获得可用于裁剪的版型数据。The body surface of the three-dimensional mannequin is provided with an adjustable sheet structure. Adjustable 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 clothing, and the clothing image with clothing elements The combined plane structure is mapped to the adjustable piece structure of the garment according to the collected position, and the parameters of the piece structure are set, such as the expression of shape, pattern, color, texture and gloss, and the fusion and reorganization operation is realized. After the fashion clothing design is completed, the pattern data that can be used for cutting can be directly obtained.
S3、具有较高时尚票评价的设计送入到时尚生成单元；时尚生成单元，根据时尚评价的排序，生成用于展示的平面或立体图形，并在展示过程中增加衬托氛围的要素，例如人体模特，立体模型之外的配饰，氛围灯光，闪烁光影等。S3. The design with high fashion ticket evaluation is sent to the fashion generation unit; the fashion generation unit, according to the order of fashion evaluation, generates plane or three-dimensional graphics for display, and adds elements to set off the atmosphere in the display process, such as human body Models, accessories other than dioramas, ambient lighting, shimmering lights, etc.
通过以上步骤，实现基于人工智能的时尚设计。Through the above steps, the fashion design based on artificial intelligence is realized.
上述的实施例仅为本发明的优选技术方案，而不应视为对于本发明的限制，本申请中的实施例及实施例中的特征在不冲突的情况下，可以相互任意组合。本发明的保护范围应以权利要求记载的技术方案，包括权利要求记载的技术方案中技术特征的等同替换方案为保护范围。即在此范围内的等同替换改进，也在本发明的保护范围之内。The above-mentioned embodiments are only the preferred technical solutions of the present invention, and should not be regarded as limitations of the present invention. The embodiments and features in the present application may be arbitrarily combined with each other without conflict. The protection scope of the present invention shall take the technical solutions described in the claims, including the equivalent alternatives of the technical features in the technical solutions described in the claims, as the protection scope. That is, equivalent replacements and improvements within this scope are also within the protection scope of the present invention.
Claims (10)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111455629.1A CN114357544A (en) | 2021-12-01 | 2021-12-01 | Fashion design system and method for clothes |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111455629.1A CN114357544A (en) | 2021-12-01 | 2021-12-01 | Fashion design system and method for clothes |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114357544A true CN114357544A (en) | 2022-04-15 |
Family
ID=81098317
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111455629.1A Pending CN114357544A (en) | 2021-12-01 | 2021-12-01 | Fashion design system and method for clothes |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114357544A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114782653A (en) * | 2022-06-23 | 2022-07-22 | 杭州彩连科技有限公司 | Method and system for automatically expanding dress design layout |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102810214A (en) * | 2011-06-03 | 2012-12-05 | 段湘华 | Method, device and system for customizing clothes online |
| CN103971407A (en) * | 2014-05-16 | 2014-08-06 | 陆剑青 | Online 3D garment customization interactive design system |
| CN110728015A (en) * | 2018-06-27 | 2020-01-24 | 国际商业机器公司 | Cognitive automation and interactive personalized fashion design |
| CN112434210A (en) * | 2020-12-14 | 2021-03-02 | 武汉纺织大学 | Clothing fashion trend prediction system and method |
| CN112488780A (en) * | 2020-10-22 | 2021-03-12 | 江南大学 | Intelligent design method for personalized customized functional clothing |
| CN113393550A (en) * | 2021-06-15 | 2021-09-14 | 杭州电子科技大学 | Fashion garment design synthesis method guided by postures and textures |
-
2021
- 2021-12-01 CN CN202111455629.1A patent/CN114357544A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102810214A (en) * | 2011-06-03 | 2012-12-05 | 段湘华 | Method, device and system for customizing clothes online |
| CN103971407A (en) * | 2014-05-16 | 2014-08-06 | 陆剑青 | Online 3D garment customization interactive design system |
| CN110728015A (en) * | 2018-06-27 | 2020-01-24 | 国际商业机器公司 | Cognitive automation and interactive personalized fashion design |
| CN112488780A (en) * | 2020-10-22 | 2021-03-12 | 江南大学 | Intelligent design method for personalized customized functional clothing |
| CN112434210A (en) * | 2020-12-14 | 2021-03-02 | 武汉纺织大学 | Clothing fashion trend prediction system and method |
| CN113393550A (en) * | 2021-06-15 | 2021-09-14 | 杭州电子科技大学 | Fashion garment design synthesis method guided by postures and textures |
Non-Patent Citations (1)
| Title |
|---|
| 段然;刘晓刚;: "少数民族服饰元素在数字化服装设计中的运用", 贵州民族研究, no. 10, 25 October 2016 (2016-10-25) * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114782653A (en) * | 2022-06-23 | 2022-07-22 | 杭州彩连科技有限公司 | Method and system for automatically expanding dress design layout |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108520206B (en) | A Recognition Method of Fungal Microscopic Image Based on Fully Convolutional Neural Network | |
| CN107316340B (en) | A fast face modeling method based on a single photo | |
| CN101576913B (en) | Tongue Image Automatic Clustering, Visualization and Retrieval Method Based on Self-Organizing Map Neural Network | |
| CN107609193A (en) | The intelligent automatic processing method and system of picture in a kind of suitable commodity details page | |
| CN109344699A (en) | Disease identification method of winter jujube based on hierarchical deep convolutional neural network | |
| CN106203533A (en) | The degree of depth based on combined training study face verification method | |
| CN109191455A (en) | A kind of field crop pest and disease disasters detection method based on SSD convolutional network | |
| CN106296695A (en) | Adaptive threshold natural target image based on significance segmentation extraction algorithm | |
| CN108416353A (en) | Crop field spike of rice fast partition method based on the full convolutional neural networks of depth | |
| CN108717732A (en) | A kind of expression method for tracing based on MobileNets models | |
| CN108520204A (en) | A face recognition method | |
| CN108109055A (en) | A kind of across scene costume retrieval method based on image rendering | |
| CN109766822A (en) | Gesture recognition method and system based on neural network | |
| CN109086768A (en) | The semantic image dividing method of convolutional neural networks | |
| CN101894263A (en) | Computer Aided Classification System and Method of Plant Species Based on Level Set and Local Sensitive Discriminant Mapping | |
| CN109961022A (en) | A deep learning-based simulation method for clothing sketch input cloth material recognition | |
| CN103745497A (en) | Plant growth modeling method and system | |
| CN109993838A (en) | Implementation method and system of virtual trial launch based on WebGL and face reconstruction | |
| CN111476241A (en) | A method and system for character clothing conversion | |
| CN109446985A (en) | Multi-angle plants identification method based on vector neural network | |
| CN115168745B (en) | Virtual character image recreating method and system based on image technology | |
| CN112434210A (en) | Clothing fashion trend prediction system and method | |
| CN100487732C (en) | Method for generating cartoon portrait based on photo of human face | |
| CN106960457A (en) | A kind of colored paintings creative method extracted and scribbled based on image, semantic | |
| CN114357544A (en) | Fashion design system and method for clothes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |