Data generation methods, apparatus, electronic devices and computer storage media

Abstract

This invention provides a data generation method, apparatus, electronic device, and computer storage medium. The data generation method includes: acquiring constituent elements of a target object; determining element expansion rules corresponding to the constituent elements of the target object; expanding the elements in the constituent elements of the target object according to the element expansion rules; and generating result data based on the expanded constituent elements of the target object. This invention satisfies diverse design needs, reduces design requirements and costs for designers, and improves design efficiency.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a data generation method, apparatus, electronic device and computer storage medium. Background Technology

[0002] With the rapid development of AI (Artificial Intelligence) technology, many traditional industries, such as traditional manufacturing and traditional art creation and design, are gradually transforming towards greater efficiency and intelligence. Under this trend, the demand for AI empowerment in traditional industries, such as the need for apparel design—a core component of the clothing industry—to achieve efficient, diverse, and personalized customized designs, has also emerged.

[0003] Taking fashion design as an example, pattern, material, and pattern / texture are the three major elements of fashion design, representing the three fundamental components of clothing and the three basic ways in which a designer's design concept is expressed. Traditionally, designers typically demonstrate these design concepts through sketches or physical samples. However, this traditional method often results in a lengthy cycle from conceptual design to final design, which is highly unfavorable for the rapid and low-cost instantiation of design concepts. Therefore, fashion design software such as CLO3D has gradually been applied to the field. Currently, these software programs provide resources such as textures and materials, which designers can use to design according to their needs. However, essentially, these software programs only digitize the resources; manual design by designers is still required.

[0004] Therefore, on the one hand, these clothing design software programs have relatively basic functions and cannot automatically complete clothing design, still requiring a high level of skill from designers; on the other hand, the electronic materials are also relatively monotonous and cannot meet diverse design needs.

[0005] Similarly, the same problem exists in other traditional industries, especially those related to design. Summary of the Invention

[0006] In view of this, embodiments of the present invention provide a data generation scheme to at least partially solve the above-mentioned problems.

[0007] According to a first aspect of the present invention, a data generation method is provided, comprising: acquiring constituent elements of a target object; determining element expansion rules corresponding to the constituent elements of the target object; expanding the elements in the constituent elements of the target object according to the element expansion rules; and generating result data based on the expanded constituent elements of the target object.

[0008] According to a second aspect of the present invention, a data generation apparatus is provided, comprising: an acquisition module for acquiring constituent elements of a target object; an expansion module for determining an element expansion rule corresponding to the constituent elements of the target object; and expanding elements in the constituent elements of the target object according to the element expansion rule; and a generation module for generating result data based on the expanded constituent elements of the target object.

[0009] According to a third aspect of the present invention, an electronic device is provided, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus; the memory is used to store at least one executable instruction, wherein the executable instruction causes the processor to perform an operation corresponding to the data generation method described in the first aspect.

[0010] According to a fourth aspect of the present invention, a computer storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the data generation method as described in the first aspect.

[0011] According to the data generation scheme provided in the embodiments of the present invention, in the process of generating result data based on a target object, the constituent elements of the target object are first obtained; further, the elements in the obtained constituent elements of the target object are expanded according to the corresponding element expansion rules. Thus, each element in various constituent elements of the target object can be expanded from a single element into multiple different elements, realizing the expansion and enrichment of the constituent elements of the target object; subsequently, a variety of result data can be generated based on the expanded constituent elements of the target object. Therefore, on the one hand, the element materials of the target object are greatly enriched, meeting diverse design needs; on the other hand, the richness of the elements makes the results obtained through material combinations more abundant and diverse, reducing the design requirements and costs for designers, and improving design efficiency. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0013] Figure 1 This is a flowchart of the steps of a data generation method according to Embodiment 1 of the present invention;

[0014] Figure 2A This is a flowchart of the steps of a data generation method according to Embodiment 2 of the present invention;

[0015] Figure 2B for Figure 2A A schematic diagram of a scenario example in the illustrated embodiment;

[0016] Figure 3A This is a flowchart of the steps of a data generation method according to Embodiment 3 of the present invention;

[0017] Figure 3B for Figure 3A A schematic diagram of a clothing knowledge graph in the illustrated embodiment;

[0018] Figure 4A This is a flowchart of the steps of a data generation method according to Embodiment 4 of the present invention;

[0019] Figure 4B for Figure 4A An extended schematic diagram of a pattern texture element in the illustrated embodiment;

[0020] Figure 4C for Figure 4A An extended schematic diagram of a material element in the illustrated embodiment;

[0021] Figure 4D for Figure 4A A schematic diagram illustrating the generation of a three-dimensional garment in the illustrated embodiment;

[0022] Figure 5 This is a structural block diagram of a data generation apparatus according to Embodiment 5 of the present invention;

[0023] Figure 6 This is a schematic diagram of the structure of an electronic device according to Embodiment Six of the present invention. Detailed Implementation

[0024] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art should fall within the protection scope of the present invention.

[0025] The specific implementation of the embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0026] Example 1

[0027] Reference Figure 1 The diagram shows a flowchart of a data generation method according to Embodiment 1 of the present invention.

[0028] The data generation method in this embodiment includes the following steps:

[0029] Step S100: Obtain the constituent elements of the target object.

[0030] The solutions provided in this invention can be applied to various target object application scenarios with constituent elements, including but not limited to: 3D clothing generation scenarios with clothing constituent elements, 3D building generation scenarios with architectural constituent elements, 3D decoration generation scenarios with decoration constituent elements, 3D scene generation scenarios with scene constituent elements, and so on.

[0031] The clothing elements are used to indicate the basic elements that can form 3D clothing, such as patterns, textures, materials, and patterns. The architectural elements are used to indicate the basic elements that can form 3D buildings, such as house shapes, room layouts, door and window arrangements, etc. The interior decoration elements are used to indicate the elements that can form the interior decorations of 3D buildings, such as furniture layouts, floor styles, wall styles, etc. The set design elements are used to form 3D sets within a given environment, such as background props and facility props. For example, a 3D set in a beach environment might require ocean background props, beach leisure facility props, entertainment facility props, etc., and the resulting 3D set will be applied to reality for video or image shooting.

[0032] However, this is not the only application. In practical applications, any scenario that can be designed according to its constituent elements can be applied to the solutions of this invention.

[0033] Step S200: Determine the element expansion rules corresponding to the constituent elements of the target object.

[0034] Element expansion rules are used to expand the elements that constitute the target object, in order to enrich and expand the elements.

[0035] Different target objects may have different element expansion rules. For example, when expanding the pattern and texture of clothing elements, the color and / or style can be expanded based on the existing pattern and texture to form a new pattern and texture; when expanding the material, the rendering attributes can be expanded based on the existing material or pattern and texture to form a new material; when expanding the pattern, the pattern can be transformed, disassembled or combined based on the existing pattern to form a new pattern, and so on.

[0036] When expanding the shape of a building, architectural elements can be based on an existing building shape, with changes made to the shape of one or more parts to create a new building shape; when expanding the room layout, existing room layouts can be adjusted, added to, or deleted to create new room layouts; when expanding the door and window layout, existing door and window layouts can be based on adjustments to their positions or shapes to create new door and window layouts, and so on.

[0037] When expanding the furniture layout, the decoration requirements can be based on the existing furniture layout, and one or more pieces of furniture can be expanded by changing their position or type to form a new furniture layout; when expanding the flooring style, the existing flooring style can be based on the existing flooring style, and the flooring color, material, shape, etc. can be expanded to form a new flooring style; the expansion of wall style is similar to that of flooring style.

[0038] When expanding the elements of a set design, background props can be expanded based on existing ones, including their position, style, and content, to create new background props. Similarly, when expanding the elements of facilities and props, existing facilities and props can be expanded based on their shape, pattern, and accessories, to create new facilities and props, and so on.

[0039] It should be noted that the above is only an illustrative example. In practical applications, those skilled in the art can implement a scheme to generate corresponding result data based on the data generation principle provided in the embodiments of the present invention, according to the extended results of the constituent elements of the target object.

[0040] Step S300: Based on the element expansion rules, expand the elements in the constituent elements of the target object.

[0041] As mentioned earlier, by utilizing the element expansion rules corresponding to the constituent elements of the target object, the elements in the constituent elements of the target object can be effectively expanded to enrich their quantity, content, and form.

[0042] Step S400: Generate result data based on the expanded target object constituent elements.

[0043] By appropriately combining and matching the elements in the expanded target object's constituent elements, corresponding results, i.e., result data, can be generated, such as 3D clothing, 3D architecture, 3D decoration, 3D scenery, etc.

[0044] Furthermore, in one feasible approach, after obtaining the result data, the result data can be sent to industrial control equipment for display. The industrial control equipment can then control the production of the items corresponding to the result data, thereby improving the conversion and utilization efficiency of the result data. The industrial control equipment can be any suitable device applied in the industrial field that has display and other equipment control functions; the specific implementation of the industrial control equipment is not limited in this embodiment of the invention.

[0045] For example, the generated 3D clothing data can be sent to industrial control equipment. The industrial control equipment can display the 3D clothing data and control related equipment based on this data, such as fabric rendering equipment to render the corresponding patterns and textures, and controlling fabric cutting equipment to cut the corresponding patterns, etc.

[0046] For example, the generated 3D building data can be sent to industrial control equipment. The industrial control equipment can display the 3D building data and control related equipment based on this data, such as model making equipment to create a 3D building model, and so on.

[0047] For example, the generated 3D decoration data can be sent to industrial control equipment. The industrial control equipment can display the 3D decoration data and control related equipment based on this data, such as a 3D printer printing the effect model of the 3D decoration, and so on.

[0048] For example, the generated 3D scene data can be sent to industrial control equipment. The industrial control equipment can display the 3D scene data and control related equipment based on this data, such as a 3D printer printing the effect model of the 3D scene, and so on.

[0049] As can be seen, through this embodiment, in the process of generating result data based on the target object, the constituent elements of the target object are first obtained. Further, the elements in the obtained constituent elements of the target object are expanded according to the corresponding element expansion rules. Thus, each element in various constituent elements of the target object can be expanded from a single element into multiple different elements, realizing the expansion and enrichment of the constituent elements of the target object. Consequently, based on the expanded constituent elements of the target object, a wide variety of result data can be generated. Therefore, on the one hand, the element materials of the target object are greatly enriched, meeting diverse design needs; on the other hand, the richness of the elements makes the results generated by combining the elements more abundant and diverse, reducing the design requirements and costs for designers, and improving design efficiency.

[0050] The data generation method of this embodiment can be executed by any suitable electronic device with data processing capabilities, including but not limited to: servers, mobile terminals (such as mobile phones, PADs, etc.) and PCs.

[0051] The above describes the forms and extensions of various target object constituent elements, as well as the process of generating result data based on the extension results, using multiple application scenarios as examples. In the following embodiments, a three-dimensional clothing generation scenario is used as an example to illustrate the above process. When the data generation scheme of this invention is applied to a three-dimensional clothing generation scenario, the target object constituent elements are implemented as clothing constituent elements, which include at least one type.

[0052] Example 2

[0053] Reference Figure 2A The diagram shows a flowchart of a data generation method according to Embodiment 2 of the present invention.

[0054] The data generation method in this embodiment includes the following steps:

[0055] Step S102: Based on the sample clothing image data, obtain at least one type of clothing composition element.

[0056] When the data generation scheme provided in the embodiments of the present invention is applied to a three-dimensional clothing generation scene, the acquisition of the constituent elements of the target object can be implemented as the above step S102, that is, based on the sample clothing image data, at least one type of clothing constituent element is acquired.

[0057] Clothing composition elements are the basic elements used to construct clothing; they are the components of clothing. Clothing composition elements typically include several categories, including but not limited to: pattern (the overall cut and shape of the clothing); material (the material and texture of the fabric in three-dimensional clothing, describing the details of the fabric surface such as color, texture, embossing, and light); and pattern / texture (important visual attributes of three-dimensional clothing, describing information such as color, surface pattern, and design). A garment can be designed using these three categories of clothing composition elements. However, it is not limited to these three categories. In practical applications, more optional clothing composition elements can be added, such as patterns, styles, and combinations. Those skilled in the art can set these appropriately according to actual needs.

[0058] The constituent elements of clothing can be obtained from sample clothing image data, for example, through analysis by a network model (such as the DeepFashion model). This embodiment does not limit the specific means of obtaining the constituent elements of clothing.

[0059] Generally, newly designed clothing is intended to become a future trend. Therefore, the sample clothing image data can be selected from images released more recently to analyze current trends. However, it is not limited to this. In practical applications, it can also be image data from a more distant time, or a mixture of recent and distant image data. The clothing designer can choose according to the actual needs.

[0060] Step S104: Determine the element expansion rules corresponding to at least one type of clothing constituent elements, and expand the elements in at least one type of clothing constituent elements according to the element expansion rules.

[0061] When the data generation scheme provided in this embodiment of the invention is applied to a three-dimensional clothing generation scene, the step of determining the element expansion rule corresponding to the constituent elements of the target object and expanding the elements in the constituent elements of the target object according to the element expansion rule can be implemented as step S104 above, that is, determining the element expansion rule corresponding to at least one type of clothing constituent elements and expanding the elements in at least one type of clothing constituent elements according to the element expansion rule.

[0062] The element expansion rule is used to expand the elements in the clothing composition elements to form multiple new elements based on a single element. If a category of clothing composition elements is considered as a set, then each specific element within it is considered an element. For example, if there are three patterns / textures in the pattern / texture category, then these three patterns / textures are the three elements within the pattern / texture category. In practical applications, these elements can be implemented in the form of images.

[0063] Each type of clothing element can have corresponding element expansion rules. However, this is not the only possibility; in practical applications, only some types of clothing elements may have element expansion rules. The element expansion rules for each type of clothing element need to be appropriate for the element itself. For example, the element expansion rules for pattern and texture elements may indicate the expansion of color and / or style for the elements; the element expansion rules for material elements may indicate the expansion of material rendering attributes for the elements, and so on.

[0064] Once the rules for expanding the elements corresponding to the constituent elements of clothing are determined, the elements within these constituent elements can be expanded to obtain the corresponding expanded results. Through element expansion, the elements within the constituent elements of clothing are greatly enriched, providing ample material for the design and creation of clothing works.

[0065] Step S106: Generate a three-dimensional garment based on the expanded garment components.

[0066] When the data generation scheme provided in the embodiments of the present invention is applied to a three-dimensional clothing generation scenario, the step of generating result data based on the extended target object constituent elements can be implemented as the above step S106, that is: generating three-dimensional clothing based on the extended clothing constituent elements.

[0067] As mentioned earlier, after expanding the elements that constitute clothing, sufficient materials can be obtained. Based on this, clothing designers can use these materials to freely match and combine them to generate three-dimensional clothing. Three-dimensional clothing is a virtual garment created through 3D modeling technology, typically represented as a triangular mesh and visualized through rendering algorithms. In this embodiment, the specific implementation method for generating three-dimensional clothing based on clothing constituent elements is not limited.

[0068] The following example illustrates the above process. Figure 2B As shown.

[0069] This example uses a web crawler to collect recent (e.g., within the past month) sample clothing images from websites such as fashion shows, fashion magazines, and fashion blogs. The collected sample clothing images are then analyzed using appropriate methods (such as the DeepFashion model) to obtain the clothing's constituent elements. This example sets up the acquisition of three categories of clothing constituent elements: pattern elements, material elements, and pattern / texture elements. Furthermore, this example only provides a simple illustration of expanding the pattern / texture element.

[0070] like Figure 2B As shown in the diagram, assuming the element expansion rules corresponding to the floral texture category elements indicate color expansion and style expansion for the elements within the floral texture category elements (hereinafter referred to as floral texture elements), taking a single floral texture element A as an example, assuming the original color of floral texture element A is blue (shown as a vertical line in the diagram) and the original style is elegant, then on the one hand, the floral texture element A can be expanded in color, which in this example is simply expanded to yellow (shown as a diagonal line in the diagram) and red (shown as a horizontal line in the diagram); on the other hand, the floral texture element A can be expanded in style, which in this example is simply expanded to cartoon style and avant-garde style. The specific settings and names of the various styles follow the regulations of the clothing design industry, and will not be detailed in this embodiment of the invention.

[0071] Assuming the generated 3D garment is a sleeveless dress (pattern) made of cotton, with components including an upper body and a lower body, the possible textures for the upper body are: [Elegant Style, Blue], [Elegant Style, Yellow], [Elegant Style, Red]; [Cartoon Style, Blue], [Cartoon Style, Yellow], [Cartoon Style, Red]; [Avant-garde Style, Blue], [Avant-garde Style, Yellow], [Avant-garde Style, Red]. Similarly, the lower body can also use one of the above textures. The resulting 3D garment, combining the upper and lower body, can use any combination of the nine textures for the upper and lower body (the diagram only shows combinations of the nine textures for the upper and lower body). In this example, the 3D garment uses a uniform material, but those skilled in the art should understand that expanding the elements of the material-based garment will further enrich the possible forms of the generated 3D garment.

[0072] After combining the patterns and textures of the upper and lower body parts, the clothing designer can select one or more final three-dimensional garments according to their needs. Figure 2B The image shows the selection of a 3D outfit, specifically a cotton outfit with an elegant yellow top and a cartoon blue bottom.

[0073] According to this embodiment, intelligent clothing design is achieved through the automatic generation of 3D clothing. In this generation process, at least one type of clothing composition element is first obtained based on sample clothing image data, such as pattern, material, and pattern / texture. Further, the elements in the obtained clothing composition elements are expanded according to corresponding element expansion rules. Thus, each element in various clothing composition elements can be expanded from a single element into multiple different elements, achieving the expansion and enrichment of clothing composition elements. Subsequently, a wide variety of 3D clothing can be generated based on the expanded clothing composition elements. Therefore, on the one hand, this greatly enriches the materials for clothing design, meeting diverse design needs; on the other hand, the abundance of materials makes the types and styles of clothing created through material combinations more diverse and varied, reducing the design requirements and costs for designers, and improving design efficiency.

[0074] The data generation method of this embodiment can be executed by any suitable electronic device with data processing capabilities, including but not limited to: servers, mobile terminals (such as mobile phones, PADs, etc.) and PCs.

[0075] Example 3

[0076] Reference Figure 3A The diagram shows a flowchart of a data generation method according to Embodiment 3 of the present invention.

[0077] The data generation method in this embodiment includes the following steps:

[0078] Step S202: Based on the sample clothing image data, obtain at least one type of clothing composition element.

[0079] In one feasible approach, at least one class of clothing constituent elements can be obtained by generating a clothing knowledge graph based on sample clothing image data. A knowledge graph is a semantic network that reveals the relationships between entities, providing a formal description of things in the real world and their interrelationships. For a clothing knowledge graph, clothing constituent elements and their components can be used as entities, with the relationships between elements, between elements themselves, and between elements as edges, thus constructing a graph-structured clothing knowledge graph. A simple example of a constructed clothing knowledge graph is shown below. Figure 3B As shown, by Figure 3B As can be seen, the elements constituting the clothing include: pattern elements, material elements, pattern and texture elements, style elements, and color elements. Each element category has certain sub-elements, as shown in Figure 3B, which illustrates the sub-elements of each element. Under each sub-element, there can be further sub-elements, such as "shirt" under the "top" element. Figure 3B The clothing knowledge graph shown can easily obtain the constituent elements of clothing, the elements at each level under each element, and the relationships between them, effectively improving the speed and efficiency of obtaining the constituent elements of clothing and their corresponding elements.

[0080] Optionally, when using a clothing knowledge graph, before obtaining at least one type of clothing constituent element from the clothing knowledge graph generated based on sample clothing image data, the method further includes: obtaining sample clothing image data and corresponding textual description data; obtaining information about clothing elements and element labels corresponding to the clothing elements through a neural network model for clothing element detection; obtaining the association relationships between the clothing elements based on the element labels and the textual description data; and generating a clothing knowledge graph based on the association relationships and the correspondence between the clothing elements and the clothing constituent elements. The textual description data can provide a direct semantic explanation of the clothing image data. Combining the textual description data with the clothing image data allows for a more accurate understanding of clothing constituent elements and fashion trends. The neural network model can be any suitable network model, including but not limited to CNN network models (convolutional neural network models) with clothing element detection capabilities, and the DeepFashion model for clothing image data processing. The DeepFashion model is an analysis model built based on the DeepFashion2 dataset. By pre-training it, various functions and data based on clothing image data can be obtained, including but not limited to: clothing subject detection, clothing key point estimation, clothing segmentation, clothing retrieval, clothing element detection, and clothing element label generation.

[0081] For example, a certain number of sample clothing images and their corresponding text descriptions are obtained. The sample clothing images are then input into the DeepFashion model to obtain the required clothing elements (including but not limited to one or more of the following: pattern elements, material elements, and pattern / texture elements) and their corresponding element tags. High-frequency popular element words are then obtained from the text descriptions. Using the element tags as the basic corpus, they are matched with these high-frequency popular element words to obtain a certain number of high-frequency popular element words corresponding to each element tag, such as 10. Since the text descriptions are usually related to the clothing elements in the clothing images, the element tags can establish a relationship between a clothing element and other clothing elements. For example, if the text description corresponding to a clothing image is "Cartoon style is all the rage in short skirts this season," the information contained in the clothing elements includes "short skirt" (pattern) and "cartoon style" (pattern / texture). This text description can then establish a relationship between "short skirt" and "cartoon style." Furthermore, since there is a corresponding relationship between clothing elements and clothing components, as mentioned above, "short skirt" corresponds to pattern elements, and "cartoon style" corresponds to pattern / texture elements. Based on this, a clothing knowledge graph can be established, which includes information on the constituent elements of clothing, information on the elements within those constituent elements, information on the relationships between the constituent elements, and information on the relationships between the elements within those constituent elements. Through this clothing knowledge graph, all of the above information can be quickly obtained, thereby greatly improving the efficiency of subsequent clothing data processing.

[0082] Step S204: Determine the element expansion rules corresponding to at least one type of clothing constituent elements, and expand the elements in at least one type of clothing constituent elements according to the element expansion rules.

[0083] As mentioned above, the constituent elements of clothing include at least one of the following: pattern elements, material elements, and pattern / texture elements. Some or all of these constituent elements of clothing have their own element expansion rules. In this embodiment of the invention, the element expansion of material elements and pattern / texture elements is mainly described. However, those skilled in the art should understand that for pattern elements and other types of elements, corresponding element expansion rules can also be set according to actual needs to carry out element expansion.

[0084] In one feasible approach, when the garment's constituent elements include patterned or textured elements, step S204 can be implemented as follows: determining the color expansion rules and / or style expansion rules corresponding to the patterned or textured elements; expanding the colors of the elements within the patterned or textured elements according to the color expansion rules; and / or expanding the styles of the elements within the patterned or textured elements according to the style expansion rules. Color expansion expands the color range of the elements within the patterned or textured elements and their compatibility with other colors; style expansion gives the elements within the patterned or textured elements more varied styles to meet the design needs of different garment designers.

[0085] Optionally, color expansion of elements in a pattern / texture category, according to color expansion rules, can include: expanding the colors of elements in the pattern / texture category by rotating a hue ring at a set angle, based on the colors corresponding to the elements. The specific number of rings on the hue ring and the set angle can be appropriately set by those skilled in the art according to actual needs, ensuring effective differentiation between the expanded colors and between the expanded colors and the original colors of the elements. A hue ring is a circular arrangement of hues in a spectrum, with colors arranged according to the order of their appearance in nature. Different numbers of rings on the hue ring result in different hue spacing. For example, the spacing between each hue on a twelve-hue ring is 30 degrees, and the spacing between each hue on a twenty-four-hue ring is 15 degrees. Therefore, optionally, the set angle in this embodiment is 60 degrees to make the difference between the colors before and after expansion more obvious. Color expansion using a hue ring is simple to implement and reduces the cost of color expansion.

[0086] Alternatively, the above-mentioned method of adjusting colors via a color wheel can be specifically implemented as follows: The color space of elements in the pattern / texture category is converted from RGB space to HSV space; in HSV space, the color of the element is expanded by rotating the color wheel at a set angle according to the color corresponding to the element. Converting the color space of an element from RGB space to HSV space can change the color of the element without altering the light perception, achieving color expansion and resulting in a better color expansion effect.

[0087] In a feasible method for style extension of floral and texture elements, the style extension of elements within these elements according to style extension rules can be achieved by: obtaining information about the design style to be used; and using a multi-stroke attention perception algorithm to transfer the design style indicated by the information about the design style to be used to the elements of the floral and texture elements, thereby extending the style of the elements within the floral and texture elements. The multi-stroke attention perception algorithm (Attention-aware Multi-stroke Style Transfer) uses an attention mechanism and a multi-stroke fusion strategy. First, it uses a self-attention mechanism to construct a reconstruction autoencoder with a previously unknown style, from which an attention map of the original image can be obtained. Then, by performing multi-scale style transformations on content and style features, feature maps representing multiple stroke patterns are generated. Finally, combining the attention map, a flexible mixing strategy is used to blend important feature points in the attention map with feature maps of multiple stroke patterns, outputting images with different stroke patterns. Given a pair of images containing patterned textures and images of a predetermined style, the patterned texture images are fused using the multi-stroke attention perception algorithm described above. This allows for the fusion of multi-stroke style transformation based on an attention mechanism, resulting in patterned texture images of different styles. Furthermore, it ensures the consistency of attention in the patterned texture images before and after the style transformation (i.e., the spatial consistency of visual attention distribution between the patterned texture images and the images of the predetermined style).

[0088] For example, if the design style to be used is Van Gogh style or Picasso style, it can be transferred to the elements of floral and texture elements through a multi-stroke attention perception algorithm to form Van Gogh style elements or Picasso style elements, thereby realizing the style expansion of the elements.

[0089] While the above method achieves style expansion of elements within the floral and texture category, to further enrich the styles of floral and texture elements, another feasible approach is to implement style expansion of elements within the floral and texture category according to style expansion rules. This can be achieved by: obtaining style data of the elements within the floral and texture category; using the Mondrian Random Process algorithm to concatenate the style data; and obtaining the style expansion of the elements based on the concatenation result. This creates new elements within the floral and texture category, further enriching the variety and quantity of elements. The Mondrian Random Process algorithm can be represented as a recursive generation process algorithm that randomly performs axis-aligned spatial cutting to divide a space into multiple spaces.

[0090] Specifically, in this embodiment, optionally, the step of splicing the style data using the Mondrian random process algorithm and obtaining the style extension of the element based on the splicing result includes: dividing a preset two-dimensional image space into multiple subspaces using the Mondrian random process algorithm; selecting a portion of the subspaces to fill with the style data; filling the remaining spaces in the multiple subspaces, excluding the selected portion, with the dominant color of the two-dimensional image space; splicing the style data and the dominant color based on the filling result; and obtaining the style extension of the element in the pattern texture category based on the splicing result.

[0091] Through the above process, the color, style, and pattern of floral textures have been expanded in many ways.

[0092] In addition, when the elements constituting the clothing include material-type elements, step S204 can be implemented as follows: determining the material rendering extension rules corresponding to the material-type elements; and extending the material rendering attributes of the elements in the material-type elements according to the material rendering extension rules. Material-type elements have physical properties (such as cotton, linen, leather, etc.) and rendering attributes. Physical properties are inherent properties of materials, and in the three-dimensional presentation of clothing, different material effects can be achieved through the rendering attributes of the materials. In this embodiment, by extending the material rendering attributes of the elements in the material-type elements, more diverse and richer material effects can be achieved.

[0093] Optionally, extending the material rendering attributes of elements in the material category elements according to the material rendering extension rules may include: obtaining the main color tone of the elements in the pattern texture category elements after extension; and extending the material rendering attributes of the elements according to the main color tone and the UV texture space transformation of the material sphere. In 3D design, to create realistic materials, it is necessary to understand material properties. Material refers to the simulated physical properties of objects in the virtual world, such as color, reflection, transparency, texture, etc.; while material sphere is a general term for the integrated properties of this material, commonly represented by a sphere in the industry, hence the name material sphere. Based on this, in this embodiment of the invention, the extension of material category elements can be achieved through material sphere. Specifically, extending the material rendering attributes of the elements according to the main color tone and the UV texture space transformation of the material sphere includes: performing a transformation operation in the UV texture space of the material sphere; and extending the material rendering attributes of the material sphere on the pattern texture and / or main color tone according to the transformation operation; wherein the transformation operation includes at least one of the following: rotation operation, translation operation, and scaling operation.

[0094] Elements within material-type elements will be rendered as solid colors or as the textures corresponding to elements within patterned texture-type elements with a certain probability. Based on this, the primary color tone can be used as a reference to expand the material rendering attributes of the elements within the material-type elements, forming a variety of different colors, all of which are solid colors. Furthermore, by transforming the UV texture coordinates of the primary color tone in the UV coordinate space, a series of elements with different textures can be formed to enrich the elements within the material-type elements.

[0095] Step S206: Generate a three-dimensional garment based on the expanded garment components.

[0096] The process includes: determining the pattern and texture elements to be used from the expanded pattern and texture category elements; obtaining the corresponding material elements based on the pattern and texture elements to be used; assigning material elements to each component of the 3D clothing to be generated according to preset element allocation rules; and rendering each component based on the assigned material elements to generate the 3D clothing. This achieves diversified generation of 3D clothing. The components of the clothing are semantically informational parts, usually obtained by stitching together two-dimensional patterns. They are the smallest replaceable, recombinable, and deformable units. The two-dimensional pattern, as the smallest unit of clothing, is usually represented as a two-dimensional polygon, which defines the material cutting method. A stitching algorithm can be used to generate 3D clothing.

[0097] The preset element allocation rules include one of the following: each component of the three-dimensional garment to be generated uses the same material element; the main component of the three-dimensional garment to be generated uses a first material element, and other components besides the main component use a second material element.

[0098] Alternatively, the material elements may include: a first type of material element indicating that the material is a solid color texture, and a second type of material element indicating that the material is a patterned texture; the preset element allocation rules may include: the main component of the 3D clothing to be generated uses the first type of material element, and other components other than the main component use the second type of material element; or, the main component of the 3D clothing to be generated uses the second type of material element, and other components other than the main component use the first type of material element.

[0099] This results in a more harmonious overall style for the generated 3D clothing. It should be noted that this step generates more than one type of 3D clothing; the specific number depends on the number of elements in the expanded clothing's constituent elements. Of course, any appropriate generation rules can be set to generate 3D clothing based on elements from only a portion of the clothing's constituent elements.

[0100] Step S208: Display the generated 3D clothing.

[0101] This step enables the rendering and display of 3D clothing.

[0102] According to this embodiment, intelligent clothing design is achieved through the automatic generation of 3D clothing. In this generation process, at least one type of clothing composition element is first obtained based on sample clothing image data, such as pattern, material, and pattern / texture. Further, the elements in the obtained clothing composition elements are expanded according to corresponding element expansion rules. Thus, each element in various clothing composition elements can be expanded from a single element into multiple different elements, achieving the expansion and enrichment of clothing composition elements. Subsequently, a wide variety of 3D clothing can be generated based on the expanded clothing composition elements. Therefore, on the one hand, this greatly enriches the materials for clothing design, meeting diverse design needs; on the other hand, the abundance of materials makes the types and styles of clothing created through material combinations more diverse and varied, reducing the design requirements and costs for designers, and improving design efficiency.

[0103] The data generation method of this embodiment can be executed by any suitable electronic device with data processing capabilities, including but not limited to: servers, mobile terminals (such as mobile phones, PADs, etc.) and PCs.

[0104] Example 4

[0105] Reference Figure 4A The diagram shows a flowchart of a data generation method according to Embodiment 4 of the present invention.

[0106] This embodiment illustrates the data generation method provided by the present invention through a specific example. The data generation method includes the following steps:

[0107] Step S302: Obtain the sample clothing image data and the corresponding text description data.

[0108] For example, web crawlers can be used to collect data from websites such as fashion shows, fashion magazines, social networks, and fashion blogs, resulting in a massive amount of sample clothing image data resources, including image data resources and corresponding textual descriptions. Images and text are the main training data sources for trend analysis. The model is trained using both images and text because images are rich in visual information such as color and color combinations, but lack structured semantic information; while text can provide direct semantic explanations of abstract trends. Therefore, the combination of textual and visual information from images enables a more accurate understanding of trends.

[0109] Step S304: Generate a clothing knowledge graph based on the sample clothing image data and its text description data.

[0110] This includes: using deep learning algorithms to analyze the fashion trends of sample clothing image data and its text description data, including but not limited to clothing color and texture elements, pattern elements, material elements, accessories (materials that play a decorative role in clothing design, such as zippers, buttons, linings, etc.) and suitable matching elements, and statistically analyzing the frequency of occurrence of element tags in the obtained sample clothing image data, and taking the high-frequency element tags as the fashion trend information of the current season.

[0111] For example, for sample clothing image data, the visual perception Match R-CNN network in the DeepFashion model is used to complete the detection and segmentation of clothing subjects, outputting the key points and masks of the clothing in the image. Thus, clothing subject detection and segmentation can identify multiple clothing elements (such as tops and bottoms) in the image as completely as possible, effectively avoiding interference from the background and human skin color in subsequent tasks such as element label recognition and extraction of clothing patterns, textures, and main colors. Subsequently, each segmented clothing element is input into the FashionNet (with a VGG-16 backbone) proposed in the DeepFashion model for element classification and element label recognition, resulting in a series of clothing elements and their corresponding element labels, such as style labels like T-shirt and shirt, and style labels like futuristic print and workwear casual.

[0112] For textual description data, the element tags obtained through the DeepFashion model are first used as the basic corpus data for clothing. For each element tag, a semantic fuzzy matching algorithm is used to count the high-frequency popular words in the textual description data that match the element tag, and the 10 high-frequency words that appear simultaneously with each popular word are recorded. Based on this, the relationships between clothing elements are obtained.

[0113] Then, based on the aforementioned relationships and the correspondence between clothing elements and clothing constituent elements, using the information of clothing constituent elements and element tags as entities, and the relationships between elements, between elements themselves, and between elements as edges, a graph structure is constructed to provide a structured description of clothing—this is the clothing knowledge graph. Through this clothing knowledge graph, the information needed for subsequent expansion of pattern / texture elements and material elements can be guided and obtained.

[0114] Step S306: Obtain pattern and texture elements and material elements based on the clothing knowledge graph.

[0115] In this embodiment, an example is given where the obtained clothing constituent elements include pattern texture elements and material elements.

[0116] Step S308: Expand the pattern texture elements and the material elements respectively.

[0117] It includes:

[0118] (1) Expand the pattern texture elements.

[0119] First, according to the clothing knowledge graph, obtain multiple images containing pattern texture elements, and each image is used as an element of the pattern texture elements. For each element, use it as seed data and perform color expansion and stylistic expansion on it.

[0120] Among them, when performing color expansion, the image of the element can be converted from the RGB space to the HSV space, so as to change the color of the image by rotating the hue ring without changing the light perception, thereby achieving color expansion. An element after color expansion is as shown Figure 4B in the top row on the right side of [Figure 4B]. Only two color expansions are shown in [Figure 4B]. Assume that the original RGB color of the element is blue (shown as a vertical line element in the figure), then through this method, it is expanded to yellow (shown as a diagonal line element in the figure) and red (shown as a horizontal line element in the figure).

[0121] The stylistic expansion of the element can adopt a multi-stroke attention perception algorithm to transfer the design style to be used, such as the artist style (such as Van Gogh style, Picasso style, etc.) to the element, that is, to the pattern texture of the element, to achieve the expansion of the pattern texture from 1 to N. An element after the stylistic expansion in this method is as shown Figure 4B in the middle row on the right side of [Figure 4B]. Figure 4B Only two stylistic expansions are shown in [Figure 4B], which are expanded to Van Gogh style and Picasso style respectively.

[0122] In addition, in this embodiment, the pattern texture pattern of the element is also spliced and expanded based on the Mondrian random process. Specifically, given a two-dimensional image space, it is divided into n sub-spaces through the Mondrian random process algorithm, and m (m < n) sub-spaces are randomly selected to be filled with the pattern texture pattern of the element, while the remaining m - n spaces are filled with the main color of the image of the element. Thus, the further expansion of the element is achieved. An element after the stylistic expansion in this method is as shown Figure 4B in the bottom row on the right side of [Figure 4B]. Figure 4BThe example only illustrates two possible splicing extensions. One result divides the 2D image space into two subspaces: one located at the center of the image, and the other surrounding the central subspace to form an outer subspace. The central subspace is filled with the element's patterned texture, while the outer subspace is filled with the element's dominant color (illustrated as diagonal lines). Another result divides the 2D image space into four subspaces: one located in the lower right corner of the image, filled with the element's patterned texture, and the other three subspaces filled with the element's dominant color (illustrated as diagonal lines).

[0123] (ii) Expand the material elements.

[0124] For example, randomly select a pattern from the expanded pattern patterns (I), and perform a histogram analysis of its image to obtain a color chart of the main color tones of the image (e.g., extract 5 main color tones).

[0125] Then, by transforming the main color tone and UV coordinate space of the pattern and texture, the material elements of the clothing are expanded to form a variety of clothing materials.

[0126] Taking a material sphere as an example of a material-type element, based on the principle of similar colors matching in clothing design, when a material type (such as leather, cotton, linen, etc.) is selected, the material sphere will be assigned a solid color with a 30% probability and a patterned texture image as the material's diffuse map with a 70% probability. When the material sphere is assigned a solid color, the color will be taken from the main color tone of the patterned texture image with a 50% probability and from the complementary color of the main color tone of the patterned texture image with a 50% probability. When the material sphere uses a patterned texture image as the material's diffuse map, a series of material spheres with different appearances can be generated through different rotations, translations, and scaling transformations in the UV coordinate space. The element expansion process of a material-type element is as follows: Figure 4C As shown, the expansion results in five new types of material spheres.

[0127] It should be noted that the expansion of the above-mentioned pattern and texture elements and the expansion of the material elements can be performed in any order or in parallel.

[0128] Step S310: Generate a variety of three-dimensional clothing based on the expanded elements.

[0129] For example, such as Figure 4D As shown, for a certain fixed pattern ( Figure 4DIn step (d), firstly, suitable pattern and texture images are selected as seed data based on the clothing knowledge graph; then, as mentioned earlier, they are expanded in terms of color and style diversity (generating a large number of similar but different new data from a small amount of data), and the main color tone swatches of the expanded pattern and texture images are extracted. Figure 4D (a)); Next, using the clothing knowledge graph, select a series of materials that match the style. Figure 4D (b)); Based on the selected pattern texture image and material sphere, a method is proposed to generate diverse material spheres based on the main color tone of the pattern texture and the UV coordinate space transformation of the pattern texture.

[0130] In the preprocessing stage, 3D clothing models can be segmented based on semantics, dividing them into multiple components such as the main body, sleeves, collar, skirt, pants, pockets, and decorative elements like buttons. During 3D clothing generation, for a selected patterned texture image, each component of the clothing can be assigned a randomly selected, diverse material sphere according to the following rules: 1) The entire garment can use one material sphere; 2) The main body and other related components can use different material spheres. If the main body uses a material sphere with a patterned texture, then other related components use solid-color material spheres, and vice versa; 3) Only one type of patterned texture material sphere and only one type of solid-color material sphere are allowed on the entire garment. Following these rules, different material spheres can be selected for different components of the clothing, thus achieving a diverse recombination scheme for clothing patterns, textures, and materials. Finally, the generated diverse 3D clothing is rendered into an image.

[0131] In one alternative approach, the color distribution domain in the generated 3D clothing image can be used to measure the design appeal of the 3D clothing. A wider color distribution indicates the use of more colors in the clothing design, thus filtering out design-oriented options from a large variety of clothing designs. For example, deep learning algorithms can be used to collect popular designs similar to the pattern as references. The similarity between the generated 3D clothing and the reference design images can be calculated at the image level. The higher the similarity, the more aesthetically pleasing the design, and the more likely it is to be selected as the final generated result.

[0132] This embodiment applies deep learning algorithms to the field of apparel design. Through trend analysis, a knowledge graph of apparel is constructed, which is then used to expand elements in two independent dimensions: pattern / texture and material. By combining and recombining patterns / textures and materials, more diverse apparel is generated, reducing the cost and requirements of apparel design. Furthermore, for the generated 3D apparel, intelligent aesthetic evaluation is applied. Image similarity metrics are used to select design schemes that conform to popular apparel aesthetics from numerous options, ensuring both the quantity and quality of intelligent apparel design.

[0133] The data generation method of this embodiment can be executed by any suitable electronic device with data processing capabilities, including but not limited to: servers, mobile terminals (such as mobile phones, PADs, etc.) and PCs.

[0134] Example 5

[0135] Reference Figure 5 The diagram shows a structural block diagram of a data generation apparatus according to Embodiment 5 of the present invention.

[0136] The data generation device of this embodiment includes: an acquisition module 402, used to acquire at least one type of clothing constituent elements based on sample clothing image data; an expansion module 404, used to determine the element expansion rules corresponding to the at least one type of clothing constituent elements, and to expand the elements in the at least one type of clothing constituent elements according to the element expansion rules; and a generation module 406, used to generate three-dimensional clothing based on the expanded clothing constituent elements.

[0137] Optionally, the clothing components include pattern and texture elements; the extension module 404 is used to determine the color extension rules and / or style extension rules corresponding to the pattern and texture elements; perform color extension on the elements in the pattern and texture elements according to the color extension rules; and / or perform style extension on the elements in the pattern and texture elements according to the style extension rules.

[0138] Optionally, when the expansion module 404 performs color expansion on the elements in the patterned texture category according to the color expansion rules, it performs color expansion on the elements in the patterned texture category by rotating the hue ring at a set angle according to the colors corresponding to the elements in the patterned texture category.

[0139] Optionally, when the expansion module 404 expands the color of the elements in the pattern and texture category by rotating the hue ring at a set angle according to the colors corresponding to the elements in the pattern and texture category, it converts the color space of the elements in the pattern and texture category from RGB space to HSV space; in the HSV space, it expands the color of the elements by rotating the hue ring at a set angle according to the colors corresponding to the elements.

[0140] Optionally, when the extension module 404 performs style extension on the elements in the pattern and texture category according to the style extension rules, it obtains information on the design style to be used; and uses a multi-stroke attention perception algorithm to transfer the design style indicated by the information on the design style to be used to the elements in the pattern and texture category, so as to perform style extension on the elements in the pattern and texture category.

[0141] Optionally, when the extension module 404 performs style extension on the elements in the pattern and texture category according to the style extension rules, it obtains the style data of the elements in the pattern and texture category; and uses the Mondrian random process algorithm to splice the style data, and obtains the style extension of the elements based on the splicing result.

[0142] Optionally, when the extension module 404 splices the style data using the Mondrian random process algorithm and obtains the style extension of the element based on the splicing result, it divides the preset two-dimensional image space into multiple subspaces using the Mondrian random process algorithm; selects a portion of the space from the multiple subspaces to fill it with the style data; and fills the remaining space from the multiple subspaces except for the selected portion with the main color tone of the two-dimensional image space; based on the filling result, splices the style data and the main color tone, and obtains the style extension of the element in the pattern texture category based on the splicing result.

[0143] Optionally, the clothing components include material elements; the extension module 404 is used to determine the material rendering extension rules corresponding to the material elements; and to extend the material rendering attributes of the elements in the material elements according to the material rendering extension rules.

[0144] Optionally, when the extension module 404 extends the material rendering attributes of the elements in the material category elements according to the material rendering extension rules, it obtains the main color tone of the elements in the pattern texture category elements after extension; and extends the material rendering attributes of the elements according to the main color tone and the UV texture space transformation of the material ball.

[0145] Optionally, when the extension module 404 extends the material rendering attributes of the element according to the main color and the UV texture space transformation of the material ball, it performs a transformation operation in the UV texture space of the material ball; according to the transformation operation, it extends the material rendering attributes of the material ball on the pattern texture and / or the main color; wherein the transformation operation includes at least one of the following: rotation operation, translation operation, scaling operation.

[0146] Optionally, the generation module 406 is used to determine the pattern texture elements to be used from the expanded pattern texture class elements; obtain the corresponding material elements according to the pattern texture elements to be used; assign material elements to each component of the three-dimensional clothing to be generated according to the preset element allocation rules; and render each component according to the assigned material elements to generate the three-dimensional clothing.

[0147] Optionally, the preset element allocation rules include one of the following: each component of the three-dimensional garment to be generated uses the same material element; the main component of the three-dimensional garment to be generated uses a first material element, and other components besides the main component use a second material element.

[0148] Optionally, the material elements include: a first type of material element indicating that the material is a solid color texture, and a second type of material element indicating that the material is a patterned texture; the preset element allocation rules include: the main component of the three-dimensional garment to be generated uses the first type of material element, and other components other than the main component use the second type of material element; or, the main component of the three-dimensional garment to be generated uses the second type of material element, and other components other than the main component use the first type of material element.

[0149] Optionally, the acquisition module 402 is used to acquire at least one type of clothing constituent elements by generating a clothing knowledge graph based on sample clothing image data.

[0150] Optionally, the data generation device in this embodiment further includes a graph module 408, which is used to acquire sample clothing image data and text description data corresponding to the sample clothing image data before the acquisition module 402 acquires at least one type of clothing constituent element by generating a clothing knowledge graph based on the sample clothing image data; obtain information about clothing elements and element tags corresponding to the clothing elements through a neural network model for clothing element detection; obtain the association relationship between the clothing elements based on the element tags and the text description data; and generate a clothing knowledge graph based on the association relationship and the correspondence between the clothing elements and the clothing constituent elements.

[0151] The data generation apparatus of this embodiment is used to implement the corresponding data generation methods in the foregoing method embodiments and has the beneficial effects of the corresponding method embodiments, which will not be repeated here. Furthermore, the functional implementation of each module in the data generation apparatus of this embodiment can be referred to the description of the corresponding part in the foregoing method embodiments, which will also not be repeated here.

[0152] Example 6

[0153] Reference Figure 6 The diagram shows a structural schematic of an electronic device according to Embodiment Six of the present invention. The specific embodiments of the present invention do not limit the specific implementation of the electronic device.

[0154] like Figure 6As shown, the electronic device may include: a processor 502, a communications interface 504, a memory 506, and a communications bus 508.

[0155] in:

[0156] The processor 502, communication interface 504, and memory 506 communicate with each other via communication bus 508.

[0157] Communication interface 504 is used to communicate with other electronic devices or servers.

[0158] The processor 502 is used to execute program 510, specifically to perform the relevant steps in the above-described data generation method embodiment.

[0159] Specifically, program 510 may include program code that includes computer operation instructions.

[0160] Processor 502 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The smart device may include one or more processors of the same type, such as one or more CPUs; or it may include processors of different types, such as one or more CPUs and one or more ASICs.

[0161] Memory 506 is used to store program 510. Memory 506 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one type of disk storage.

[0162] Specifically, program 510 can be used to enable processor 502 to perform the following operations: based on sample clothing image data, obtain at least one type of clothing constituent elements; determine the element expansion rules corresponding to the at least one type of clothing constituent elements, expand the elements in the at least one type of clothing constituent elements according to the element expansion rules; and generate three-dimensional clothing based on the expanded clothing constituent elements.

[0163] In one optional implementation, the clothing elements include pattern and texture elements; the program 510 is further configured to cause the processor 502, when determining an element expansion rule corresponding to at least one type of clothing element, and expanding the elements in the at least one type of clothing element according to the element expansion rule: determining a color expansion rule and / or a style expansion rule corresponding to the pattern and texture element; expanding the elements in the pattern and texture element by color according to the color expansion rule; and / or expanding the elements in the pattern and texture element by style according to the style expansion rule.

[0164] In an optional implementation, program 510 is further configured to cause processor 502 to perform color expansion on elements in the patterned texture category according to the color expansion rules: based on the colors corresponding to the elements in the patterned texture category, perform color expansion on the elements in the patterned texture category by rotating a hue ring at a set angle.

[0165] In an optional implementation, program 510 is further configured to cause processor 502 to, when expanding the color of elements in the patterned texture category by rotating a color wheel at a set angle according to the colors corresponding to the elements in the patterned texture category: convert the color space of the elements in the patterned texture category from RGB space to HSV space; and in the HSV space, expand the color of the elements by rotating the color wheel at a set angle according to the colors corresponding to the elements.

[0166] In an optional implementation, program 510 is further configured to cause processor 502, when performing style expansion on elements in the patterned texture category according to the style expansion rules, to: obtain information on the design style to be used; and, through a multi-stroke attention perception algorithm, transfer the design style indicated by the information on the design style to be used to the elements in the patterned texture category, so as to perform style expansion on the elements in the patterned texture category.

[0167] In an optional implementation, program 510 is further configured to cause processor 502, when performing style expansion on elements in the patterned texture category elements according to the style expansion rules, to: obtain style data of elements in the patterned texture category elements; and to concatenate the style data using the Mondrian random process algorithm, and obtain the style expansion of the elements based on the concatenation result.

[0168] In an optional implementation, program 510 is further configured to cause processor 502, when splicing the style data using the Mondrian random process algorithm and obtaining the style extension of the element based on the splicing result: dividing a preset two-dimensional image space into multiple subspaces using the Mondrian random process algorithm; selecting a portion of the space from the multiple subspaces to fill it with the style data; filling the remaining space from the multiple subspaces except for the selected portion with the main color tone of the two-dimensional image space; splicing the style data and the main color tone based on the filling result; and obtaining the style extension of the element in the pattern texture category based on the splicing result.

[0169] In one optional implementation, the clothing components include material elements; the program 510 is further configured to cause the processor 502, when determining an element expansion rule corresponding to at least one type of clothing component and expanding the elements in the at least one type of clothing component according to the element expansion rule, to: determine a material rendering expansion rule corresponding to the material element; and expand the material rendering attributes of the elements in the material element according to the material rendering expansion rule.

[0170] In an optional implementation, program 510 is further configured to cause processor 502, when performing material rendering attribute extension on elements in the material class element according to the material rendering extension rules, to: obtain the main color tone of the elements in the pattern texture class element after extension; and perform material rendering attribute extension on the elements according to the main color tone and the UV texture space transformation of the material ball.

[0171] In an optional implementation, program 510 is further configured to cause processor 502 to perform a transformation operation in the UV texture space of the material sphere when expanding the material rendering attributes of the element according to the main color and the UV texture space transformation of the material sphere: performing a transformation operation in the UV texture space of the material sphere; expanding the material rendering attributes of the material sphere on the pattern texture and / or main color according to the transformation operation; wherein the transformation operation includes at least one of the following: rotation operation, translation operation, scaling operation.

[0172] In an optional implementation, program 510 is further configured to cause processor 502, when generating a three-dimensional garment based on the expanded garment constituent elements, to: determine the pattern texture element to be used from the expanded pattern texture elements; obtain the corresponding material element based on the pattern texture element to be used; allocate material elements to each component of the three-dimensional garment to be generated according to a preset element allocation rule; and render each component based on the allocated material elements to generate the three-dimensional garment.

[0173] In one optional implementation, the preset element allocation rule includes one of the following: each component of the three-dimensional garment to be generated uses the same material element; the main component of the three-dimensional garment to be generated uses a first material element, and other components besides the main component use a second material element.

[0174] In one optional implementation, the material elements include: a first type of material element indicating that the material is a solid color texture, and a second type of material element indicating that the material is a patterned texture; the preset element allocation rules include: the main component of the three-dimensional garment to be generated uses the first type of material element, and other components other than the main component use the second type of material element; or, the main component of the three-dimensional garment to be generated uses the second type of material element, and other components other than the main component use the first type of material element.

[0175] In an alternative implementation, program 510 is further configured to cause processor 502 to obtain at least one type of clothing constituent element by means of a clothing knowledge graph generated from the sample clothing image data when obtaining at least one type of clothing constituent element based on the sample clothing image data.

[0176] In an optional implementation, program 510 is further configured to cause processor 502 to acquire sample clothing image data and text description data corresponding to the sample clothing image data before acquiring at least one type of clothing constituent element through the clothing knowledge graph generated from the sample clothing image data; obtain information about clothing elements and element labels corresponding to the clothing elements through a neural network model for clothing element detection; obtain the association relationship between the clothing elements based on the element labels and the text description data; and generate a clothing knowledge graph based on the association relationship and the correspondence between the clothing elements and the clothing constituent elements.

[0177] The specific implementation of each step in program 510 can be found in the corresponding steps and units described in the above-mentioned data generation method embodiments, and will not be repeated here. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the devices and modules described above can be referred to the corresponding process descriptions in the aforementioned method embodiments, and will not be repeated here.

[0178] The electronic device in this embodiment enables intelligent clothing design through the automatic generation of 3D clothing. In this generation process, at least one type of clothing element is first obtained based on sample clothing image data, such as pattern, material, and pattern / texture. Further, the elements within these clothing elements are expanded according to corresponding element expansion rules. Thus, each element in various clothing elements can be expanded from a single element into multiple different elements, achieving expansion and enrichment of clothing elements. Subsequently, a wide variety of 3D clothing can be generated based on the expanded clothing elements. Therefore, on the one hand, this greatly enriches the materials available for clothing design, meeting diverse design needs; on the other hand, the abundance of materials allows for a greater variety and diversity of clothing types and styles created through material combinations, reducing design requirements and costs for designers and improving design efficiency.

[0179] It should be noted that, depending on the implementation needs, the various components / steps described in the embodiments of the present invention can be broken down into more components / steps, or two or more components / steps or parts of the operation of components / steps can be combined into new components / steps to achieve the purpose of the embodiments of the present invention.

[0180] The methods described above according to embodiments of the present invention can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium (such as a CD-ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code originally stored on a remote recording medium or a non-transitory machine-readable medium and subsequently stored on a local recording medium, downloaded via a network. Thus, the methods described herein can be stored as software processing on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It is understood that the computer, processor, microprocessor controller, or programmable hardware includes storage components (e.g., RAM, ROM, flash memory, etc.) capable of storing or receiving software or computer code that, when accessed and executed by the computer, processor, or hardware, implements the data generation methods described herein. Furthermore, when a general-purpose computer accesses code used to implement the data generation methods shown herein, the execution of the code transforms the general-purpose computer into a dedicated computer for executing the data generation methods shown herein.

[0181] Those skilled in the art will recognize that the units and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the embodiments of the present invention.

[0182] The above embodiments are only used to illustrate the embodiments of the present invention, and are not intended to limit the embodiments of the present invention. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the embodiments of the present invention. Therefore, all equivalent technical solutions also fall within the scope of the embodiments of the present invention, and the patent protection scope of the embodiments of the present invention should be defined by the claims.

Claims

1. A data generation method, comprising: Obtain the constituent elements of the target object, wherein the constituent elements of the target object are clothing constituent elements, and the clothing constituent elements include at least pattern and texture elements; Determine the element expansion rules corresponding to the constituent elements of the target object; Based on the aforementioned element expansion rules, the elements in the constituent elements of the target object are expanded. Generate result data based on the expanded constituent elements of the target object; The rule for expanding elements corresponding to the constituent elements of the target object is determined. According to the aforementioned element expansion rules, the elements in the constituent elements of the target object are expanded, including: Determine the style extension rules corresponding to the aforementioned pattern and texture elements; According to the style expansion rules, the elements in the pattern and texture category are style-expanded, including: obtaining style data of the elements in the pattern and texture category; and using the Mondrian random process algorithm to concatenate the style data, and obtaining the style expansion of the elements based on the concatenation result. The method further includes: obtaining information about clothing elements through a neural network model for clothing element detection; generating a clothing knowledge graph based on the relationships between clothing elements and the correspondence between the clothing elements and the clothing constituent elements; wherein the clothing knowledge graph is used to obtain the clothing constituent elements and the information required for element expansion.

2. The method according to claim 1, wherein, The method is used for generating 3D clothing.

3. The method according to claim 2, wherein, The step of determining the element expansion rules corresponding to the constituent elements of the target object, and expanding the elements in the constituent elements of the target object according to the element expansion rules, further includes: Determine the color extension rules corresponding to the aforementioned pattern and texture elements; According to the color expansion rules, the elements in the pattern and texture category are expanded in color.

4. The method according to claim 3, wherein, According to the color expansion rules, the elements in the pattern and texture category are color expanded, including: Based on the colors corresponding to the elements in the color and texture category, the colors of the elements in the color and texture category are expanded by rotating the color wheel at a set angle.

5. The method according to claim 4, wherein, The step of expanding the colors of the elements in the pattern and texture category by rotating the color wheel at a set angle according to the colors corresponding to the elements in the pattern and texture category includes: Convert the color space of the elements in the aforementioned floral texture category from RGB space to HSV space; In the HSV space, the color of the element is expanded by rotating the color wheel according to a set angle.

6. The method according to claim 1, wherein, The step of concatenating the style data using the Mondrian random process algorithm and obtaining the style extension of the element based on the concatenation result includes: The pre-defined two-dimensional image space is divided into multiple subspaces using the Mondrian random process algorithm; A portion of the subspaces are selected and filled with the style data; the remaining spaces in the subspaces other than the selected spaces are filled with the dominant color of the two-dimensional image space. Based on the filling result, the style data and the main color are spliced ​​together, and the style extension of the elements in the pattern and texture category is obtained based on the splicing result.

7. The method according to claim 1, wherein, The elements constituting the clothing include material elements; The step of determining the element expansion rules corresponding to the constituent elements of the target object; and expanding the elements in the constituent elements of the target object according to the element expansion rules, includes: Determine the material rendering extension rules corresponding to the material class elements; extend the material rendering attributes of the elements in the material class elements according to the material rendering extension rules.

8. The method according to claim 7, wherein, The step of extending the material rendering attributes of elements in the material class element according to the material rendering extension rules includes: Obtain the primary color tone of the elements in the expanded pattern texture category; based on the primary color tone and the UV texture space transformation of the material sphere, expand the material rendering attributes of the elements.

9. The method according to claim 8, wherein, The step of extending the material rendering attributes of the element based on the main color tone and the UV texture space transformation of the material sphere includes: Perform transformation operations in the UV texture space of the material sphere; based on the transformation operations, extend the material rendering attributes of the material sphere in terms of pattern texture and / or main color tone; The transformation operation includes at least one of the following: rotation operation, translation operation, and scaling operation.

10. The method according to any one of claims 7-9, wherein, The step of generating result data based on the expanded target object constituent elements includes: From the expanded pattern and texture category elements, determine the pattern and texture elements to be used; Based on the pattern and texture elements to be used, obtain the corresponding material elements; According to the preset element allocation rules, material elements are assigned to each component of the 3D clothing to be generated; Based on the assigned material elements, each component is rendered to generate 3D clothing.

11. The method according to claim 10, wherein, The preset element allocation rules include one of the following: The components of the three-dimensional garment to be generated use the same material elements; The main component of the 3D clothing to be generated uses a first material element, while other components besides the main component use a second material element.

12. The method according to claim 10, wherein, The material elements include: a first type of material element used to indicate that the material is a solid color texture, and a second type of material element used to indicate that the material is a patterned texture; The preset element allocation rules include: The main component of the 3D clothing to be generated uses a first type of material element, while other components besides the main component use a second type of material element. or, The main component of the 3D clothing to be generated uses a second type of material element, while other components besides the main component use a first type of material element.

13. The method according to any one of claims 2-9, wherein, The acquisition of the constituent elements of the target object includes: Based on the sample clothing image data, obtain at least one type of clothing composition element.

14. The method according to claim 13, wherein, The step of obtaining at least one type of clothing composition element based on sample clothing image data includes: By generating a clothing knowledge graph based on sample clothing image data, at least one type of clothing constituent element can be obtained.

15. The method according to claim 1, wherein, The method further includes: Obtain sample clothing image data and corresponding text description data; Information about clothing elements and corresponding element labels are obtained through a neural network model for clothing element detection. The relationships between the clothing elements are obtained based on the element tags and the text description data.

16. The method according to claim 1, wherein, The constituent elements of the target object include one of the following: architectural constituent elements, decoration constituent elements, and set design constituent elements.

17. The method according to claim 1, wherein, The method further includes: The result data is sent to an industrial control device to display the result data and to control the production of the item corresponding to the result data.

18. A data generation apparatus, comprising: The acquisition module is used to acquire the constituent elements of the target object, wherein the constituent elements of the target object are clothing constituent elements, and the clothing constituent elements include at least pattern and texture elements. An extension module is used to determine the element extension rules corresponding to the constituent elements of the target object; and to extend the elements in the constituent elements of the target object according to the element extension rules. The generation module is used to generate result data based on the constituent elements of the expanded target object; The expansion module is used to determine the element expansion rules corresponding to the constituent elements of the target object; According to the element expansion rules, the elements in the constituent elements of the target object are expanded, including: determining the style expansion rules corresponding to the pattern and texture elements; obtaining the style data of the elements in the pattern and texture elements; and using the Mondrian random process algorithm to concatenate the style data, and obtaining the style expansion of the elements based on the concatenation result. The knowledge graph module is used to obtain information about clothing elements through a neural network model for clothing element detection; and to generate a clothing knowledge graph based on the relationships between clothing elements and the correspondence between the clothing elements and the clothing constituent elements; wherein the clothing knowledge graph is used to obtain the clothing constituent elements and the information required for element expansion.

19. An electronic device comprising: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction that causes the processor to perform an operation corresponding to the data generation method as described in any one of claims 1-17.

20. A computer storage medium having a computer program stored thereon, which, when executed by a processor, implements the data generation method as described in any one of claims 1-17.

Images