Fabric fusion method and apparatus, and product color matching method and apparatus

Abstract

One or more embodiments of the present description provide a fabric fusion method and apparatus, and a product color matching method and apparatus. The fabric fusion method comprises: acquiring a standard fabric figure and a product style figure; fusing the product style figure and the standard fabric figure to obtain an intermediate effect figure; determining, as a fabric figure to undergo fusion, image content in the standard fabric figure for fusion with the product style figure; and fusing the intermediate effect figure and the fabric figure to undergo fusion to obtain a final effect figure.

Description

Fabric fusion method and device, product color matching method and device Technical Field

[0001] One or more embodiments of this specification relate to the field of computer application technology, and in particular, to a fabric fusion method and device, and a product color matching method and device. Background Art

[0002] In the manufacturing industry, the manufacture of many products requires the use of fabrics and color matching, such as clothing, luggage, home textiles, toys and so on.

[0003] Taking the clothing industry as an example, designers typically don't just design a single color palette for a particular style; they offer a comprehensive suite of color combinations to meet the aesthetic needs and fashion trends of diverse consumers. Fabric-to-garment refers to the entire manufacturing process from raw fabric to finished garment. This meticulous process encompasses multiple key steps, from selecting the appropriate fabric, through cutting, sewing, and embellishment, all the way to the final finished product. This process integrates the fabric's characteristics with the design concept to create a flawless fashion piece.

[0004] The clothing design process requires determining matching fabrics and styles, and determining the desired effect, which requires extensive trial and error. Experimenting with different fabrics and colorways for each style often requires significant investment in the production of a single garment. Similarly, for other fabric-based products like luggage, home textiles, and toys, experimenting with different fabrics and colorways for each style often requires significant investment in the production of the final product.

[0005] This method often requires a lot of manpower and has poor real-time performance. Summary of the Invention

[0006] In view of this, one or more embodiments of this specification provide a fabric fusion method and device, as well as a product color matching method and device.

[0007] One or more embodiments of this specification provide the following technical solutions:

[0008] According to a first aspect of one or more embodiments of the present specification, a fabric fusion method is proposed, including: obtaining a standard fabric image and a product style image; fusing the product style image and the standard fabric image to obtain an intermediate rendering; determining image content in the standard fabric image for fusion with the product style image as a fabric image to be fused; and fusing the intermediate rendering and the fabric image to be fused to obtain a final rendering.

[0009] Optionally, the fusing of the intermediate rendering and the fabric image to be fused to obtain a final rendering includes: inputting the intermediate rendering and the fabric image to be fused into a preset model to obtain a final rendering output by the preset model; the preset model is used to reduce the loss of fabric information; the loss of fabric information in the final rendering is less than the loss of fabric information in the intermediate rendering.

[0010] Optionally, the method further includes: obtaining target fabric information of the standard fabric image; fusing the intermediate rendering and the fabric image to be fused to obtain a final rendering, including: fusing the obtained target fabric information, the intermediate rendering and the fabric image to be fused to obtain a final rendering.

[0011] Optionally, the fusing of the intermediate rendering and the fabric image to be fused to obtain the final rendering includes: looping through the following steps until a preset loop stop condition is met: fusing the current intermediate rendering and the fabric image to be fused to obtain a new rendering; determining the obtained new rendering as the current intermediate rendering; and after the loop stops, determining the current intermediate rendering as the final rendering.

[0012] Optionally, the determining of the image content in the standard fabric map for merging with the product style map as the fabric map to be fused includes: determining the image content in the standard fabric map for merging with the product style map as the fabric map to be fused based on the intermediate rendering; or overlapping the product outline in the product style map with the standard fabric map in a preset manner, and determining the image content contained in the product outline in the standard fabric map as the fabric map to be fused.

[0013] Optionally, the obtaining of the standard fabric map includes: obtaining an initial fabric map; and preprocessing the initial fabric map to obtain a standard fabric map.

[0014] Optionally, the method for obtaining the product style map includes: obtaining an initial product image; performing image recognition on the initial product image to determine the product image content in the initial product image; and determining the product style map based on the product image content recognized in the initial product image.

[0015] Optionally, the method further includes: obtaining target parameters; obtaining a standard fabric map, including: obtaining an initial fabric map; determining repeated content of the fabric map in the initial fabric map based on the target parameters; and splicing several copies of the repeated content of the fabric map to obtain a standard fabric map.

[0016] Optionally, the method also includes: obtaining target parameters; determining the image content in the standard fabric map used for fusion with the product style map as the fabric map to be fused, including: overlapping the product outline in the product style map with the standard fabric map according to the relative position in the target parameters, and determining the image content contained in the product outline in the standard fabric map as the fabric map to be fused.

[0017] According to a second aspect of one or more embodiments of the present specification, another fabric fusion method is proposed, including: obtaining a standard fabric image and a clothing style image; fusing the clothing style image and the standard fabric image to obtain an intermediate clothing rendering image; determining the image content in the standard fabric image used for fusion with the clothing style image as the fabric image to be fused; fusing the intermediate clothing rendering image and the fabric image to be fused to obtain a final clothing rendering image.

[0018] According to a third aspect of one or more embodiments of the present specification, a fabric fusion device is proposed, including: an acquisition unit for acquiring a standard fabric image and a product style image; a first fusion unit for fusing the product style image and the standard fabric image to obtain an intermediate rendering; a determination unit for determining image content in the standard fabric image for fusion with the product style image as a fabric image to be fused; and a second fusion unit for fusing the intermediate rendering and the fabric image to be fused to obtain a final rendering.

[0019] Optionally, the second fusion unit is used to: input the intermediate rendering and the fabric image to be fused into a preset model to obtain a final rendering output by the preset model; the preset model is used to reduce the loss of fabric information; the loss of fabric information in the final rendering is less than the loss of fabric information in the intermediate rendering.

[0020] Optionally, the acquisition unit is further used to: acquire target fabric information of the standard fabric image; and the second fusion unit is used to: fuse the acquired target fabric information, the intermediate rendering and the fabric image to be fused to obtain a final rendering.

[0021] Optionally, the second fusion unit is used to: loop through the following steps until a preset loop stop condition is met: fuse the current intermediate rendering and the fabric image to be fused to obtain a new rendering; determine the obtained new rendering as the current intermediate rendering; after the loop stops, determine the current intermediate rendering as the final rendering.

[0022] Optionally, the determination unit is used to: determine the image content in the standard fabric map used for fusion with the product style map as the fabric map to be fused based on the intermediate rendering; or overlap the product outline in the product style map with the standard fabric map in a preset manner, and determine the image content contained in the product outline in the standard fabric map as the fabric map to be fused.

[0023] Optionally, the acquisition unit is used to: acquire an initial fabric image; and preprocess the initial fabric image to obtain a standard fabric image.

[0024] Optionally, the acquisition unit is used to: acquire an initial product image; perform image recognition on the initial product image to determine product image content in the initial product image; and determine a product style image based on the product image content recognized in the initial product image.

[0025] Optionally, the acquisition unit is further configured to: acquire target parameters; acquire an initial fabric map; determine repeated content of the fabric map in the initial fabric map based on the target parameters; and splice together several copies of the repeated content of the fabric map to obtain a standard fabric map.

[0026] Optionally, the acquisition unit is also used to: acquire target parameters; the determination unit is used to: overlap the product outline in the product style diagram with the standard fabric diagram according to the relative position in the target parameters, and determine the image content contained in the product outline in the standard fabric diagram as the fabric diagram to be fused.

[0027] According to a fourth aspect of one or more embodiments of the present specification, another fabric fusion device is proposed, including: an image acquisition unit, used to acquire a standard fabric image and a clothing style image; a third fusion unit, used to fuse the clothing style image and the standard fabric image to obtain an intermediate clothing effect image; a unit to be fused, used to determine the image content in the standard fabric image for fusion with the clothing style image as the fabric image to be fused; a fourth fusion unit, used to fuse the intermediate clothing effect image and the fabric image to be fused to obtain a final clothing effect image.

[0028] According to a third aspect of one or more embodiments of this specification, a product color matching method is proposed, including: obtaining a product style image and color matching information; the color matching information includes at least one color information; fusing the product style image and the color matching information to obtain an intermediate rendering; fusing the intermediate rendering and the color matching information to obtain a final rendering.

[0029] Optionally, the fusing of the intermediate rendering and the color matching information to obtain the final rendering includes: inputting the intermediate rendering and the color matching information into a preset model to obtain a final rendering output by the preset model; the preset model is used to reduce the loss of color information; the loss of color information in the final rendering is less than the loss of color information in the intermediate rendering.

[0030] Optionally, the method further includes: obtaining atmosphere information corresponding to the color matching information; fusing the intermediate rendering and the color matching information to obtain a final rendering, including: fusing the obtained atmosphere information, the intermediate rendering and the color matching information to obtain a final rendering.

[0031] Optionally, the fusing of the intermediate rendering and the color matching information to obtain the final rendering includes: looping through the following steps until a preset loop stop condition is met: fusing the current intermediate rendering and the color matching information to obtain a new rendering; determining the obtained new rendering as the current intermediate rendering; after the loop stops, determining the current intermediate rendering as the final rendering.

[0032] Optionally, the color matching information includes: a set of regional color matching correspondences; the regional color matching correspondence set includes at least one set of correspondences between colors and regions in the product style diagram; the fusion of the product style diagram and the color matching information to obtain an intermediate rendering includes: according to the regional color matching correspondence set, for the regions in the product style diagram, fusing corresponding colors to obtain an intermediate rendering; the fusion of the intermediate rendering and the color matching information to obtain a final rendering includes: according to the regional color matching correspondence set, for the regions in the intermediate rendering, fusing corresponding colors to obtain a final rendering.

[0033] Optionally, the method further includes: filling corresponding colors for the areas in the product style diagram according to the regional color correspondence set to obtain a color reference diagram; fusing corresponding colors for the areas in the intermediate rendering according to the regional color correspondence set to obtain a final rendering, including: fusing the intermediate rendering and the color reference diagram to obtain a final rendering.

[0034] Optionally, obtaining a product style diagram includes: obtaining an initial product image; performing image recognition on the initial product image to determine product image content in the initial product image; and determining a product style diagram based on the product image content recognized in the initial product image.

[0035] Optionally, the method for determining the regional color correspondence set includes any one of the following items: obtaining a color specified for the region in the product style diagram; based on the specified color, constructing at least one set of correspondences between colors and regions in the product style diagram, to obtain a set of regional color correspondences; obtaining an atmosphere prompt specified for the region in the product style diagram, and automatically generating a corresponding color based on the obtained atmosphere prompt; based on the generated color, constructing at least one set of correspondences between colors and regions in the product style diagram, to obtain a set of regional color correspondences; obtaining a color scheme; for the region in the product style diagram, randomly selecting a color from the obtained color scheme as the corresponding color, constructing at least one set of correspondences between colors and regions in the product style diagram, to obtain a set of regional color correspondences.

[0036] According to a fourth aspect of one or more embodiments of this specification, a product color matching device is proposed, including: an acquisition unit for acquiring a product style image and color matching information; the color matching information includes at least one color information; a first fusion unit for fusing the product style image and the color matching information to obtain an intermediate rendering; a second fusion unit for fusing the intermediate rendering and the color matching information to obtain a final rendering.

[0037] Optionally, the second fusion unit is used to: input the intermediate rendering and the color matching information into a preset model to obtain a final rendering output by the preset model; the preset model is used to reduce the loss of color information; the loss of color information in the final rendering is less than the loss of color information in the intermediate rendering.

[0038] Optionally, the acquisition unit is further configured to: acquire atmosphere information corresponding to the color matching information; and the second fusion unit is configured to: fuse the acquired atmosphere information, the intermediate rendering and the color matching information to obtain a final rendering.

[0039] Optionally, the second fusion unit is used to: loop through the following steps until a preset loop stop condition is met: fuse the current intermediate rendering and the color matching information to obtain a new rendering; determine the obtained new rendering as the current intermediate rendering; after the loop stops, determine the current intermediate rendering as the final rendering.

[0040] Optionally, the color information includes: a set of regional color correspondences; the regional color correspondence set includes at least one set of correspondences between colors and regions in the product style diagram; the first fusion unit is used to: according to the regional color correspondence set, for the region in the product style diagram, fuse the corresponding colors to obtain an intermediate rendering; the second fusion unit is used to: according to the regional color correspondence set, for the region in the intermediate rendering, fuse the corresponding colors to obtain a final rendering.

[0041] Optionally, the device also includes a reference unit, which is used to: fill the corresponding colors for the areas in the product style image according to the regional color matching correspondence set to obtain a color reference image; and a second fusion unit is used to: fuse the intermediate rendering and the color reference image to obtain a final rendering.

[0042] Optionally, the acquisition unit is used to: acquire an initial product image; perform image recognition on the initial product image to determine product image content in the initial product image; and determine a product style image based on the product image content recognized in the initial product image.

[0043] Optionally, the method for determining the regional color correspondence set includes any one of the following items: obtaining a color specified for the region in the product style diagram; based on the specified color, constructing at least one set of correspondences between colors and regions in the product style diagram, to obtain a set of regional color correspondences; obtaining an atmosphere prompt specified for the region in the product style diagram, and automatically generating a corresponding color based on the obtained atmosphere prompt; based on the generated color, constructing at least one set of correspondences between colors and regions in the product style diagram, to obtain a set of regional color correspondences; obtaining a color scheme; for the region in the product style diagram, randomly selecting a color from the obtained color scheme as the corresponding color, constructing at least one set of correspondences between colors and regions in the product style diagram, to obtain a set of regional color correspondences.

[0044] According to the above embodiments, the effect diagram is obtained by automatically fusing fabric images to achieve fabric fusion, and the effect diagram of product color matching is obtained by automatically fusing color matching colors and product style images, thereby reducing manual participation and improving the efficiency and real-time performance of viewing product effects.

[0045] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0047] FIG1 is a schematic diagram showing the principle of a fabric fusion process provided by an exemplary embodiment.

[0048] FIG2 is a flow chart of a fabric fusion method provided by an exemplary embodiment.

[0049] FIG. 3 is an example fabric diagram provided by an exemplary embodiment.

[0050] FIG4 is a schematic diagram showing a principle of obtaining a fabric image to be fused, provided by an exemplary embodiment.

[0051] FIG5 is a schematic diagram showing the principle of a fabric fusion method provided by an exemplary embodiment.

[0052] FIG6 is a flow chart of another fabric fusion method provided by an exemplary embodiment.

[0053] FIG7 is a schematic diagram showing the principle of a fabric fusion system provided by an exemplary embodiment.

[0054] FIG8 is a schematic diagram showing a principle of an integrated module provided by an exemplary embodiment.

[0055] FIG9 is a schematic structural diagram of a fabric fusion device provided by an exemplary embodiment.

[0056] FIG10 is a schematic structural diagram of another fabric fusion device provided by an exemplary embodiment.

[0057] FIG11 is a schematic diagram showing a principle of a color fusion process provided by an exemplary embodiment.

[0058] FIG12 is a flow chart of a product color matching method provided by an exemplary embodiment.

[0059] FIG13 is a schematic diagram showing the principle of a method for generating a color reference image provided by an exemplary embodiment.

[0060] FIG14 is a schematic diagram showing the principle of a clothing color matching method provided by an exemplary embodiment.

[0061] FIG15 is a flow chart of another product color matching method provided by an exemplary embodiment.

[0062] FIG16 is a schematic diagram showing the principle of a color matching system provided by an exemplary embodiment.

[0063] FIG17 is a schematic diagram showing the principle of a color preprocessing module provided by an exemplary embodiment.

[0064] FIG18 is a schematic diagram showing the principle of a fusion module provided by an exemplary embodiment.

[0065] FIG19 is a schematic structural diagram of a product color matching device provided by an exemplary embodiment.

[0066] FIG20 is a schematic structural diagram of another product color matching device provided by an exemplary embodiment.

[0067] FIG21 is a schematic structural diagram of a device provided by an exemplary embodiment. DETAILED DESCRIPTION

[0068] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The implementations described in the following exemplary embodiments are not intended to represent all implementations consistent with one or more embodiments of this specification. Rather, they are merely examples of apparatuses and methods consistent with certain aspects of one or more embodiments of this specification, as detailed in the appended claims.

[0069] It should be noted that in other embodiments, the steps of the corresponding method are not necessarily performed in the order shown and described in this specification. In some other embodiments, the method may include more or fewer steps than those described in this specification. In addition, a single step described in this specification may be broken down into multiple steps for description in other embodiments, and multiple steps described in this specification may be combined into a single step for description in other embodiments.

[0070] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of this specification are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0071] In the manufacturing industry, fabrics are required for the manufacture of many products, such as clothing, luggage, home textiles, and so on.

[0072] Taking the apparel industry as an example, fabric garment manufacturing refers to the entire process from raw fabric to finished garment. This sophisticated production process includes multiple key steps, from selecting the appropriate fabric, cutting, sewing, and embellishment, all the way to the final finished product, integrating the fabric's characteristics with the design concept to create a perfect fashion piece.

[0073] The clothing design process requires determining matching fabrics and styles, and determining the desired effect, which requires extensive trial and error. Experimenting with different fabrics for specific styles often requires significant investment in the production of a single garment. Similarly, for other fabric-based products like luggage or home textiles, experimenting with different fabrics for different styles often requires significant investment in the production of the final product.

[0074] This method often requires a lot of manpower and has poor real-time performance.

[0075] In order to solve the above problems, the embodiments of this specification provide a fabric fusion method.

[0076] The method can obtain a fabric image and a product style image, and the device automatically fuses the fabric image and the product style image to obtain an effect image, that is, a simulated effect image of a product manufactured with the fabric in the fabric image and the style in the product style image.

[0077] This method can obtain effect pictures by fusing images through devices, thereby reducing manual participation and improving the efficiency and real-time performance of viewing product effects.

[0078] In addition, when specifically fusing the fabric image and the product style image, the intermediate rendering obtained from the first fusion can be further combined with the image content in the fabric image for a second fusion to obtain the final rendering, so as to improve the fabric performance in the final rendering and reduce the loss of fabric information during the fusion process.

[0079] For ease of understanding, an example is provided below with reference to the accompanying drawings.

[0080] As shown in FIG1 , FIG1 is a schematic diagram of a fabric fusion process provided by an exemplary embodiment.

[0081] Taking clothing as an example, the fabric map and the clothing style map can be integrated to obtain a clothing rendering, that is, a simulated rendering of a clothing product manufactured with the fabric in the fabric map and the style in the clothing style map.

[0082] For ease of understanding, a fabric fusion method is explained below with reference to the accompanying drawings.

[0083] As shown in FIG2 , FIG2 is a flow chart of a fabric fusion method provided by an exemplary embodiment.

[0084] The embodiments of this specification do not limit the execution subject of this method.

[0085] Optionally, the method can be applied to a computing device or a module in the device. For example, the method can be applied to a server, a terminal, or other devices.

[0086] The method may include the following steps.

[0087] S101: Obtain standard fabric drawings and product style drawings.

[0088] S102: The product style image and the standard fabric image are integrated to obtain an intermediate rendering.

[0089] S103: Determine the image content in the standard fabric image that is used for fusion with the product style image as the fabric image to be fused.

[0090] S104: Fusing the intermediate rendering and the fabric image to be fused to obtain a final rendering.

[0091] Among them, the loss of fabric information in the final rendering can be smaller than the loss of fabric information in the intermediate rendering.

[0092] The above method process can obtain the effect map by automatically fusing images, thereby reducing manual participation and improving the efficiency and real-time performance of viewing product effects.

[0093] In addition, the above method process can also improve the fabric performance in the final rendering by fusing the fabric image to be fused with the intermediate rendering, thereby reducing the loss of fabric information during the fusion process.

[0094] In this method, the order of the steps is not limited, and the order of execution of S102 and S103 is not limited. Optionally, S102 and S103 can be executed in parallel, in sequence, or in sequence.

[0095] S103 can be executed after S102, specifically by determining the fabric image to be fused based on the actual fusion process in S102. S102 can also be executed after S103, specifically by first determining the fabric image to be fused and then performing the actual fusion based on the fabric image to be fused. S102 and S103 can also be executed in parallel, with S102 directly merging the product style image and the standard fabric image, while S103 directly determines the fabric image to be fused.

[0096] Furthermore, this method is not limited to the specific products in the product style diagram. Optionally, the product style diagram may specifically include clothing, luggage, home textiles, and other products, including products made of fabrics. Of course, for products that are not made of fabrics, you can also view the effects of using fabrics to make products by integrating fabrics. For example, for paper products, you can view the effects of products made of fabrics, such as desk calendars, portraits, notebooks, and so on.

[0097] This method flow explains the steps for fusing a standard fabric image with a product style image. It is understood that when fusing multiple standard fabric images into a single product style image, different standard fabric images can also be fused into different regions of the single product style image. For specific steps, refer to this method flow. When fusing a single standard fabric image with multiple product style images, this method flow can be executed separately for each product style image.

[0098] The following is a detailed explanation of each step.

[0099] 1. S101: Obtain standard fabric drawings and product style drawings.

[0100] 1. About the content of standard fabric drawings.

[0101] The present method does not limit the content of the standard fabric image. It only needs to include fabric. For example, the standard fabric image can be an image of clothing that includes the fabric of the clothing itself, or an image of luggage that includes the fabric used in the luggage. In a more specific example, the standard fabric image can be an image of an actual product or actual fabric. The standard fabric image can also include a pattern, such as a pattern on the fabric.

[0102] Optionally, the standard fabric image may include image content of the fabric, which may specifically reflect at least one of the following: fabric material, fabric pattern, fabric texture, fabric pattern, fabric color, etc.

[0103] For ease of understanding, the embodiments of this specification provide examples of various fabric diagrams. As shown in Figure 3, Figure 3 is an example fabric diagram provided by an exemplary embodiment. It includes four example fabric diagrams, namely Example Fabric Diagrams 1-4.

[0104] To facilitate subsequent fusion, the standard fabric image can be made to contain only fabrics, allowing for more efficient fusion using the fabric information within the standard fabric image. The lighting within the standard fabric image can also be made uniform, minimizing the impact of the lighting on the fabric information. The fabrics within the standard fabric image can also be laid flat, minimizing the impact of wrinkles or deformation on the fabric information.

[0105] Therefore, optionally, the standard fabric image may satisfy at least one of the following conditions: only containing fabrics, uniform lighting, and fabrics being laid flat.

[0106] 2. How to obtain standard fabric patterns.

[0107] This method does not limit the specific method or source for obtaining the standard fabric image. Alternatively, the standard fabric image can be directly obtained, or other fabric images can be obtained and processed to obtain the standard fabric image. Alternatively, the standard fabric image can be obtained by user input or specified, or by random acquisition.

[0108] In an optional embodiment, the fabric image input by the user can be directly obtained as the standard fabric image, or the fabric image selected by the user can be obtained as the standard fabric image.

[0109] In an optional embodiment, obtaining a standard fabric image may specifically include: obtaining an initial fabric image; and preprocessing the initial fabric image to obtain a standard fabric image. Optionally, the obtained standard fabric image may satisfy at least one of the following conditions: containing only fabrics, having uniform lighting, and having fabrics laid flat.

[0110] This embodiment does not limit the specific content and source of the initial fabric image. For example, the initial fabric image can be a fabric image obtained by actually photographing the fabric, or a fabric image rendered by software simulation, etc. Subsequently, through preprocessing, all initial fabric images, regardless of content and source, can be standardized to obtain a standard fabric image, facilitating subsequent processing.

[0111] This embodiment does not limit the specific method of preprocessing. For example, preprocessing may include affine transformation, thereby obtaining a tiled fabric image from an initial fabric image; preprocessing may also include light processing, which may adjust uneven light in the initial fabric image to uniform light; preprocessing may also include size expansion, which may expand the fabric portion of the initial fabric image into a fabric image containing only the fabric; preprocessing may also include replication and combination, specifically, the fabric portion of the initial fabric image to be replicated may be copied to obtain multiple copies, and further combined into a fabric image containing only the fabric.

[0112] In a specific example, there are often repeated pattern parts in the fabric pattern. Therefore, the fabric parts that need to be repeated can be determined for the initial fabric pattern, and then a standard fabric pattern can be obtained by further copying and combining.

[0113] The above preprocessing methods can be combined with each other. Therefore, optionally, the preprocessing can include at least one of the following: affine transformation, light uniformity processing, size expansion, copy combination, etc.

[0114] The above embodiment can obtain a standard fabric image through preprocessing, thereby improving the flexibility of fabric image acquisition and facilitating the processing of various types of fabric images. In particular, for actual fabric images or fabric images simulated by software, standard fabric images can be obtained through preprocessing to achieve fabric fusion and view the fabric clothing effect.

[0115] To facilitate understanding, designers, users, manufacturers and other users can input different standard fabric images to meet their different needs. After processing, a standard fabric image can be obtained, or it can be directly used as a standard fabric image for subsequent fabric integration, conveniently meeting the needs of different users.

[0116] Of course, in the process of obtaining the standard fabric pattern, processing can also be performed according to specified parameters.

[0117] Optionally, to unify the resolution or size of the standard fabric images, a parameter of the resolution or size may be specified so that a standard fabric image of the specified resolution or size is obtained. Optionally, to facilitate determination of the repeated fabric portion, the fabric portion to be repeated may be adjusted by a user.

[0118] Therefore, optionally, the above method flow may further include: obtaining target parameters; obtaining a standard fabric map, specifically: obtaining an initial fabric map; determining the repeated content of the fabric map in the initial fabric map based on the target parameters; and splicing several copies of the repeated content of the fabric map to obtain a standard fabric map.

[0119] This embodiment does not limit the method and source for obtaining the target parameters. Optionally, the target parameters can be input by the user or pre-stored. This embodiment also does not limit the content of the target parameters. Optionally, the target parameters can specifically include a rectangular box input by the user to determine the repeated content of the fabric pattern; the target parameters can also include a pre-stored rectangular box to determine the repeated content of the fabric pattern. This embodiment also does not limit the specific splicing method. Optionally, splicing can be performed according to the size or form of a standard fabric pattern. The splicing method can also be included in the target parameters.

[0120] Among them, based on the target parameters input by the user, the fabric map can be easily adjusted according to the user's needs, and then the fabric fusion effect can be adjusted to meet the user's actual needs.

[0121] In an optional example, the target parameters may also include the size of the standard fabric pattern to be generated, so that according to the size information in the target parameters, several copies of the repeated content of the fabric pattern may be spliced ​​together to obtain a standard fabric pattern that meets the size information.

[0122] In another optional example, the target parameters may further include adjusting the size information of the initial fabric picture, so that the size information of the initial fabric picture may be adjusted according to the size information in the target parameters, and the repeated content of the fabric picture may be further determined to generate a standard fabric picture.

[0123] For easier understanding, this specification also provides a specific example.

[0124] For example, users can choose to input fabric images in a variety of forms, such as: on-site real-life fabric images, 3D modeling simulation renderings, etc. If you choose to upload a 2D fabric image, it will first pass through the affine transformation processing module to adjust the 2D fabric image to the required tiled image. Then, through the light preprocessing module, the uneven lighting conditions that may exist in the real-life image are converted into uniform lighting for the next step of processing. The key to this process is to make the real-life fabric image suitable for digital processing and ensure the consistency of light and material. Through the affine transformation module, the fabric image can be normalized to adapt to subsequent simulations. The role of the light preprocessing module is to eliminate light unevenness in order to more accurately simulate material effects and pattern textures. After processing the angle and light, the 2D fabric image or the user-input fabric rendering is sent to the fabric preprocessing module, and the fabric preprocessing image is output.

[0125] 3. About the content of product style pictures.

[0126] This method does not limit the content of the product style diagram, as long as the product style diagram contains product styles.

[0127] For example, the product may be clothing, and the clothing style image may be an image of a model wearing clothing, or may only include clothing, or may be an image of clothing obtained through software modeling and simulation.

[0128] For another example, the product may be luggage or home textiles. The product style image may include an image of the luggage or home textiles, specifically an image with a model or background, or an image containing only the product itself, that is, an image containing only luggage or home textiles.

[0129] Optionally, the product style image can include products, which are subsequently used for fabric fusion. To facilitate subsequent fabric fusion, the product style image can be limited to only products. Specifically, this can be achieved by extracting the product image content from the image through image segmentation or image recognition to serve as the product style image.

[0130] For example, if an image contains a person carrying a bag, you can extract the bag image content as the product style image. Or if an image contains a model trying on clothes, you can extract the clothing image content as the product style image.

[0131] Taking clothing as an example, a clothing style graph can optionally include clothing items, specifically clothing items that will be used for subsequent fabric fusion. To facilitate subsequent fabric fusion, the clothing style graph can be restricted to clothing items. Specifically, this can be achieved by extracting clothing content from an image through image segmentation or image recognition, thereby creating a clothing style graph.

[0132] Therefore, optionally, the product style diagram can only contain the product, so as to facilitate subsequent fabric fusion.

[0133] To facilitate understanding, designers, users, manufacturers, and other users can input product images of different standards to meet their different needs. After processing, product style images can be obtained, or directly used as product style images for subsequent fabric integration, conveniently meeting the needs of different users.

[0134] Furthermore, it's important to note that fabric fusion is performed on the product. After obtaining the final rendering, the product in the final rendering can be copied back into the original image. This means that the original image containing the product (the image used to extract the product image content) can be replaced with the fabric-fused product rendering, improving the product display.

[0135] For example, for an image of a model trying on clothes, you can first extract the clothing part as a clothing style image. After the fabrics are fused to obtain the final clothing effect image of the new fabric, the clothing using the new fabric in the final clothing effect image can be copied again to the original image of the model trying on clothes, so that you can replace the clothes worn by the model in the image, making it easier to check the fitting effect.

[0136] In an optional embodiment, a product style diagram may include identification information for different product regions. For example, taking a top as an example, the product style diagram may include identification information for different regions of the top, specifically including identification information for the collar, cuffs, chest, hem, and body, thereby making it easier to identify the style of the product.

[0137] Taking luggage products as an example, the luggage style image can include identification information of different parts of the luggage, specifically including the sides, bottom, interior, top, etc., so as to facilitate the clarification of the style information of the product itself.

[0138] Taking home textile products as an example, the home textile style map can include identification information of different parts of the home textile, specifically including bedding, sheets, pillows and other areas, so as to facilitate the clarification of the style information of the product itself.

[0139] Based on the identification information of different parts of the product in the product style diagram, on the one hand, it is convenient to determine which parts of the area need to be fused with fabrics according to preset instructions, that is, multiple standard fabric diagrams can be merged into different areas of a product style diagram; on the other hand, the product style can be determined in combination with the identification information, and the product style information can be introduced to facilitate improving the fabric fusion effect.

[0140] 4. How to obtain product style pictures.

[0141] This method does not limit the method or source for obtaining product style images. Alternatively, product style images can be directly obtained, or other images containing products can be obtained and processed to obtain product style images. Alternatively, product style images can be obtained from user input or specified, or randomly obtained.

[0142] In an optional embodiment, an image containing a product input by a user may be directly obtained as a product style image, or an image containing a product selected by a user may be obtained as a product style image.

[0143] In an optional embodiment, a method for obtaining a product style image may include: obtaining an initial product image; performing image recognition on the initial product image to determine the product image content in the initial product image; and determining the product style image based on the product image content recognized in the initial product image.

[0144] This embodiment does not limit the specific image recognition method. Specifically, image recognition can be performed based on a deep learning model or neural network for identifying products.

[0145] This embodiment does not limit the content and source of the initial product image, as long as the initial product image contains the product. For example, taking clothing as an example, the initial product image can be an initial clothing image, which can be an image of a model trying on the clothing, an image of the clothing itself, or a photographed image of the clothing. Alternatively, taking home textiles as an example, the initial product image can be an initial home textile image, specifically an image of bed sheets and quilts, or a photographed image of other bedding.

[0146] This embodiment also does not limit the specific form of the product image content. Optionally, the product image content can be the image content occupied by the product in the initial product image.

[0147] For ease of understanding, the product image content may specifically be product detection performed on the initial product image to determine a product detection frame in the initial product image, and the image content in the product detection frame is determined as the product image content.

[0148] For example, taking clothing products as an example, clothing detection can be performed on the initial clothing image to determine a clothing detection frame in the initial clothing image, and the image content in the clothing detection frame is determined as the clothing image content.

[0149] This embodiment does not limit the specific method of determining the product style image based on the product image content. Optionally, the product image content can be directly determined as the product style image, or the product image content can be processed to obtain the product style image.

[0150] A specific processing method, for example, is contour segmentation, which can specifically determine the product contour in the product image content and obtain the image content in the product contour as a product style diagram to improve the accuracy of the product style. For another example, a more detailed identification and segmentation can be performed on the product image content to determine the product style, that is, for each part in the product image content, it is necessary to identify which part of the product it is. Taking clothing as an example, for the image content of a top, the collar image content, sleeve image content, main body image content, etc. in the top image content can be further identified, or the collar area, sleeve area, main body area, etc. can be identified. The content of these identified areas can be used as information and determined as a product style diagram together with the product image content.

[0151] Therefore, optionally, the product style diagram is determined based on the product image content identified in the initial product image. Specifically, it can be that regional recognition is performed on the product image content identified in the initial product image to determine the identification information of different parts of the product area; and the product style diagram is determined based on the determined identification information and product image content.

[0152] This embodiment does not limit the specific method of determining the product style map based on the determined identification information and product image content. Optionally, the determined identification information and product image content can be directly determined as the product style map, or the regional content with identification information can be combined to determine the product style map.

[0153] Furthermore, it is understood that when performing image recognition on the initial product image, different product regions in the initial product image may also be directly identified and determined. The determined product image content may include identification information of different product regions.

[0154] Optionally, based on the region identification information in the product image content, the region where the fabrics need to be fused can be further determined. This embodiment does not limit the specific determination method, and the region where the fabrics need to be fused can be determined based on user designation or pre-stored information.

[0155] For example, you can fuse fabrics only for the cuff area of ​​a top garment based on user specifications, or you can fuse fabrics for the entire area of ​​a trouser garment based on pre-stored settings.

[0156] Of course, the specific determination step can be performed in the subsequent fusion step. Please refer to the following explanation for details.

[0157] For ease of understanding, this specification also provides a specific example. Taking clothing products as an example, users can choose to input various style images or model images. Among them, style images include multiple categories, such as clothing categories (such as windbreakers, sweatshirts, dresses, etc.), or other categories such as luggage, home textiles, etc.

[0158] The input for a style image can be a pure style image or a model wearing the style image. Considering that users may need to see how the style will look in a real-life scenario, the user can directly input a model wearing the style image, which effectively reduces the cost investment in the actual production stage.

[0159] The input style image or model image will then be fed into the parsing model for style parsing, resulting in the required style parsing graphs. These style parsing graphs will then be fed into the integration module along with the style image for subsequent processing.

[0160] The input style image or model image will be parsed by the parsing model. The function of the parsing model is to parse and partition the various areas of the style image. For example, the collar, sleeves, cuffs, pockets and other parts of the clothing will be accurately parsed into separate areas, providing detailed information for the subsequent fabric fusion of different areas.

[0161] In addition to identifying individual regions, the parsing model also helps the network model understand their composition, providing a better foundation for subsequent fabric fusion. Understanding and parsing style images at a higher level of detail improves the system's understanding of style structure, enabling more customized fabric fusion.

[0162] 2. S102: Integrate the product style diagram and the standard fabric diagram to obtain an intermediate rendering.

[0163] The present method does not limit the specific fusion method. Optionally, the intermediate rendering can be a simulation rendering of a product manufactured using the fabric in the standard fabric drawing and the style in the product style drawing.

[0164] Optionally, you can use layer operations to merge the product style image with the standard fabric image to obtain an intermediate rendering.

[0165] Optionally, a deep learning approach can be used to input the standard fabric image and the product style image into a pre-trained model for fusion to obtain an intermediate rendering.

[0166] This embodiment does not limit the specific model structure or training method. It is understandable that fabric fusion can be performed by training a neural network model or a generative model, specifically a generative model that generates images from images.

[0167] In an optional embodiment, the product style image and the standard fabric image may be directly fused, or the fabric image to be fused may be first determined based on S103, and then the product style image and the fabric image to be fused may be fused to obtain an intermediate effect image.

[0168] It can be understood that, optionally, the intermediate rendering can be a product simulation rendering based on the product style in the product style diagram and combined with the fabric in the standard fabric diagram, and the product style information in the intermediate rendering can be the same as the product style information in the product style diagram, and only the fabric used is replaced.

[0169] Optionally, the intermediate rendering can retain the product style information from the product style image. This embodiment does not limit the specific product style information. Product style information includes, for example, the product shape, pattern, wrinkles, and outline. The texture of a specific product can be combined or replaced with the fabric in the standard fabric image, resulting in an intermediate rendering that retains the product style information but incorporates the new fabric from the standard fabric image.

[0170] Of course, optionally, when the product style diagram contains product style information, the product styles under different viewing angles can be simulated, and the fabrics in the standard fabric diagram can be further integrated to obtain an intermediate rendering.

[0171] In an optional embodiment, the intermediate rendering obtained in S102 is a preliminary fusion effect, which can be further fused in S104. Therefore, the fusion process in S102 can prioritize preserving the product style information or other product information in the product style image, such as product wrinkles, product lighting, product shape, product patterns, etc.

[0172] Optionally, during the specific fusion, the fabric can be fused in accordance with the product's own wrinkles, or it can be fused in accordance with the product's own light information or illumination information, thereby improving the fabric fusion effect of the intermediate rendering.

[0173] In an optional embodiment, a pre-trained model may be used to fuse product style images and standard fabric images to obtain an intermediate rendering.

[0174] This embodiment does not limit the specific model structure and training method. Alternatively, a generative model or a neural network model can be used. For ease of description, the model used to fuse the standard fabric image and the product style image to obtain the intermediate rendering is referred to as the first fusion model.

[0175] For ease of understanding, this embodiment can provide an embodiment of a training method. Optionally, images can be taken of different products made of the same product style using different fabrics, and fabric images can be extracted as training samples for training the above-mentioned first fusion model. Of course, it is also possible not to manufacture actual products, but to simulate the same product style in simulation software (for example, product style design software) by replacing multiple different fabric effects and obtaining images by screenshots or photography.

[0176] In a specific example, for the same style of clothing, garments made of cotton and leather fabrics can be simulated, respectively, referred to as cotton clothing and leather clothing. Furthermore, images of the cotton clothing and leather clothing can be captured. Specifically, these can be different images taken from the same perspective with the same style information. Alternatively, they can be different images taken from different perspectives. Because the clothing is simulated by software, the specific style information can be consistent across different images at the same perspective, including the same folds, shape, pattern, and lighting information.

[0177] Furthermore, a standard fabric image of cotton fabric and a standard fabric image of leather fabric can be obtained, which can be directly captured from the cotton clothing image and the leather clothing image, or obtained by additional photography.

[0178] Afterwards, you can use the standard fabric image of cotton fabric and the leather garment image as sample features, and the cotton garment image as the sample label. You can also use the standard fabric image of leather fabric and the cotton garment image as sample features, and the leather garment image as the sample label. This way, you can obtain two training samples.

[0179] By analogy, we can combine any two different fabrics to obtain two training samples.

[0180] The obtained training samples can be used to train the first fusion model.

[0181] Therefore, optionally, the method for constructing a training sample set for the first fusion model may include: obtaining different product images of the same product style using different fabrics, and obtaining different standard fabric images of the different fabrics used; using the product images using the first fabric and the standard fabric images of the second fabric as sample features, and using the product images using the second fabric as sample labels, constructing training samples, and adding them to the training sample set. The first fabric and the second fabric are different fabrics. Of course, the first fabric can specifically be any fabric, and the second fabric can be any fabric other than the first fabric, which is merely for illustrative purposes. Optionally, the method may also use product images using any fabric and the standard fabric images of any other fabric as sample features, and use the product images of the other fabric as sample labels, to construct training samples, and add them to the training sample set.

[0182] The constructed training sample set can be used to train the first fusion model. This embodiment can improve the training effect of the first fusion model and improve the fabric fusion effect of the intermediate rendering.

[0183] In an optional embodiment, in addition to the standard fabric image and product style image that need to be fused, other information can also be fused during the specific fusion process. This embodiment does not limit the specific content of the other information. For example, the other information can be fabric information in the standard fabric image, such as material, pattern, texture, etc. It can also be other information used to guide the fusion, such as regional information on the product style image where the fabric needs to be fused, specifying the region in the product style image where the fabric needs to be fused; for example, fusion method information, which can specify whether to use deep learning or layer operation for fusion; for example, information on the fabric image to be fused, which is used to determine the fabric image to be fused in the standard fabric image.

[0184] This embodiment does not limit the specific form of other information, which may be text, image, vector, etc.

[0185] Optionally, the above method may further include obtaining target fabric information from a standard fabric image. Fusion of the product style image with the standard fabric image to obtain an intermediate rendering may specifically include fusing the obtained target fabric information, the product style image, and the standard fabric image to obtain the intermediate rendering. This embodiment can enhance the fabric fusion effect by adding fabric information from the standard fabric image.

[0186] This embodiment does not limit the specific content and format of the target fabric information. Optionally, the target fabric information may include at least one of the following: fabric material, fabric pattern, fabric texture, fabric drape, etc. The specific format may be text, image, or vector. For example, a textual description of the fabric material "cotton" may be used as the target fabric information for fusion.

[0187] This embodiment does not limit the method for obtaining the target fabric information, which may be obtained from a standard fabric image, extracted based on a standard fabric image, or obtained by obtaining pre-stored target fabric information.

[0188] Furthermore, the above method may optionally further include: obtaining information about an area in the product style image where fabrics need to be integrated. Fusion of the product style image with the standard fabric image to obtain an intermediate rendering may specifically include: fusing the standard fabric image with respect to the area in the product style image where fabrics need to be integrated to obtain the intermediate rendering.

[0189] This embodiment does not limit the form of the region information, and specifically it can be in image form or text form.

[0190] This embodiment does not limit the source of obtaining the region information. Optionally, it can be obtained from data input by the user or from pre-stored data.

[0191] In an optional embodiment, based on the product style map and the standard fabric map, it can be achieved through more than one fusion.

[0192] Therefore, optionally, the fusion between the product style image and the standard fabric image can be specifically one or more fusions to obtain an intermediate effect image.

[0193] The present method is not limited to a specific fusion process, and may be implemented using a cyclic process.

[0194] Optionally, the product style image and the standard fabric image are combined to obtain an intermediate rendering. Specifically, the following steps may be looped until a preset loop stop condition is met: the current product style image and the standard fabric image are combined to obtain a new rendering; and the obtained new rendering is determined as the current product style image. After the loop stops, the current product style image is determined as the intermediate rendering.

[0195] It is understood that the fusion method for each cycle can be explained above. For example, a preset model can be input for fusion. Specifically, it can be a generative model, such as a diffusion model.

[0196] This embodiment does not limit the preset loop stop condition. Specifically, it can be set to the maximum number of loops, or to whether the difference between the new rendering and the current product style is small enough, or it can be set based on the display of computing power resources.

[0197] Optionally, the preset loop stop condition may include at least one of the following: 1) the number of loops is greater than a preset number of loops; 2) the difference between the new rendering and the current product style image is less than a preset difference, etc. The specific method of quantifying the difference between the images is not limited in this embodiment.

[0198] 3. S103: Determine the image content in the standard fabric image that is used for fusion with the product style image as the fabric image to be fused.

[0199] This embodiment does not limit the specific method for determining the fabric image to be fused. The image content to be fused with the product style image can be determined based on the fusion method. If the fusion method utilizes a model, the fabric image to be fused can be determined based on the model input. If the fusion method utilizes layer operations, the fabric image to be fused can be determined based on the layer operation requirements.

[0200] In an optional embodiment, since the standard fabric image may contain image content that is not used for fusion with the product style image, the image content that is used for fusion with the product style image can be determined as the fabric image to be fused. At least in the subsequent step S104, it can be fused again to reduce the loss of fabric information and the impact of fabric information that is not used for fusion.

[0201] In a specific example, there may be two different patterns in the standard fabric image, and in the fusion process of S102, only one of the patterns is used for fusion, and the other pattern should not be used for re-fusion in S104, thereby improving the overall fusion effect and the fabric fusion effect of the final rendering.

[0202] In another specific example, there may be multiple different colors in the standard fabric image, and if one of the colors is not used for fusion in the fusion process of S102, then this color should not be used for re-fusion in S104, thereby improving the overall fusion effect and the fabric fusion effect of the final rendering.

[0203] In addition, S103 may be optionally executed before S102, and S102 may specifically be to fuse the product style image and the fabric image to be fused to obtain an intermediate rendering. Therefore, S103 may be to determine the image content to be subsequently fused with the product style image.

[0204] Specifically, the standard fabric image can be directly determined as the fabric image to be merged, that is, the entire standard fabric image is merged with the product style image. Alternatively, part of the standard fabric image can be extracted according to user needs and used as the fabric image to be merged.

[0205] Therefore, optionally, a cutting method can be used to cut out the fabric image content corresponding to the product outline from the standard fabric image, which is used as the fabric image to be fused. Specifically, the product outline can be first identified and then superimposed on the standard fabric image. This allows the image content contained in the standard fabric image of the product outline to be conveniently determined as the fabric image to be fused.

[0206] Of course, when overlapping, the fabric pattern to be fused can be determined by adjusting the position and size of the overlapping product outlines.

[0207] Optionally, a portion of the image content in the standard fabric image to be fused may be determined based on the intermediate rendering to serve as the fabric image to be fused. Specifically, the fused portion selected during the fusion process in S102 may be determined as the fabric image to be fused. Alternatively, the fabric image to be fused may be determined based on the fabric image content used in the intermediate rendering.

[0208] In summary, optionally, the image content in the standard fabric image used for fusion with the product style image is determined as the fabric image to be fused, which may specifically be any of the following.

[0209] 1) Based on the intermediate rendering, the image content in the standard fabric image used for fusion with the product style image is determined as the fabric image to be fused.

[0210] 2) The product outline in the product style image is overlapped with the standard fabric image in a preset manner, and the image content contained in the product outline in the standard fabric image is determined as the fabric image to be fused.

[0211] This embodiment can facilitate fabric fusion in subsequent steps by determining the fabric image to be fused, thereby improving the fabric fusion effect and efficiency, and reducing the impact of fabric information that is not used for fusion.

[0212] This embodiment does not limit the preset method of overlap. Optionally, the overlap position of the product outline can be directly set in the standard fabric image, so that the image content contained in the product outline after overlap can be determined, and the fabric image to be fused can be determined.

[0213] This embodiment also does not limit the source of the preset method. Optionally, the overlap can be performed according to a preset method input or specified by the user, or according to a pre-stored preset method.

[0214] Therefore, the above method flow may optionally further include: obtaining target parameters. Determining image content in the standard fabric image to be fused with the product style image as the fabric image to be fused. Specifically, the method may include: overlaying the product outline in the product style image with the standard fabric image according to the relative position in the target parameters, and determining the image content contained in the product outline in the standard fabric image as the fabric image to be fused.

[0215] The relative position in this embodiment can specifically be a pre-set overlap position between the product outline and the standard fabric image, thereby enabling alignment. This embodiment does not limit the specific form or source of the relative position; alternatively, the relative position can include four coordinate points in the standard fabric image. The target parameters to which the relative position belongs can be user-entered or pre-stored.

[0216] Of course, the fabric image to be fused may be specifically determined by extracting image content that matches the product outline from the standard fabric image without overlapping.

[0217] For ease of understanding, the embodiments of this specification also provide a specific illustration example, as shown in FIG4 , which is a schematic diagram of the principle of obtaining a fabric image to be fused provided by an exemplary embodiment.

[0218] Taking clothing as an example, the clothing outline can be obtained from the clothing style image, and the image content in the shape of the clothing outline can be further extracted from the standard fabric image to serve as the fabric image to be fused. The fabric image to be fused can then be used to perform step S102 for fusion, or the fabric image to be fused can be used to perform step S104 for fusion.

[0219] In addition, regarding the target parameters in the present method flow, the present method flow is not limited to specific content and source.

[0220] Optionally, the target parameters may be input or specified by the user, and specifically, the fabric fusion may be performed in real time according to the parameters input or specified by the user. The target parameters may also be pre-stored.

[0221] Of course, the target parameters may include at least one of the following: repeated content of fabric images, resizing of fabric images, size and relative position of standard fabric images, information about the area of ​​product style images where fabrics need to be merged, etc. For detailed explanations, please refer to the above.

[0222] It will be appreciated that target parameters may be used to assist in executing each step of the method flow.

[0223] 4. S104: Fusing the intermediate rendering and the fabric image to be fused to obtain a final rendering.

[0224] Among them, the loss of fabric information in the final rendering can be smaller than the loss of fabric information in the intermediate rendering.

[0225] The present method does not limit the specific method of fusion. For details, please refer to the detailed explanation of S102. Optionally, the final rendering can be a simulated rendering of a product manufactured according to the style in the product style diagram using the fabric in the standard fabric diagram or the fabric in the fabric diagram to be fused.

[0226] Since the fabric image to be fused is fused again based on the intermediate rendering, the fabric fusion effect in the final rendering can be improved and fabric loss can be reduced.

[0227] Optionally, layer operations can be used, or deep learning methods can be used.

[0228] The fusion method of S104 may be the same as or different from the fusion method of S102, and the present method flow is not limited thereto.

[0229] Optionally, the intermediate rendering and the fabric image to be fused may be fused by layer operation to obtain the final rendering.

[0230] Optionally, a deep learning approach may be used to input the intermediate rendering and the fabric image to be fused into a pre-trained model for fusion to obtain the final rendering.

[0231] This embodiment does not limit the specific model structure or training method. It is understandable that fabric fusion can be performed by training a neural network model or a generative model, specifically a generative model that generates images from images.

[0232] It can be understood that, optionally, the final rendering can be a product simulation rendering based on the product style in the product style diagram, combined with the fabric in the fabric diagram to be fused, and the product style information in the final rendering can be the same as the product style information in the product style diagram, and remain consistent, with only the fabric used being replaced.

[0233] Optionally, the final rendering can retain the product style information in the product style image. This embodiment is not limited to specific product style information. Product style information includes, for example, the product shape, pattern, wrinkles, and outline. The texture of a specific product can be combined or replaced with the fabric in the fabric image to be fused, resulting in a final rendering that retains the product style information but incorporates the new fabric in the fabric image to be fused.

[0234] Of course, optionally, when the product style image contains product style information, the product styles under different viewing angles can be simulated, and the fabrics in the fabric image to be fused can be further fused to obtain the final effect image.

[0235] It should be emphasized that the fusion in S104 is mainly based on reducing the loss of fabric information in the intermediate renderings, that is, it is hoped to further improve the effect of fabric fusion. Therefore, the product style information in the final renderings also needs to be retained, mainly to optimize and improve the effect of fabric fusion.

[0236] In an optional embodiment, the intermediate effect image and the fabric image to be fused may be fused using a pre-trained model to obtain a final effect image.

[0237] Optionally, the intermediate rendering and the fabric image to be fused are fused to obtain a final rendering. Specifically, the intermediate rendering and the fabric image to be fused are input into a preset model to obtain a final rendering output by the preset model; the preset model can be used to reduce the loss of fabric information.

[0238] This embodiment does not limit the specific preset model structure and training method. Optionally, a generative model or a neural network model can be used. For ease of description, the model used to fuse the intermediate rendering and the fabric image to be fused to obtain the final rendering is referred to as the second fusion model.

[0239] The training method of the second fusion model can refer to the training method of the first fusion model. The second fusion model can be the same as the first fusion model or different from the first fusion model.

[0240] For ease of understanding, this embodiment can provide an embodiment of a training method. Optionally, images of different products obtained by using different fabrics to manufacture different product styles can be captured separately, and fabric images can be extracted as training samples for training the above-mentioned second fusion model. Of course, it is also possible not to manufacture actual products, but to simulate different product styles in simulation software (for example, product style design software) by replacing multiple different fabric effects and obtaining images by screenshots or photography.

[0241] Of course, optionally, the input and output of the second fusion model may be different from those of the first fusion model, and the second fusion model may be used to reduce the loss of fabric information, and therefore, different training samples may be used.

[0242] Specifically, the loss of fabric information can be taken into consideration, and the intermediate rendering has already integrated the fabric rendering in the standard fabric image. Therefore, for a product made of a certain fabric, an image can be taken and further blurred to lose some fabric information, and then a training sample can be constructed.

[0243] For example, for a product made of a specified fabric, an original image can be captured and the fabric in the original image can be blurred or noised to lose fabric information, resulting in a processed image. The processed image and the fabric image of the specified fabric can then be used as sample features, and the original image can be used as a sample label to construct a training sample for training. Of course, the fabric image can specifically be the fabric image to be fused, which can specifically be obtained according to the method in S103.

[0244] Therefore, optionally, a method for constructing a training sample set for the second fusion model may include: obtaining an original image of a product using a specified fabric, and obtaining a fabric image to be fused of the specified fabric; processing the original product image to increase the loss of fabric information to obtain a product loss image; constructing training samples using the product loss image and the fabric image to be fused as sample features and the original product image as a sample label, and adding the training samples to the training sample set. The specified fabric can be any fabric.

[0245] The constructed training sample set can be used to train the second fusion model. This embodiment can improve the training effect of the second fusion model and improve the fabric fusion effect of the intermediate rendering.

[0246] This embodiment does not limit the method of obtaining the original product image, which can be obtained by direct photography or through software modeling. This embodiment also does not limit the method of obtaining the fabric image to be fused of the specified fabric, which can be obtained based on the original product image. The specific method can be referred to the explanation in S101.

[0247] This embodiment does not limit the method of processing the original product image, as long as the fabric information can be lost. For example, blurring or noise addition can be used.

[0248] Of course, in addition to the above-mentioned training sample set construction method, other methods can also be used to construct the training sample set, as long as the loss of product image fabric information in the sample features is greater than the loss of product images in the sample labels.

[0249] For example, a product original image can be processed multiple times to obtain multiple images with different fabric information losses. The images are sorted according to the amount of fabric information loss, and then training samples can be obtained by combining them.

[0250] For example, a product original image can be processed once to obtain a first loss image, and then processed again to obtain a second loss image, and so on. A training sample can then be obtained using the first loss image as the sample feature and the product original image as the sample label; a training sample can be obtained using the second loss image as the sample feature and the first loss image as the sample label; a training sample can be obtained using the second loss image as the sample feature and the product original image as the sample label; and so on.

[0251] This method does not limit the training method for the second fusion model. It can be trained based on the training sample set constructed above, or it can be trained using the same method as the first fusion model. It should be noted that the first fusion model can fuse fabrics to improve fabric fusion effects and can also be used to reduce fabric information loss. Therefore, it can serve as the second fusion model to further fuse the intermediate renderings and the fabric image to be fused.

[0252] In an optional embodiment, in the specific fusion process, in addition to the fabric image to be fused and the intermediate effect image, other information may also be fused. For details, please refer to the explanation in S102.

[0253] This embodiment does not limit the specific content of the other information. For example, the other information may be fabric information in the standard fabric image, such as material, pattern, texture, etc. It may also be other information used to guide fusion, such as information about the area in the product style image where the fabric needs to be fused, which specifies the area in the product style image where the fabric needs to be fused; information about the fusion method, which may specify whether the fusion is performed using a deep learning approach or a layer operation approach; and information about the fabric image to be fused, which is used to determine the fabric image to be fused in the standard fabric image.

[0254] This embodiment does not limit the specific form of other information, which may be text, image, vector, etc.

[0255] Optionally, the above method may further include: obtaining target fabric information of the standard fabric image. And fusing the intermediate rendering image and the fabric image to be fused to obtain the final rendering image, which may specifically include: fusing the obtained target fabric information, the intermediate rendering image, and the fabric image to be fused to obtain the final rendering image.

[0256] This embodiment does not limit the specific content and format of the target fabric information. Optionally, the target fabric information may include at least one of the following: fabric material, fabric pattern, fabric texture, fabric drape, etc. The specific format may be text, image, or vector. For example, a textual description of the fabric material "cotton" may be used as the target fabric information for fusion.

[0257] This embodiment does not limit the method for obtaining the target fabric information, which may be obtained from a standard fabric image, extracted based on a standard fabric image, or obtained by obtaining pre-stored target fabric information.

[0258] Furthermore, the above method may optionally further include: obtaining information about a region in the intermediate rendering where fabrics need to be fused. Fusion of the intermediate rendering with the fabric image to be fused to obtain the final rendering may specifically include: fusing the fabric image to be fused with respect to the region in the intermediate rendering where fabrics need to be fused to obtain the final rendering.

[0259] This embodiment does not limit the form of the region information, and specifically it can be in image form or text form.

[0260] This embodiment does not limit the source of obtaining the region information. Optionally, it can be obtained from data input by the user or from pre-stored data.

[0261] In an optional embodiment, based on the intermediate effect image and the fabric image to be fused, the loss of fabric information can be reduced. In order to further reduce the loss of fabric information and improve the fabric fusion effect, more than one fusion can be performed.

[0262] Therefore, optionally, the fusion between the intermediate effect image and the fabric image to be fused may be one or more fusions to obtain the final effect image.

[0263] The present method is not limited to a specific fusion process, and may be implemented using a cyclic process.

[0264] Optionally, the intermediate rendering and the fabric image to be fused are fused to obtain the final rendering. Specifically, the following steps may be looped until a preset loop stop condition is satisfied: fusing the current intermediate rendering and the fabric image to be fused to obtain a new rendering; and determining the obtained new rendering as the current intermediate rendering. After the loop stops, the current intermediate rendering is determined as the final rendering.

[0265] It is understood that the fusion method for each cycle can be explained above. For example, a preset model can be input for fusion. Specifically, it can be a generative model, such as a diffusion model.

[0266] This embodiment does not limit the preset loop stop condition. Specifically, it can be set to the maximum number of loops, or to determine whether the difference between the new rendering and the current intermediate rendering is small enough, or it can be set based on the display of computing resources.

[0267] Optionally, the preset loop stop condition may include at least one of the following: 1) the number of loops is greater than a preset number of loops; 2) the difference between the new rendering and the current intermediate rendering is less than a preset difference; 3) the loss of fabric information in the current intermediate rendering is less than a preset loss, etc. The specific method of quantifying the difference between images and the method of quantifying the loss of fabric information are not limited in this embodiment.

[0268] It should be noted that, precisely because the fabric image to be fused is obtained in S103 for fusion with the product style image to obtain the intermediate rendering, it is equivalent to removing the unfused part of the standard fabric image, so that the intermediate rendering and the fabric image to be fused can be further fused through S104 to improve the fabric fusion effect, reduce the loss of fabric information, and reduce the influence of image content that does not participate in fabric fusion.

[0269] Therefore, in summary, the process of this method can obtain the effect map by automatically fusing images, thereby reducing manual participation and improving the efficiency and real-time performance of viewing product effects.

[0270] In addition, the above method process can also extract the fabric image to be fused, that is, the part of the fabric image used to fuse the product style image to obtain the intermediate rendering, and fuse the fabric image to be fused again with the intermediate rendering. This can reduce the influence of the image content that does not participate in the fabric fusion, improve the fabric performance effect in the final rendering, and reduce the loss of fabric information in the fusion process.

[0271] For easier understanding, the present specification also provides drawings to explain the overall fusion process. As shown in FIG5 , FIG5 is a schematic diagram of the principle of a fabric fusion method provided by an exemplary embodiment.

[0272] Taking clothing as an example, the fabric map and the clothing style map can be integrated to obtain an intermediate clothing rendering, that is, a simulated rendering of a clothing product manufactured with the fabric in the fabric map and the style in the clothing style map.

[0273] After that, the clothing outline in the clothing style image can be obtained, and the image content with the shape of the clothing outline can be further extracted from the standard fabric image as the fabric image to be fused.

[0274] The fabric image to be fused and the intermediate clothing rendering are further fused to obtain the final clothing rendering.

[0275] The loss of fabric information in the final garment rendering can be less than that in the intermediate garment rendering, which can be specifically reflected in the fineness of the fabric. The final garment rendering is superior to the intermediate garment rendering in terms of fabric fineness.

[0276] The embodiments of this specification also provide another embodiment of the fabric fusion method using clothing products as an example.

[0277] As shown in FIG6 , FIG6 is a flow chart of another fabric fusion method provided by an exemplary embodiment.

[0278] The embodiments of this specification do not limit the execution subject of this method.

[0279] Optionally, the method can be applied to a computing device or a module in the device. For example, the method can be applied to a server, a terminal, or other devices.

[0280] The method may include the following steps.

[0281] S201: Obtain standard fabric drawings and clothing style drawings.

[0282] S202: The clothing style diagram and the standard fabric diagram are integrated to obtain an intermediate clothing effect diagram.

[0283] S203: Determine the image content in the standard fabric image that is used for fusion with the clothing style image as the fabric image to be fused.

[0284] S204: Fusing the intermediate garment-making rendering and the fabric image to be fused to obtain a final garment-making rendering.

[0285] Among them, the loss of fabric information in the final clothing rendering can be smaller than the loss of fabric information in the intermediate clothing rendering.

[0286] The above method process can obtain clothing effect pictures by automatically fusing images, thereby reducing manual participation and improving the efficiency and real-time performance of viewing clothing product effects.

[0287] In addition, the above method process can also improve the fabric performance in the final clothing rendering by fusing the fabric image to be fused with the intermediate clothing rendering, thereby reducing the loss of fabric information during the fusion process.

[0288] For a detailed explanation of the process of this method, please refer to the explanation of S101-S104.

[0289] Optionally, the intermediate clothing rendering and the fabric image to be fused are fused to obtain the final clothing rendering. Specifically, the intermediate clothing rendering and the fabric image to be fused are input into a preset model to obtain the final clothing rendering output by the preset model; the preset model can be used to reduce the loss of fabric information; the loss of fabric information in the final clothing rendering can be less than the loss of fabric information in the intermediate clothing rendering.

[0290] Optionally, the above method process may also include: obtaining target fabric information of the standard fabric picture; fusing the intermediate clothing effect picture and the fabric picture to be fused to obtain the final effect picture, which may be specifically: fusing the obtained target fabric information, the intermediate clothing effect picture and the fabric picture to be fused to obtain the final clothing effect picture.

[0291] Optionally, the intermediate clothing rendering and the fabric drawing to be fused are fused to obtain the final clothing rendering, which can be specifically: looping the following steps until a preset loop stop condition is met: fusing the current intermediate clothing rendering and the fabric drawing to be fused to obtain a new clothing rendering; determining the obtained new clothing rendering as the current intermediate clothing rendering; after the loop stops, determining the current intermediate clothing rendering as the final clothing rendering.

[0292] Optionally, the image content in the standard fabric image used for fusion with the clothing style image is determined as the fabric image to be fused. Specifically, it can be: based on the intermediate clothing effect image, the image content in the standard fabric image used for fusion with the clothing style image is determined as the fabric image to be fused; or the clothing outline in the clothing style image is overlapped with the standard fabric image in a preset manner, and the image content contained in the clothing outline in the standard fabric image is determined as the fabric image to be fused.

[0293] Optionally, obtaining a standard fabric map may specifically include: obtaining an initial fabric map; and preprocessing the initial fabric map to obtain a standard fabric map.

[0294] Optionally, the method for acquiring the clothing style diagram may specifically be: acquiring an initial clothing image; performing image recognition on the initial clothing image to determine clothing image content in the initial clothing image; and determining the clothing style diagram based on the clothing image content recognized in the initial clothing image.

[0295] Optionally, the above method flow may further include: obtaining target parameters; obtaining a standard fabric map, specifically: obtaining an initial fabric map; determining repeated content of the fabric map in the initial fabric map based on the target parameters; and splicing several copies of the repeated content of the fabric map to obtain a standard fabric map.

[0296] Optionally, the above method process may also include: obtaining target parameters; determining the image content in the standard fabric image used for fusion with the clothing style image as the fabric image to be fused, specifically: overlapping the product outline in the clothing style image with the standard fabric image according to the relative position in the target parameters, and determining the image content contained in the clothing outline in the standard fabric image as the fabric image to be fused.

[0297] For a detailed explanation of the process of this method, please refer to the explanation of S101-S104.

[0298] The following provides an application embodiment taking clothing products as an example.

[0299] Fabric-to-garment refers to the entire process of garment manufacturing, from raw fabric to finished garment. This meticulous process encompasses multiple key steps, from selecting the right fabric to cutting, sewing, and embellishment, all the way to the final finished product. The process integrates the fabric's characteristics with the design concept to create a flawless fashion piece.

[0300] The production of finished garments from fabrics is a complex process that requires skill and precision to ensure that each garment meets design requirements and is of high quality. This comprehensive process encompasses every aspect of garment manufacturing, starting with the selection and sourcing of fabrics and continuing through to the production and delivery of the finished garments. Each step is crucial to the success of the final product.

[0301] 3D simulation software plays an indispensable role in the fabric and garment design and production process. With its simulation, modeling, and visualization capabilities, these software programs can accurately reproduce how fabrics will behave on garments of various shapes and sizes. This provides designers, manufacturers, and fashion designers with a deeper understanding of the effects of the fabrics they create for their fashion creations.

[0302] With 3D simulation software, designers can instantly transform their creative ideas into visual models, observing the appearance, texture, and flow of fabrics in different garment constructions, allowing them to quickly evaluate and improve their designs. Such tools not only save time but also improve design accuracy and reduce prototyping costs, as many problems can be solved in a virtual environment.

[0303] Fabric fitting is a process that takes garment prototypes made from selected fabrics and aims to simulate how the fabric will look on a model. This process typically occurs in the early stages of garment manufacturing, with the primary purpose of helping designers, manufacturers, and fashion designers better understand how the selected fabrics will perform in actual wear, ensuring they meet design requirements and provide a comfortable wearing experience.

[0304] Fabric fitting is a crucial step in ensuring a smooth garment manufacturing process. It provides the design team with tangible visual and tactile feedback, helping to fine-tune the design, reduce potential issues, and ensure that the final garment perfectly reflects the design vision. This process is an integral part of the fashion industry and helps ensure the highest level of garment quality.

[0305] 3D simulation software can also be used to render the effects of different fabrics on 3D clothing models and digital mannequins, allowing for a preview of how they will look when worn. This virtual try-on experience not only provides consumers with a more convenient shopping experience, but also helps reduce risk and alleviate uncertainty when shopping online. For retailers and brands, this technology not only improves sales conversion rates and reduces returns, but also provides them with more opportunities for customization and personalization.

[0306] Furthermore, this embodiment proposes a method for image processing and fabric fusion that not only preserves the fabric's material and pattern texture, but also enables a one-step transition from fabric images to garment images, requiring only 2D graphics input. Furthermore, this method can transform fabrics into garments and display them on models, thus enabling a comprehensive display of fabrics and garments on the body.

[0307] This approach is unique in that it not only addresses the high costs, slow production speeds, and sustainability challenges of traditional fabric garment manufacturing and fabric upholstering, but also overcomes the technical challenges of digital 3D simulation, such as the high modeling threshold. It provides a simple way for anyone to easily get started with fabric garment manufacturing and fabric upholstering, without having to master complex technical skills. This approach allows for zero-to-no application, freeing fabric sales and manufacturers from traditional sales models.

[0308] This method ensures that the fabric's texture and pattern are reproduced to the greatest extent possible within the garment, preserving the fabric's natural fit and its coordination with the style. Furthermore, this technology offers flexible input conditions, meeting diverse needs across different links. It can accept inputs such as a single fabric image, as well as multiple inputs such as fabric and garment style images, and images of model clothing, to meet the diverse needs of different manufacturers and consumers. Furthermore, the technology's wide range of applications makes it widely applicable, benefiting everyone from clothing manufacturers to luggage manufacturers, fabric producers, and end consumers. It provides opportunities for low-cost, low-threshold, high-efficiency, and sustainable development across all links, achieving coordinated development within the industrial chain.

[0309] This method proposes a fabric fitting solution based on a generative network that preserves fabric material and pattern texture. This solution consists of four modules: a fabric preprocessing module, a style parsing module, an integration module, and a fabric fitting generation module. The fabric preprocessing module accepts multiple inputs, such as fabric images, fabric materials, and the location and size of fabric pattern textures. The clothing parsing module also accepts two inputs: style images and model images.

[0310] As shown in FIG. 7 , FIG. 7 is a schematic diagram showing the principle of a fabric fusion system provided by an exemplary embodiment.

[0311] This includes a fabric pre-processing module, a style parsing module, an integration module, and a fabric fitting generation module. The detailed process is explained below.

[0312] Step 1: Fabric pretreatment module.

[0313] Users can select from a variety of fabric image input formats, including live photos, 3D modeling simulation renderings, and more. If you choose to upload a 2D fabric image, it will first be processed through the Affine Transformation Module to adjust the 2D fabric image to the desired tiled image. The Light Preprocessing Module then transforms any uneven lighting conditions in the live photo into uniform lighting for further processing.

[0314] The key to this process is making the real-life fabric image suitable for digital processing, ensuring consistency in lighting and material. Through the affine transformation module, the fabric image is normalized for subsequent simulation. The lighting preprocessing module eliminates lighting unevenness to more accurately simulate material effects and pattern textures.

[0315] After angle and light processing, the 2D fabric image or the fabric rendering image input by the user is sent to the fabric preprocessing module, and the fabric preprocessing image is output.

[0316] Users can also choose to input regulator parameters and fabric material information, such as cotton, velvet, or leather. This information is used in the integration module and fabric fitting generation module to more precisely control the fabric's appearance in the final product. This includes the size and position of the pattern, as well as the texture and light and dark variations of the material. This provides more possibilities for personalization and customization of finished fabrics, giving each piece a unique style and texture.

[0317] Specifically, the fabric image can be expanded and the fabric fusion can be performed based on the pattern size, position, relative position and other information in the regulator parameters. In other words, the regulator parameters can be input into the integration module for the first fabric fusion.

[0318] Fabric material information can be in text or vector form, specifically used to describe the fabric material information of the fabric image. This information can then be input into the fabric fitting generation module for subsequent fusion. It can also be input into the integration module for the first fabric fusion.

[0319] Step 2: Style parsing module.

[0320] Users can choose to input various style images or model images. Among them, the style images include various categories, such as clothing categories (such as windbreakers, sweatshirts, dresses, etc.), or other categories such as luggage, home textiles, etc.

[0321] The input style image or model image will then be fed into the parsing model for style parsing, resulting in the required style parsing graphs. These style parsing graphs will then be fed into the integration module along with the style image for subsequent processing.

[0322] The style parsing module can be implemented based on algorithm models such as image segmentation, detection, and matting.

[0323] Step 3: Integrate modules.

[0324] As shown in FIG8 , FIG8 is a schematic diagram of a principle of an integrated module provided by an exemplary embodiment.

[0325] The integration module may include a style and fabric fusion module and a style and fabric reference module.

[0326] The input of the integration module comes from the fabric preprocessing module, including the fabric preprocessing graph, regulator parameters, and the style graph and style parsing graph in the style parsing module.

[0327] The regulator parameters include information such as fabric pattern texture position, size, pattern loop position, etc.

[0328] The integration module itself consists of a style and fabric fusion module and a style and fabric reference module. The style and fabric fusion module is responsible for fusing the style parsed region in the style image with the fabric. This module also uses the position and size of the fabric pattern texture as control conditions. After layer manipulation, it generates a preliminary fabric style fusion image. At this stage, the fusion image contains the style fold information of the guide image and a certain amount of fabric pattern texture, but may contain some missing details and color loss.

[0329] The inputs for the Style and Fabric Reference module are the same as those for the Style and Fabric Fusion module. At this stage, the generated style and fabric reference image retains the maximum possible fabric information, including pattern, texture, and color information. Specifically, this involves extracting the fabric image content based on the product outline in the style image.

[0330] Next, the fusion image and reference image from the style and fabric fusion module and the style and fabric reference module are sent to the fabric fitting generation module for subsequent generation processing.

[0331] The fusion image here is the intermediate effect image, and the reference image is the fabric image to be fused.

[0332] Step 4: Fabric fitting generation module.

[0333] In the integrated module, the style and fabric fusion image and the style and fabric reference image serve as the base input image and reference image, respectively, for the generative model. The base input image guides the generative model in generating corresponding style information, such as shape, folds, and contours. The reference image, on the other hand, guides the generative model in generating corresponding pattern, texture, and color information.

[0334] Furthermore, the fabric material provided by the fabric preprocessing module serves as a control condition for the generative model. This condition guides the generative model to produce results that are more consistent with the user's selected fabric material. This guidance allows the generated fitting effects to better meet requirements for fabric texture, stripe characteristics, pattern, color, garment pattern, drape, texture, and light and shade variations, resulting in detailed and realistic results.

[0335] This comprehensive generation process ensures that the final effect is highly consistent with the user's design vision while ensuring that the fabric is highly restored to the input fabric.

[0336] The generation model may be a diffusion model or other models.

[0337] The control conditions of the generated model can not only include fabric material, reference image and fusion image, but also more control conditions, such as fabric splicing process.

[0338] There is no particular order between step 1 and step 2; the style image and model image input in step 2 can be taken in real life or generated by other generative models.

[0339] This solution can achieve the following beneficial effects.

[0340] 1. Through the above process, this invention enables end-to-end generation of fabric images to fabric fitting images, without requiring the extensive human, material, and time investment required by traditional fabric manufacturing processes or digital modeling methods. This method is simpler and more efficient, quickly meeting the diverse needs of fabric manufacturers. The advantage of this technology is that it greatly simplifies the design and manufacturing process for fabric fittings, bringing greater convenience and flexibility to the industry.

[0341] The end-to-end, image-based fabric fitting model helps fabric manufacturers break free from the constraints of traditional garment manufacturing methods and 3D modeling. Without the need for extensive human and material resources, simply input a fabric image and instantly view the final fitting results. This not only improves efficiency but also reduces costs. It also provides manufacturers with greater flexibility, enabling them to more quickly meet market demands and more easily demonstrate how fabrics will perform across different styles and models.

[0342] 2. This solution can achieve consistency in fitting styles and fabrics from multiple perspectives. With the help of the style parsing module and the integration module, styles and fabrics can be better integrated. Under multiple different perspectives of the same style, the fitting effect of the same fabric can maintain good consistency, and inconsistent generated effects will not be caused by inconsistent perspectives. This is of great help to the final production and manufacturing, and can also help customers better view the effect display of the fabric from different perspectives of the style. Specifically, this can be done by using style pictures or model pictures from multiple perspectives to integrate fabrics separately. It can also be done by inputting guidance information for perspective transformation into the generation model, so that the effect diagram of the perspective transformation can be generated.

[0343] The fusion of the style parsing module and the integration module can ensure good consistency of the same style under different viewing angles, and will not lead to inconsistent final fabric fitting presentation due to inconsistent viewing angles. This can greatly help the final generation and manufacturing of styles and fabrics as well as the early display of effects.

[0344] 3. The style fabric fusion image processed by the style fabric fusion module and the style fabric reference image processed by the style fabric reference module serve as control conditions for the generative model, enabling optimal preservation of both fabric and style attributes. This means that the style's outline, shape, and wrinkle information can be retained, while also maintaining important information such as the fabric's pattern, texture, and color. The fabric material guidance control condition is of great significance for the final fabric fitting image. It can more realistically simulate the feel of fabric on clothing and people in the real world, including drape, texture, light and shadow, and so on. This means that the generated fabric fitting images are more realistic and closer to the actual item, helping to improve the quality of design and manufacturing and meet consumer needs and expectations.

[0345] By leveraging fusion images, reference images, and fabric material control as conditions, this method can meet a wide range of requirements to a great extent. This method preserves the shape, contours, and wrinkle information of the style while also retaining the pattern, texture, and color information of the fabric. This comprehensive conditional generation method aims to achieve the optimal balance between the finished garment and the wearable effect, ensuring a perfect combination of the designed style and fabric, further enhancing the realism and visual quality of the finished garment.

[0346] The embodiments of this specification also provide corresponding device embodiments.

[0347] As shown in FIG9 , FIG9 is a structural schematic diagram of a fabric fusion device provided by an exemplary embodiment.

[0348] The apparatus may include the following units.

[0349] The acquisition unit 301 is used to acquire standard fabric patterns and product style patterns.

[0350] The first fusion unit 302 is used to fuse the product style image and the standard fabric image to obtain an intermediate effect image.

[0351] The determining unit 303 is configured to determine the image content in the standard fabric image to be fused with the product style image as the fabric image to be fused.

[0352] The second fusion unit 304 is used to fuse the intermediate effect image and the fabric image to be fused to obtain a final effect image.

[0353] Optionally, the second fusion unit 304 is configured to:

[0354] The intermediate rendering and the fabric image to be fused are input into the preset model to obtain the final rendering output by the preset model; the preset model is used to reduce the loss of fabric information; the loss of fabric information in the final rendering is less than the loss of fabric information in the intermediate rendering.

[0355] Optionally, the acquiring unit 301 is further configured to: acquire target fabric information of the standard fabric image;

[0356] The second fusion unit 304 is used for:

[0357] The obtained target fabric information, the intermediate rendering and the fabric image to be fused are integrated to obtain the final rendering.

[0358] Optionally, the second fusion unit 304 is configured to:

[0359] The following steps are executed in a loop until a preset loop stop condition is met: the current intermediate rendering image and the fabric image to be fused are fused to obtain a new rendering image; the obtained new rendering image is determined as the current intermediate rendering image;

[0360] After the loop stops, the current intermediate rendering is determined as the final rendering.

[0361] Optionally, the determining unit 303 is configured to:

[0362] Based on the intermediate rendering, the image content in the standard fabric image used for fusion with the product style image is determined as the fabric image to be fused; or

[0363] The product outline in the product style diagram is overlapped with the standard fabric diagram in a preset manner, and the image content contained in the product outline in the standard fabric diagram is determined as the fabric diagram to be fused.

[0364] Optionally, the acquiring unit 301 is used to: acquire an initial fabric image;

[0365] The initial fabric graph is preprocessed to obtain a standard fabric graph.

[0366] Optionally, the acquiring unit 301 is configured to:

[0367] Get initial product images;

[0368] Performing image recognition on the initial product image to determine product image content in the initial product image;

[0369] A product style image is determined based on the product image content identified in the initial product image.

[0370] Optionally, the acquiring unit 301 is further configured to: acquire target parameters;

[0371] An initial fabric map is obtained; repeated content of the fabric map in the initial fabric map is determined based on target parameters; and several copies of the repeated content of the fabric map are spliced ​​together to obtain a standard fabric map.

[0372] Optionally, the acquisition unit 301 is further configured to: acquire target parameters;

[0373] The determining unit 303 is used to: overlap the product outline in the product style map with the standard fabric map according to the relative position in the target parameter, and determine the image content contained in the product outline in the standard fabric map as the fabric map to be fused.

[0374] The detailed explanation of the above device embodiment can be found in the explanation of the method embodiment.

[0375] As shown in FIG10 , FIG10 is a schematic structural diagram of another fabric fusion device provided by an exemplary embodiment.

[0376] The apparatus may include the following units.

[0377] The image acquisition unit 401 is used to acquire standard fabric images and clothing style images.

[0378] The third fusion unit 402 is used to fuse the clothing style image and the standard fabric image to obtain an intermediate clothing effect image.

[0379] The unit to be fused 403 is used to determine the image content in the standard fabric image to be fused with the clothing style image as the fabric image to be fused.

[0380] The fourth fusion unit 404 is used to fuse the intermediate clothing rendering and the fabric image to be fused to obtain a final clothing rendering.

[0381] Optionally, the fourth fusion unit 404 is used to: input the intermediate clothing rendering and the fabric image to be fused into a preset model to obtain a final clothing rendering output by the preset model; the preset model can be used to reduce the loss of fabric information; the loss of fabric information in the final clothing rendering can be less than the loss of fabric information in the intermediate clothing rendering.

[0382] Optionally, the image acquisition unit 401 is further used to: acquire target fabric information of the standard fabric image; the fourth fusion unit 404 is used to: fuse the acquired target fabric information, the intermediate clothing effect image and the fabric image to be fused to obtain the final clothing effect image.

[0383] Optionally, the fourth fusion unit 404 is used to: loop through the following steps until a preset loop stop condition is met: fuse the current intermediate clothing rendering and the fabric image to be fused to obtain a new clothing rendering; determine the obtained new clothing rendering as the current intermediate clothing rendering; after the loop stops, determine the current intermediate clothing rendering as the final clothing rendering.

[0384] Optionally, the unit to be fused 403 is used to: determine the image content in the standard fabric image used for fusion with the clothing style image as the fabric image to be fused based on the intermediate clothing effect image; or overlap the clothing outline in the clothing style image with the standard fabric image in a preset manner, and determine the image content contained in the clothing outline in the standard fabric image as the fabric image to be fused.

[0385] Optionally, the image acquisition unit 401 is configured to: acquire an initial fabric image; and preprocess the initial fabric image to obtain a standard fabric image.

[0386] Optionally, the image acquisition unit 401 is configured to: acquire an initial clothing image; perform image recognition on the initial clothing image to determine clothing image content in the initial clothing image; and determine a clothing style image based on the clothing image content recognized in the initial clothing image.

[0387] Optionally, the image acquisition unit 401 is further configured to: acquire target parameters; acquire an initial fabric image; determine repeated content of the fabric image in the initial fabric image based on the target parameters; and splice several copies of the repeated content of the fabric image to obtain a standard fabric image.

[0388] Optionally, the image acquisition unit 401 is also used to: obtain target parameters; the unit to be fused 403 is used to: overlap the product outline in the clothing style image with the standard fabric image according to the relative position in the target parameters, and determine the image content contained in the clothing outline in the standard fabric image as the fabric image to be fused.

[0389] In the manufacturing industry, many products require color matching, such as clothing, luggage, home textiles, toys, and so on.

[0390] Taking the clothing industry as an example, designers usually do not just design one color scheme for a certain style, but provide a whole set of diverse color combinations to meet the aesthetic needs and fashion trends of different consumers.

[0391] The clothing design process requires determining matching color schemes and styles, and determining the desired effect, which requires extensive trial and error. Experimenting with different color schemes for specific clothing styles often requires significant costs to produce a single garment. Similarly, for products like luggage, home textiles, and toys, experimenting with different color schemes for different styles often requires significant labor and resources to produce the actual product.

[0392] This method often requires a lot of manpower and has poor real-time performance.

[0393] In order to solve the above problems, the embodiments of this specification provide a product color matching method.

[0394] The method can obtain a color scheme and a product style diagram, and the device automatically fuses the color scheme and the product style diagram to obtain a rendering, that is, a simulated rendering of a product manufactured in the color specified by the color scheme according to the style in the product style diagram.

[0395] This method can obtain effect pictures by fusing images through devices, thereby reducing manual participation and improving the efficiency and real-time performance of viewing product color matching effects.

[0396] In addition, when specifically fusing color schemes and product style images, the intermediate rendering obtained from the first fusion can be further combined with the color scheme for a second fusion to obtain the final rendering, so as to improve the color performance effect in the final rendering and reduce the loss of color information during the fusion process.

[0397] For ease of understanding, an example is provided below with reference to the accompanying drawings.

[0398] As shown in FIG11 , FIG11 is a schematic diagram showing a principle of a color fusion process provided by an exemplary embodiment.

[0399] Taking clothing as an example, the color matching and clothing style diagram can be integrated to obtain a clothing rendering, that is, a simulated rendering of a clothing product manufactured according to the style in the clothing style diagram using the colors specified by the color matching.

[0400] Among them, the sleeves and collar in the clothing style picture are gray, while the main part is white.

[0401] Specifically, the sleeves and collar in the clothing style image may be configured to be black, while the main body may be configured to be gray. Specifically, after the colors specified by the color matching are merged, an effect image may be obtained, in which the sleeves and collar are merged to black, while the main body is merged to gray.

[0402] It should be noted that only black, white and gray are used as examples here, and a variety of colors can actually be used for color matching, such as green, red, purple, etc.

[0403] For ease of understanding, a product color matching method is explained below with reference to the accompanying drawings.

[0404] As shown in FIG12 , FIG12 is a flow chart of a product color matching method provided by an exemplary embodiment.

[0405] The embodiments of this specification do not limit the execution subject of this method.

[0406] Optionally, the method can be applied to a computing device or a module in the device. For example, the method can be applied to a server, a terminal, or other devices.

[0407] The method may include the following steps.

[0408] S1101: Obtain product style pictures and color matching information; the color matching information includes at least one color information.

[0409] S1102: Integrate the product style image and color matching information to obtain an intermediate rendering.

[0410] S1103: Fusing the intermediate rendering and color matching information to obtain the final rendering.

[0411] The intermediate rendering and the final rendering can both be simulated renderings of a product manufactured in accordance with the style in the product style diagram using the specified color scheme. The color information loss in the final rendering can be less than the color information loss in the intermediate rendering.

[0412] The above method process can obtain the effect diagram of product color matching by automatically integrating color matching colors and product style diagrams, thereby reducing manual participation and improving the efficiency and real-time performance of viewing product color matching effects.

[0413] In addition, the above method process can also improve the color performance effect in the final rendering by fusing the color matching colors again for the intermediate rendering, thereby reducing the loss of color information during the fusion process.

[0414] This method is not limited to the specific products in the product style diagram. Alternatively, the product style diagram can specifically include clothing, luggage, home textiles, toys, and other products. For example, clothing style diagrams, toy style diagrams, luggage style diagrams, and so on. These products all require color matching to verify the manufacturing effect, so that simulated renderings can be obtained based on the above method.

[0415] This method flow explains the steps of integrating color schemes for a product style image. It can be understood that color schemes can be integrated separately for multiple product style images. For specific method steps, please refer to this method flow.

[0416] In this method, both S1102 and S1103 perform color fusion to obtain the effect image. This method does not limit the number of fusions, methods, objects, and whether the methods are the same. Optionally, S1102 and S1103 can use the same method to perform color fusion, or different methods can be used to perform color fusion.

[0417] Optionally, for the two fusions between S1102 and S1103 , the same color information may be used to fuse the same area of ​​the product style image and the intermediate rendering, thereby reducing the possibility of conflict between the fused colors.

[0418] The following is a detailed explanation of each step.

[0419] 1. S1101: Obtain product style images and color matching information; the color matching information includes at least one color information.

[0420] 1. About the content of product style pictures.

[0421] This method does not limit the content of the product style diagram, as long as the product style diagram contains product styles.

[0422] For example, the product may be clothing, and the clothing style image may be an image of a model wearing clothing, or may only include clothing, or may be an image of clothing obtained through software modeling and simulation.

[0423] For another example, the product may be luggage or home textiles. The product style image may include an image of the luggage or home textiles, specifically an image with a model or background, or an image containing only the product itself, that is, an image containing only luggage or home textiles.

[0424] Optionally, the product style image can include the product, which is used for subsequent color fusion. To facilitate subsequent color fusion, the product style image can be limited to the product. Specifically, this can be done by extracting the product image content from the image through image segmentation or image recognition to use as the product style image.

[0425] For example, if an image contains a person carrying a bag, you can extract the bag image content as the product style image. Or if an image contains a model trying on clothes, you can extract the clothing image content as the product style image.

[0426] Taking clothing as an example, a clothing style map can optionally include clothing items, specifically clothing items that will be used for subsequent color fusion. To facilitate subsequent color fusion, the clothing style map can be limited to clothing items. Specifically, this can be achieved by extracting clothing image content from an image through image segmentation or image recognition to create the clothing style map.

[0427] Therefore, optionally, the product style diagram may only contain the product, thereby facilitating subsequent color fusion.

[0428] To facilitate understanding, product images of different standards can be input to meet the different needs of different users such as designers, users, and manufacturers. After processing, product style images can be obtained, or they can be directly used as product style images for subsequent color fusion to meet the needs of different users.

[0429] Furthermore, it should be noted that color fusion is performed on the product. After obtaining the final rendering, the product in the final rendering can be copied back into the original image. In other words, the original image containing the product (the image used to extract the product image content) can be replaced with the color-fused product rendering to improve the product display effect.

[0430] For example, for an image of a model trying on clothes, you can first extract the clothing part as a clothing style image. After color fusion to obtain the final clothing rendering with a new color scheme, you can copy the clothing with the new color scheme in the final clothing rendering to the original image of the model trying on clothes, so that you can replace the clothes worn by the model in the image, making it easier to check the fitting effect.

[0431] In an optional embodiment, a product style diagram may include identification information for different product regions. For example, taking a top as an example, the product style diagram may include identification information for different regions of the top, specifically including identification information for the collar, cuffs, chest, hem, and body, thereby making it easier to identify the style of the product.

[0432] Taking luggage products as an example, the luggage style image can include identification information of different parts of the luggage, specifically including the sides, bottom, interior, top, etc., so as to facilitate the clarification of the style information of the product itself.

[0433] Taking home textile products as an example, the home textile style map can include identification information of different parts of the home textile, specifically including bedding, sheets, pillows and other areas, so as to facilitate the clarification of the style information of the product itself.

[0434] Of course, the product style diagram can also include identification information of the entire product area, so that it is convenient to match the color of the entire product.

[0435] Based on the identification information of different parts of the product in the product style diagram, on the one hand, it is convenient to determine which parts of the area need to be blended with colors and the colors that need to be blended according to preset instructions; on the other hand, the product style can be determined in combination with the identification information, and the product style information can be introduced to facilitate improving the color fusion effect.

[0436] 2. How to obtain product style pictures.

[0437] This method does not limit the method or source for obtaining product style images. Alternatively, product style images can be directly obtained, or other images containing products can be obtained and processed to obtain product style images. Alternatively, product style images can be obtained from user input or specified, or randomly obtained.

[0438] In an optional embodiment, an image containing a product input by a user may be directly obtained as a product style image, or an image containing a product selected by a user may be obtained as a product style image.

[0439] In an optional embodiment, a method for obtaining a product style image may include: obtaining an initial product image; performing image recognition on the initial product image to determine the product image content in the initial product image; and determining the product style image based on the product image content recognized in the initial product image.

[0440] This embodiment does not limit the specific image recognition method. Specifically, image recognition can be performed based on a deep learning model or neural network for identifying products.

[0441] This embodiment does not limit the content and source of the initial product image, as long as the initial product image contains the product. For example, taking clothing as an example, the initial product image can be an initial clothing image, which can be an image of a model trying on the clothing, an image of the clothing itself, or a photographed image of the clothing. Alternatively, taking home textiles as an example, the initial product image can be an initial home textile image, specifically an image of bed sheets and quilts, or a photographed image of other bedding.

[0442] This embodiment also does not limit the specific form of the product image content. Optionally, the product image content can be the image content occupied by the product in the initial product image.

[0443] For ease of understanding, the product image content may specifically be product detection performed on the initial product image to determine a product detection frame in the initial product image, and the image content in the product detection frame is determined as the product image content.

[0444] For example, taking clothing products as an example, clothing detection can be performed on the initial clothing image to determine a clothing detection frame in the initial clothing image, and the image content in the clothing detection frame is determined as the clothing image content.

[0445] This embodiment does not limit the specific method of determining the product style image based on the product image content. Optionally, the product image content can be directly determined as the product style image, or the product image content can be processed to obtain the product style image.

[0446] A specific processing method, for example, is contour segmentation, which can specifically determine the product contour in the product image content and obtain the image content in the product contour as a product style diagram to improve the accuracy of the product style. For another example, a more detailed identification and segmentation can be performed on the product image content to determine the product style, that is, for each part in the product image content, it is necessary to identify which part of the product it is. Taking clothing as an example, for the image content of a top, the collar image content, sleeve image content, main body image content, etc. in the top image content can be further identified, or the collar area, sleeve area, main body area, etc. can be identified. The content of these identified areas can be used as information and determined as a product style diagram together with the product image content.

[0447] Therefore, optionally, the product style diagram is determined based on the product image content identified in the initial product image. Specifically, it can be that regional recognition is performed on the product image content identified in the initial product image to determine the identification information of different areas of the product; and the product style diagram is determined based on the determined identification information and product image content.

[0448] This embodiment does not limit the specific method of performing region recognition. Optionally, image segmentation, image detection and other algorithms can be used to perform region recognition.

[0449] This embodiment does not limit the specific method of determining the product style map based on the determined identification information and product image content. Optionally, the determined identification information and product image content can be directly determined as the product style map, or the regional content with identification information can be combined to determine the product style map.

[0450] Furthermore, it is understood that when performing image recognition on the initial product image, different product regions in the initial product image may also be directly identified and determined. The determined product image content may include identification information of different product regions.

[0451] Optionally, based on the region identification information in the product image content, the region to be fused can be further determined. This embodiment does not limit the specific determination method, and the region to be fused can be determined based on user designation or pre-stored information.

[0452] For example, the colors can be blended only for the cuff area of ​​a top garment according to user specifications, or for the entire area of ​​a trouser garment according to pre-stored settings.

[0453] For ease of understanding, this specification also provides a specific example. Taking clothing products as an example, users can choose to input various style images or model images. Among them, style images include multiple categories, such as clothing categories (such as windbreakers, sweatshirts, dresses, etc.), or other categories such as luggage, home textiles, etc.

[0454] The input for a style image can be a pure style image or a model wearing the style image. Considering that users may need to see how the style will look in a real-life scenario, the user can directly input a model wearing the style image, which effectively reduces the cost investment in the actual production stage.

[0455] The input style image or model image will then be fed into the parsing model for style parsing, resulting in the required style parsing graphs. These style parsing graphs will then be fed into the integration module along with the style image for subsequent processing.

[0456] The input style image or model image will be parsed by the parsing model. The function of the parsing model is to parse and partition the various areas of the style image. For example, the collar, sleeves, cuffs, pockets and other parts of the clothing will be accurately parsed into separate areas, providing detailed information for the subsequent color fusion of different areas.

[0457] In addition to identifying individual regions, the parsing model also helps the network model understand their composition, providing a better foundation for subsequent color fusion. Understanding and parsing style images at a higher level of detail improves the system's understanding of style structure, enabling more customized color fusion.

[0458] 3. About color matching information.

[0459] This method does not limit the specific content and format of the color matching information. Optionally, the color matching information may include at least one color information, thereby facilitating product color matching. Optionally, the color matching information may also include other information, such as the area in the product style diagram that requires color matching, the area that requires a specific color, or the color matching method.

[0460] In an optional embodiment, since the color matching information needs to be used for product color matching, the color matching information may include at least one color information to facilitate product color matching.

[0461] If the color matching information includes only one color, the color matching can be performed based on this single color. The specific color matching method can be determined based on the product area that requires color matching. If the color matching information includes multiple colors, specific color matching methods can be considered, such as color matching using a gradient blend of multiple colors or multiple colors in parallel. In particular, when matching multiple colors for a single product area, color matching methods such as multiple colors in parallel, multiple color blending, and multiple color gradients can be considered.

[0462] As you can see, in the subsequent product color matching process, in addition to the colors to be matched, the product areas and color matching methods must also be determined. For example, for luggage products, in addition to determining that the colors to be matched are white and red, it is also necessary to determine that the side areas of the luggage need to be matched, and further determine that the color matching method can be red and white. Then, the red and white colors can be integrated into the side areas of the luggage to obtain the luggage effect after color matching.

[0463] Of course, this information may be included in the color matching information, or may be determined in a subsequent step, specifically by selection, designation, or random determination.

[0464] Therefore, the color matching information may optionally include at least one of the following: at least one color information, color information corresponding to a region in the product style image, a color matching method corresponding to a region in the product style image, etc. Optionally, the color matching information may specifically include a set of regional color matching correspondences. The set of regional color matching correspondences may include at least one set of correspondences between colors and regions in the product style image.

[0465] Furthermore, the form of the color matching information is not limited, and may be in text form, table form, vector form, etc.

[0466] After the content and form of the color matching information are clearly explained, the present method does not limit the method for obtaining and determining the color matching information.

[0467] Optionally, the color information may be obtained from a local device, an external device, or an external input, etc. Specifically, the color information may be obtained from user input, pre-stored color information, or locally automatically randomly generated color information.

[0468] In a specific example, color information can be user-input, that is, user-specified or selected. For example, a user may specify black and green as the overall color scheme for a clothing product, and may also specify a gradient blending scheme. Color information can also be pre-stored, and the specific method for determining it is not limited. For example, pre-stored color information can be obtained, such as "the overall color scheme for a toy product is red and blue, and a multi-color juxtaposition scheme can also be specified." Pre-stored color information can be associated with a product style or product type.

[0469] Optionally, the color matching information may be determined by user input or automatically determined by the device. The automatic determination of the color matching information by the device may be generated randomly, based on a preset correspondence, or based on other information, such as keywords.

[0470] For example, the color matching information determined and input by the user can specifically be color matching information specified by the user, and the user can input or select a color as the color matching information. The randomly generated color matching information can be directly randomly selecting color information as the color matching information, so as to expand the possibility of color matching. In addition, color information can also be generated based on specified information. For example, color information is generated based on specified keywords. Taking the keyword "warm" as an example, colors such as red, orange, and yellow can be generated. Of course, the specified information can specifically be keywords specified, input, or selected by the user, or it can be pre-stored keywords, which is not limited here.

[0471] This embodiment does not limit the specific method of generating color matching information based on other information. Optionally, the color matching information can be searched based on a predetermined correspondence between the other information and the color matching information, or the color matching information can be generated based on the other information using a machine learning trained model. The model can specifically be a color generation model.

[0472] This embodiment does not limit the specific content and form of the other information. Optionally, the other information may specifically be information that characterizes the atmosphere or information that characterizes color matching requirements. The specific form may be text or vector. Other information, for example, atmosphere keywords, may include "energizing," "cold," "warm," "lively," and so on.

[0473] Therefore, color information can be obtained directly, for example, by directly obtaining color information input by a user or by directly obtaining pre-stored color information. Alternatively, color information can be generated indirectly, for example, by first obtaining other information and then generating color information based on the other information. Other information can include atmosphere keywords. Alternatively, the other information can be input into a color generation model to obtain color information output by the color generation model.

[0474] In addition, in order to facilitate subsequent color fusion, more information can be further mined from the color matching information for subsequent color fusion.

[0475] Alternatively, taking atmosphere as an example, the colors configured in the color matching information often have a certain atmosphere, such as warm, lively, fresh, etc. Therefore, atmosphere information can be further extracted from the color matching information for use in subsequent fusion steps. The atmosphere information here can be used to characterize the atmosphere of the colors in the color matching information. The specific form is not limited and can be text or vector form.

[0476] Of course, since the color matching information itself can also be determined based on the atmosphere information, such as the above-mentioned atmosphere keywords, the atmosphere information can be directly used in the subsequent fusion steps.

[0477] Furthermore, this embodiment does not limit the specific method for extracting atmosphere information. Specifically, it can be performed by searching based on the correspondence between pre-stored atmosphere information and color matching information, or it can be determined based on a model trained by machine learning. For example, a color atmosphere encoding model can be used to encode the color information in the color matching information into atmosphere information. The specific form can be vector form.

[0478] The above examples are based on atmosphere, but they can also be based on emotions and other information.

[0479] For easier understanding, a specific example is given below.

[0480] Users can specify their desired color palette in two ways: directly by providing a color list, or by using emotionally expressive color keywords, such as "energized," "calm," and "vital." If the user chooses to enter a color keyword, the system will use a pre-trained color generation model to generate a specific color palette based on the user's provided keywords. This generated color palette, based on the user's input keywords, effectively conveys the desired emotion and message.

[0481] Users can directly provide a customized color palette to ensure the desired color scheme meets their specific requirements and tastes. This approach is suitable for users who have already clearly defined their desired colors. If users prefer to express emotions rather than specify colors, they can use color-related emotional keywords such as vitality, calmness, and vitality. This method is more flexible, allowing users to guide color generation through emotional keywords without having to deeply understand specific color combinations. The pre-trained color generation model can automatically generate corresponding color palettes based on the emotional keywords entered by the user. It has a certain degree of intelligence and creativity, and can better meet user expectations.

[0482] The user-specified color series, or the color series generated by inputting color keywords, is then fed into the color encoding model. In this step, the color encoding model generates a coding vector that characterizes the emotional atmosphere of the color series. This coding vector serves as the input control condition for the subsequent generation model, precisely controlling the atmospheric characteristics conveyed by the generated image. The color encoding model is responsible for converting the user-specified or generated color series into a specific coding vector. This model not only considers the visual effects of the colors, but also their underlying emotional and atmospheric characteristics.

[0483] 4. About the regional color matching correspondence set.

[0484] The color matching information may specifically include a regional color matching relationship set, which is further explained below. The at least one color information included in the regional color matching relationship set may also be the at least one color information included in the color matching information.

[0485] The present method flow does not limit the content of the set of regional color matching correspondences. The set of regional color matching correspondences may include at least one set of correspondences between regions and colors in a product style diagram. More specifically, the present method flow does not limit the content of regions and colors. Optionally, the region may specifically be the entire region of the product image content in the product style diagram, thereby enabling color matching for the entire product. The region may also be the region occupied by the image content of different parts of the product in the product style diagram, such as the collar region of clothing, the side region of luggage, the component region of a plastic toy, and so on, thereby enabling color matching for different parts of the product. Specific colors may also include one or more colors. For example, a single color may be configured for a region, or multiple colors may be configured for a region. More specifically, how to configure the colors can be further determined based on the actual fusion method. For example, the hue of the image content in the region may be directly adjusted to the corresponding color, or color fusion may be performed through machine learning.

[0486] This method does not limit the form of the above-mentioned correspondence. Specifically, it can be in the form of text, image, or vector. More specifically, this method does not limit the form of region and color. The form of the region can be determined based on the product style diagram and can be in the form of text or image.

[0487] For example, for a hat, you can use text to specify the colors for the brim, body, and accessories. For a car, you can use images to specify the colors for the front, body, and rear of the car.

[0488] Based on the above-mentioned regional color correspondence set, in an optional embodiment, the color information may include: a regional color correspondence set; the regional color correspondence set may include at least one set of correspondences between colors and regions in the product style diagram. S1102: Fusion of the product style diagram and the color information to obtain an intermediate rendering, specifically: according to the regional color correspondence set, for the regions in the product style diagram, the corresponding colors are fused to obtain an intermediate rendering. S1103: Fusion of the intermediate rendering and the color information to obtain a final rendering, specifically: according to the regional color correspondence set, for the regions in the intermediate rendering, the corresponding colors are fused to obtain a final rendering.

[0489] This embodiment can improve the efficiency and accuracy of color fusion based on the regional color matching correspondence set.

[0490] This method does not limit the source of the region information in the region color correspondence set. It can be pre-stored, user-entered, or obtained from a product style image or identified from a product style image.

[0491] Optionally, the pre-stored region information may specifically be a pre-designated region, such as the entire product area. Optionally, the region information input by the user may specifically be a region designated by the user for color matching, such as the side, bottom, or top surface of the product. The specific region information may be determined based on user needs.

[0492] It should be noted that different product types and styles may have different area divisions. Therefore, it is possible to obtain pre-stored area information or user-input area information based on the specific product type or style in the product style diagram. For example, for a top in a clothing style diagram, the pre-stored area information corresponding to the top can be obtained, which may specifically be areas such as the collar, cuffs, and body. For a car in a toy style diagram, the area information input by the user can be obtained, which may specifically be areas such as the front, body, and rear of the car. Of course, it is also possible to obtain pre-stored general area information or obtain user-input general area information. For example, the overall area of ​​the product.

[0493] Optionally, if the product style image includes region information, pre-stored region information may be obtained, or region information input or selected by the user may be obtained. Specifically, the region information may be selected by the user from the region information included in the product style image. Of course, the region information included in the product style image may also be obtained directly.

[0494] If the product style image does not contain region information, the region information can be further identified and obtained from the product style image. This embodiment does not limit the specific region identification method. Alternatively, image segmentation, image detection, and other algorithms can be used for region identification. For a detailed explanation, please refer to the above explanation.

[0495] This method does not limit the source of the color information in the regional color matching relationship set. It can be pre-stored, user-entered, randomly generated, etc. For details, please refer to the explanation of color matching information above.

[0496] Optionally, the pre-stored color information may specifically be color information that is pre-specified to be configured, such as red, white, and the like. Optionally, the color information input by the user may specifically be color difference information specified by the user, for example, the user may input or select a color as the color information. Optionally, the randomly generated color information may be color information that is directly randomly selected as the color information in order to expand the possibility of color matching, or color information may be randomly generated based on specified information. For example, color information may be randomly generated based on specified keywords. Taking the keyword "warm" as an example, colors such as red, orange, and yellow may be randomly generated. Of course, the specified information may specifically be keywords specified, input, or selected by the user, which is not limited here.

[0497] After explaining the sources of the region information and color information in the region color correspondence set, the source of the correspondence between the region information and color information can be further explained, which can be pre-stored, user-input, randomly generated, etc.

[0498] Optionally, the correspondence between regional information and color information can be pre-stored. For example, the overall image area of ​​the product can be pre-stored, corresponding to the black color, so that the product in any product style diagram can be integrated with the black viewing effect. Optionally, the correspondence between regional information and color information can be constructed based on the information input by the user. Specifically, the user can specify the color information based on the regional information, specify the regional information based on the color information, or directly input the correspondence between the regional information and the color information. For example, for a set of regional color matching correspondences, the user can select or input the corresponding color information for the preset regional information, or the user can select or input the corresponding regional information for the preset color information, or the user can directly input a set of correspondences between regional information and color information.

[0499] Alternatively, the correspondence may be randomly generated, specifically by randomly combining the region information and the color information, that is, randomly assigning the corresponding color information to the region information, randomly assigning the corresponding region information to the color information, or directly randomly generating the correspondence between the region information and the color information.

[0500] The method does not limit the specific method of obtaining the regional color matching relationship set. Optionally, the method may directly obtain a pre-stored regional color matching relationship set, or determine the regional color matching relationship based on user input or product style images, and then further obtain the regional color matching relationship set.

[0501] This method does not limit the specific method of determining the set of regional color matching correspondences. Please refer to the above explanation for details.

[0502] Optionally, the method for determining the regional color matching relationship set may include any of the following:

[0503] 1) Obtaining colors specified for regions in a product style diagram; constructing at least one set of correspondences between regions and colors in the product style diagram based on the specified colors, and obtaining a set of region color matching correspondences.

[0504] 2) Obtaining atmosphere cues specified for regions in the product style diagram, and automatically generating corresponding colors based on the obtained atmosphere cues; constructing at least one set of correspondences between regions and colors in the product style diagram based on the generated colors, to obtain a set of regional color matching correspondences.

[0505] 3) Obtaining a color scheme; for each region in the product style diagram, randomly selecting a color from the obtained color scheme as a corresponding color, constructing at least one set of correspondences between the region and the color in the product style diagram, and obtaining a set of region color correspondences.

[0506] Among them, optionally, specifically obtaining the color specified for the area in the product style diagram may be obtaining the specified color input by the user, that is, the user may input the color specified for the area in the product style diagram, further facilitating the automatic construction of the regional color matching correspondence.

[0507] This embodiment does not limit the form of the atmosphere prompt. Specifically, it can be in text or vector form. For example, warm, energetic, cold, etc. Specifically, machine learning can be used to generate colors based on the atmosphere prompt, or corresponding colors can be found based on the preset correspondence between atmosphere prompts and colors. The atmosphere prompt can be input or specified by the user, or it can be randomly selected.

[0508] This embodiment does not limit the form and content of the color scheme. Optionally, the color scheme may include one or more color information, and when randomly selecting colors, one or more colors may also be randomly selected. The source of the color scheme is also not limited, and may be user input or model generated. For example, a user-input color scheme of "red, yellow, and white" may be obtained, and further, a random selection may be made from three colors as the corresponding color for an area of ​​a product style image.

[0509] Of course, the method of determining the set of regional color matching correspondences is not limited to the above three examples. You can also select regions in the product style diagram for color matching, or directly randomly generate regional color matching correspondences, etc.

[0510] Furthermore, the above embodiment explains the region information in the color matching information and the correspondence between the region information and the color information. In an optional embodiment, the color matching information may also include a color matching method corresponding to the region of the product style image. This embodiment does not limit the content, format, or acquisition method of the color matching method. Optionally, it can be user-entered, pre-stored, randomly selected, etc.

[0511] Color matching methods include single color gradient, single color tiling, multiple colors juxtaposed, multiple colors gradient mixing, etc.

[0512] It is understood that including region information corresponding to the color information, or the color matching method corresponding to the color information, in the color matching information can help with color fusion. Even if the color matching information does not include region information and color matching method, the region and color matching method can be determined or selected in real time during the specific fusion process for color fusion.

[0513] 2. S1102: Integrate the product style image and color matching information to obtain an intermediate rendering.

[0514] The method is not limited to a specific method of blending colors. Optionally, the intermediate rendering can be a simulated rendering of a product manufactured in accordance with the style in the product style diagram using the colors specified in the color matching.

[0515] Optionally, a layer operation can be used to merge the product style image with the color matching information to obtain an intermediate rendering. Specifically, the color or tone of the product style image can be changed according to the color matching information.

[0516] Optionally, a deep learning approach can be used to input the color matching information and the product style image into a pre-trained model for fusion to obtain an intermediate rendering.

[0517] This embodiment does not limit the specific model structure or training method. It is understandable that color fusion can be performed by training a neural network model or a generative model, specifically a generative model of image-to-image.

[0518] Taking the generative model as an example, color matching information can be used as a guide to perform color fusion based on product style images.

[0519] It can be understood that, optionally, the intermediate rendering can be a product simulation rendering based on the product style in the product style diagram and combined with the color matching information, and the product style information in the intermediate rendering can be the same as the product style information in the product style diagram, and only the color matching used is replaced.

[0520] Optionally, the intermediate rendering can retain the product style information in the product style image. This embodiment does not limit the specific product style information. Product style information includes, for example, the product shape, the product pattern, the product wrinkles, the product outline, etc. The specific product color can be combined or replaced with the color matching information to obtain an intermediate rendering that retains the product style information but incorporates the color information in the color matching information.

[0521] Of course, optionally, when the product style image includes product style information, the product styles under different viewing angles can be simulated, and the color matching information can be further integrated to obtain an intermediate rendering.

[0522] In an optional embodiment, the intermediate rendering obtained in step S1102 is a preliminary fusion effect, which can be further fused in step S1103. Therefore, the fusion process in step S1102 can prioritize preserving the product style information or other product information in the product style image, such as product wrinkles, product lighting, product shape, product patterns, etc.

[0523] Optionally, during the specific fusion, the colors can be fused to fit the product's own wrinkles, or the colors can be fused to fit the product's own light information or illumination information, thereby improving the color fusion effect of the intermediate rendering.

[0524] In an optional embodiment, a pre-trained model may be used to fuse product style images and color matching information to obtain an intermediate rendering.

[0525] This embodiment does not limit the specific model structure and training method. Alternatively, a generative model or a neural network model can be used. For ease of description, the model used to fuse color matching information and product style images to obtain the intermediate renderings is referred to as the first fusion model.

[0526] For ease of understanding, this embodiment can provide an embodiment of a training method. Optionally, images of different products obtained by manufacturing the same product style using different color combinations can be taken separately, and the color matching information can be extracted as training samples for training the above-mentioned first fusion model. Of course, it is also possible not to manufacture actual products, but to simulate the same product style in simulation software (for example, product style design software) by replacing multiple different color matching effects, and obtain images by screenshots or photography.

[0527] In a specific example, for the same style of clothing, clothing manufactured using different color schemes (a first color scheme and a second color scheme) can be simulated, respectively, referred to as first color-matched clothing and second color-matched clothing. Furthermore, an image of the first color-matched clothing and a second color-matched clothing can be captured. Specifically, these can be different images captured from the same perspective with the same style information. Alternatively, they can be different images captured from different perspectives. Since the clothing is simulated by software, at the same perspective, the specific style information can be such that the folds, shape, pattern, etc. of the clothing are the same between different images, and the lighting information can also be the same.

[0528] Furthermore, the first color scheme and the second color scheme may be obtained, which may be directly captured from the first color matching clothing image and the second color matching clothing image, or may be obtained by additional photography.

[0529] Afterwards, the first color scheme and the second color scheme clothing image can be used as sample features, and the first color scheme clothing image can be used as a sample label. Alternatively, the second color scheme and the first color scheme clothing image can be used as sample features, and the second color scheme clothing image can be used as a sample label. Thus, two training samples can be obtained.

[0530] By analogy, we can combine any two different color schemes to obtain two training samples.

[0531] The obtained training samples can be used to train the first fusion model.

[0532] Therefore, optionally, the method for constructing a training sample set for the first fusion model may include: obtaining different product images using different color schemes under the same product style, and obtaining information on the different color schemes used; using the product images using the first color scheme and the second color scheme information as sample features, and the product images using the second color scheme as sample labels, to construct training samples, and add them to the training sample set. The first color scheme and the second color scheme are different color schemes. Of course, the first color scheme can be any color scheme, and the second color scheme can be any color scheme other than the first color scheme. This is only for illustrative purposes. Optionally, it is also possible to use product images using any color scheme and any other color scheme information as sample features, and use product images using the other color schemes as sample labels, to construct training samples, and add them to the training sample set.

[0533] The constructed training sample set can be used to train the first fusion model. This embodiment can improve the training effect of the first fusion model and improve the color fusion effect of the intermediate effect image.

[0534] Therefore, optionally, the product style image and the color matching information are fused to obtain an intermediate rendering, which can be specifically: the product style image and the color matching information are input into the first fusion model to obtain the intermediate rendering output by the first fusion model.

[0535] In an optional embodiment, in addition to the color information and product style images that need to be fused, other information may also be fused during the specific fusion process. This embodiment does not limit the specific content of the other information. For example, the other information may be a specific description of the color information, such as the color matching method, the regional color matching correspondence, and so on. It may also be other information used to guide the fusion, such as the area information of the product style image where the colors need to be fused, specifying the area in the product style image where the colors need to be fused; for example, the fusion method information may specify whether the fusion is performed by deep learning or layer operation.

[0536] In other words, the color information can optionally be further integrated with information such as the correspondence between the colors in the color matching information and the product style image area, as well as the color matching method corresponding to the product style image area. This information can be used as a guide for color fusion. Of course, this information can also be determined in real time or randomly, rather than included in the color matching information. The specific determination method can be explained above.

[0537] This embodiment does not limit the specific form of other information, which may be text, image, vector, etc.

[0538] Optionally, you can obtain atmosphere information corresponding to the color information and fuse the product style image and color information to produce an intermediate rendering. Specifically, you can fuse the acquired atmosphere information, product style image, and color information to produce an intermediate rendering. The atmosphere information can be extracted or searched based on the color information. Specifically, the atmosphere information can include keywords such as "warm," "lively," and "calm" that represent the atmosphere. For a detailed explanation, see above. Incorporating atmosphere information can improve the effect of color fusion.

[0539] This embodiment does not limit the specific method of integrating the atmosphere information. Optionally, taking the generation model as an example, the atmosphere information can be used as a guide to integrate the product style image and color matching information to obtain an intermediate effect image.

[0540] Furthermore, this method does not limit the number of times color information is fused to obtain an intermediate rendering. Specifically, for a product style image, color information can be fused once or multiple times to obtain an intermediate rendering, thereby improving the color fusion effect through multiple fusions.

[0541] Optionally, the product style image and color matching information are integrated to obtain an intermediate rendering, which can be specifically: looping the following steps until a preset loop stop condition is met: integrating the current product style image and color matching information to obtain a new rendering; determining the obtained new rendering as the current product style image; after the loop stops, determining the current product style image as the intermediate rendering.

[0542] It is understood that the fusion method for each cycle can be explained above. For example, a preset model can be input for fusion. Specifically, it can be a generative model, such as a diffusion model.

[0543] This embodiment does not limit the preset loop stop condition. Specifically, it can be set to the maximum number of loops, or to whether the difference between the new rendering and the current product style is small enough, or it can be set based on the display of computing power resources.

[0544] Optionally, the preset loop stop condition may include at least one of the following: 1) the number of loops is greater than a preset number of loops; 2) the difference between the new rendering and the current product style image is less than a preset difference. The specific method of quantifying the difference between the images is not limited in this embodiment.

[0545] In an optional embodiment, in the process of specifically fusing the product style image and the color information, the form of the color information is not limited. To facilitate the fusion, the color information can be converted into an image form.

[0546] This embodiment does not limit the method of converting the color matching information into an image format.

[0547] Optionally, taking the example that the color matching information only includes one color information, a pure color image filled with the color information can be generated for integration with the product style image.

[0548] Optionally, taking the regional color correspondence set as an example, the above method flow may further include: filling the regions in the product style image with corresponding colors based on the regional color correspondence set to obtain a color reference image; correspondingly, fusing the corresponding colors for the regions in the product style image based on the regional color correspondence set to obtain an intermediate rendering, specifically fusing the product style image and the color reference image to obtain the intermediate rendering. The filling may specifically be based on the outline of the region in the product style image, with the corresponding color.

[0549] This embodiment can combine color matching information in image form to improve the efficiency and accuracy of color fusion.

[0550] 3. S1103: Fusing the intermediate rendering and color matching information to obtain the final rendering.

[0551] The loss of color information in the final rendering may be smaller than the loss of color information in the intermediate rendering.

[0552] The present method does not limit the specific method of fusion. For details, please refer to the detailed explanation of S1102. Optionally, the final rendering can be a simulated rendering of the product manufactured according to the style in the product style diagram using the color matching information.

[0553] Since the color matching information is fused again based on the intermediate rendering, the color fusion effect in the final rendering can be improved and the color loss can be reduced.

[0554] Optionally, the fusion method can adopt layer operation or deep learning method.

[0555] The fusion method of S1103 may be the same as or different from the fusion method of S1102, and this method process is not limited thereto.

[0556] Optionally, the intermediate rendering and color matching information can be merged by layer operation to obtain the final rendering.

[0557] Optionally, a deep learning approach can be used to input the intermediate rendering and color matching information into a pre-trained model for fusion to obtain the final rendering.

[0558] This embodiment does not limit the specific model structure or training method. It is understandable that color fusion can be performed by training a neural network model or a generative model, specifically a generative model of image-to-image.

[0559] It can be understood that, optionally, the final rendering can be a product simulation rendering based on the product style in the product style diagram and combined with color matching information, and the product style information in the final rendering can be the same as the product style information in the product style diagram, and only the color matching used is replaced.

[0560] Optionally, the final rendering may retain the product style information from the product style image. This embodiment is not limited to specific product style information. Product style information may include, for example, the product's shape, pattern, wrinkles, and outline. The specific product color scheme may be combined or replaced with the color scheme information to produce a final rendering that retains the product style information but incorporates the new color scheme from the color scheme information.

[0561] Of course, optionally, when the product style image includes product style information, the product styles under different viewing angles can be simulated, and the color matching in the color matching information can be further integrated to obtain the final rendering.

[0562] It should be emphasized that the fusion in S1103 is mainly based on reducing the loss of color information in the intermediate renderings, that is, it is hoped to further improve the effect of color fusion. Therefore, the product style information in the final renderings also needs to be retained, mainly to optimize and improve the effect of color fusion.

[0563] In an optional embodiment, the final rendering may be obtained by fusing the intermediate rendering and the color matching information using a pre-trained model.

[0564] Optionally, the intermediate rendering and color matching information are fused to obtain the final rendering. Specifically, the intermediate rendering and color matching information can be input into a preset model to obtain the final rendering output by the preset model. The preset model can be used to reduce the loss of color information. The loss of color information in the final rendering can be less than the loss of color information in the intermediate rendering.

[0565] This embodiment does not limit the specific preset model structure and training method. Optionally, a generative model or a neural network model can be used. For the convenience of description, the model used to fuse the intermediate rendering and color matching information to obtain the final rendering is called the second fusion model.

[0566] The training method of the second fusion model can refer to the training method of the first fusion model. The second fusion model can be the same as the first fusion model or different from the first fusion model.

[0567] For ease of understanding, this embodiment can provide an embodiment of a training method. Optionally, images of different products obtained by using different color combinations to manufacture different product styles can be taken separately, and color matching information can be extracted as training samples for the above-mentioned second fusion model. Of course, it is also possible not to manufacture actual products, but to simulate different product styles in simulation software (for example, product style design software) by replacing multiple different color matching effects, and obtain images by screenshots or photography.

[0568] Of course, optionally, the input and output of the second fusion model may be different from those of the first fusion model, and the second fusion model may be used to reduce the loss of color matching information, and therefore, different training samples may be used.

[0569] Specifically, the loss of color information can be taken into consideration, and the intermediate rendering is already a rendering that integrates the color information. Therefore, for a product manufactured using one color scheme, an image can be taken and further blurred to lose some color information, and then a training sample can be constructed.

[0570] For example, for a product manufactured using a specified color scheme, you can take an original image and further blur or add noise to the colors in the original image to cause a loss of color information, thereby obtaining a processed image. The processed image and the color information of the specified color scheme can then be used as sample features, and the original image as a sample label to construct a training sample for training.

[0571] Therefore, optionally, a method for constructing a training sample set for the second fusion model may include: obtaining an original image of a product using a specified color scheme and obtaining color information for the specified color scheme; processing the original image of the product to increase the loss of the color information to obtain a product loss image; constructing a training sample using the product loss image and the color information for the specified color scheme as sample features and the original image of the product as a sample label, and adding the sample to the training sample set. The specified color scheme may be any color scheme.

[0572] The constructed training sample set can be used to train the second fusion model. This embodiment can improve the training effect of the second fusion model and improve the color fusion effect of the intermediate effect image.

[0573] This embodiment does not limit the method of obtaining the original product image, which can be obtained by direct photography or through software modeling. This embodiment also does not limit the method of obtaining the color information of the specified color matching, which can be obtained based on the original product image. The specific method can be referred to the explanation in S1101.

[0574] This embodiment does not limit the method of processing the original product image, as long as the color information can be lost. For example, blurring or noise addition can be used.

[0575] Of course, in addition to the above-mentioned training sample set construction method, other methods can also be used to construct the training sample set, as long as the loss of product image color information in the sample features is greater than the product image in the sample labels.

[0576] For example, a product original image can be processed multiple times to obtain multiple images with different color information loss. The images are sorted according to the amount of color information loss, and then training samples can be obtained by combining them.

[0577] For example, a product original image can be processed once to obtain a first loss image, and then processed again to obtain a second loss image, and so on. A training sample can then be obtained using the first loss image as the sample feature and the product original image as the sample label; a training sample can be obtained using the second loss image as the sample feature and the first loss image as the sample label; a training sample can be obtained using the second loss image as the sample feature and the product original image as the sample label; and so on.

[0578] This method does not limit the training method of the second fusion model. It can be trained based on the training sample set constructed above, or it can be trained using the same method as the first fusion model. It should be noted that the first fusion model can fuse colors to improve the color fusion effect and can also be used to reduce the loss of color information. Therefore, it can be used as the second fusion model to further fuse the intermediate rendering and color matching information.

[0579] In an optional embodiment, in the specific fusion process, in addition to the color matching information and the intermediate renderings that need to be fused, other information may also be fused. For details, please refer to the explanation in S1102.

[0580] This embodiment does not limit the specific content of other information. For example, the information of the area in the product style image where the colors need to be fused can specify the area in the product style image where the colors need to be fused; for example, the fusion method information can specify whether the fusion is performed by deep learning or layer operation.

[0581] In other words, optionally, the correspondence between the colors in the color matching information and the product style image area (or the intermediate rendering area), as well as the color matching method corresponding to the product style image area (or the intermediate rendering area) and other information can be further integrated. Specifically, color fusion can be guided by this information. Of course, this information may not be included in the color matching information, but can be determined in real time or randomly. The specific determination method can be explained above. Since the product style information between the intermediate rendering and the product style image is the same, the identified areas can also be the same.

[0582] This embodiment does not limit the specific form of other information, which may be text, image, vector, etc.

[0583] Optionally, the above method flow may further include: obtaining atmosphere information corresponding to the color matching information; fusing the intermediate rendering and the color matching information to obtain the final rendering, specifically: fusing the obtained atmosphere information, the intermediate rendering and the color matching information to obtain the final rendering. For specific explanations, please refer to the above. The atmosphere information may be extracted or searched based on the color matching information, and the specific atmosphere information may be keywords such as "warm", "lively", "calm", etc. that characterize the atmosphere. The effect of color fusion can be improved by combining atmosphere information. This embodiment does not limit the specific way of fusing atmosphere information. Optionally, taking the generation model as an example, the intermediate rendering and the color matching information may be fused with the atmosphere information as a guide to obtain the final rendering.

[0584] In an optional embodiment, based on the intermediate rendering and color matching information, the loss of color information can be reduced by fusion. In order to further reduce the loss of color information and improve the color fusion effect, more than one fusion can be performed.

[0585] Therefore, optionally, the fusion between the intermediate rendering and the color matching information may be performed once or multiple times to obtain the final rendering.

[0586] The present method is not limited to a specific fusion process, and may be implemented using a cyclic process.

[0587] Optionally, the intermediate rendering and color matching information are fused to obtain the final rendering, which can be specifically: looping the following steps until a preset loop stop condition is met: fusing the current intermediate rendering and color matching information to obtain a new rendering; determining the obtained new rendering as the current intermediate rendering; after the loop stops, determining the current intermediate rendering as the final rendering.

[0588] It is understood that the fusion method for each cycle can be explained above. For example, a preset model can be input for fusion. Specifically, it can be a generative model, such as a diffusion model.

[0589] This embodiment does not limit the preset loop stop condition. Specifically, it can be set to the maximum number of loops, or to determine whether the difference between the new rendering and the current intermediate rendering is small enough, or it can be set based on the display of computing resources.

[0590] Optionally, the preset loop stop condition may include at least one of the following: 1) the number of loops is greater than a preset number of loops; 2) the difference between the new rendering and the current intermediate rendering is less than a preset difference; 3) the loss of color information in the current intermediate rendering is less than a preset loss, etc. The specific method of quantifying the difference between images and the method of quantifying the loss of color information are not limited in this embodiment.

[0591] In an optional embodiment, during the specific fusion of the intermediate rendering and the color information, the form of the color information is not limited. To facilitate fusion, the color information can be converted into an image form.

[0592] This embodiment does not limit the method of converting the color matching information into an image format.

[0593] Optionally, taking the example that the color matching information only includes one color information, a pure color image filled with the color information can be generated, which can be used for fusion with the intermediate effect image.

[0594] Optionally, taking the regional color correspondence set as an example, the above method flow may further include: filling the regions in the product style image with corresponding colors based on the regional color correspondence set to obtain a color reference image; correspondingly, fusing the corresponding colors for the regions in the intermediate rendering image based on the regional color correspondence set to obtain a final rendering, specifically fusing the intermediate rendering image with the color reference image to obtain the final rendering. The filling may specifically be based on the outline of the region in the product style image, filling with the corresponding color.

[0595] This embodiment can combine color matching information in image form to improve the efficiency and accuracy of color fusion.

[0596] For easier understanding, the present specification also provides figures to explain the process of generating a color reference image. As shown in FIG13 , FIG13 is a schematic diagram of a method for generating a color reference image provided by an exemplary embodiment.

[0597] Taking clothing as an example, we can generate a color reference map based on the correspondence between areas and colors in the color matching information, that is, the sleeve and collar areas are black, and the main body area is dark gray, and further combine it with the areas in the product style diagram. Specifically, we can fill in the corresponding colors based on the outlines of the areas in the product style diagram.

[0598] It should be noted that only black, white and gray are used as examples here, and a variety of colors can actually be used for color matching, such as green, red, purple, etc.

[0599] For easier understanding, the present specification also provides drawings to explain the overall fusion process. As shown in FIG14 , FIG14 is a schematic diagram of the principle of a clothing color matching method provided by an exemplary embodiment.

[0600] Taking clothing as an example, the color matching information and the clothing style diagram can be integrated to obtain an intermediate clothing rendering, that is, a simulated rendering of a clothing product manufactured according to the color matching in the color matching information and the style in the clothing style diagram.

[0601] Afterwards, the outline of the clothing area in the clothing style image can be obtained, and the corresponding color can be further filled to obtain a color reference image.

[0602] The color reference image and the intermediate clothing rendering are further integrated to obtain the final clothing rendering.

[0603] The loss of color information in the final garment rendering can be less than that in the intermediate garment rendering, which can be specifically reflected in the fineness of the colors. The final garment rendering is superior to the intermediate garment rendering in terms of fineness of color.

[0604] The embodiments of this specification also provide another embodiment of the fabric fusion method using clothing products as an example.

[0605] As shown in FIG15 , FIG15 is a flow chart of another product color matching method provided by an exemplary embodiment.

[0606] The embodiments of this specification do not limit the execution subject of this method.

[0607] Optionally, the method can be applied to a computing device or a module in the device. For example, the method can be applied to a server, a terminal, or other devices.

[0608] The method may include the following steps.

[0609] S2201: Acquire clothing style pictures and color matching information; the color matching information includes at least one color information.

[0610] S2202: Integrate the clothing style image and color matching information to obtain an intermediate clothing effect image.

[0611] S2203: Fusing the intermediate clothing rendering and color matching information to obtain the final clothing rendering.

[0612] The loss of color information in the final clothing rendering may be smaller than the loss of color information in the intermediate clothing rendering.

[0613] The above method process can obtain clothing effect pictures by automatically fusing images, thereby reducing manual participation and improving the efficiency and real-time performance of viewing clothing product effects.

[0614] In addition, the above method process can also improve the color expression effect in the final clothing rendering by fusing color matching information with the intermediate clothing rendering, and reduce the loss of color information during the fusion process.

[0615] For a detailed explanation of the process of this method, please refer to the explanation of S1101-S1103.

[0616] Optionally, the intermediate clothing rendering and the color matching information are fused to obtain the final clothing rendering, which can be specifically: the intermediate clothing rendering and the color matching information are input into a preset model to obtain the final clothing rendering output by the preset model; the preset model is used to reduce the loss of color information; the loss of color information in the final clothing rendering is less than the loss of color information in the intermediate clothing rendering.

[0617] Optionally, the above method process may also include: obtaining atmosphere information corresponding to the color matching information; fusing the intermediate clothing rendering and the color matching information to obtain a final clothing rendering, including: fusing the obtained atmosphere information, the intermediate clothing rendering and the color matching information to obtain a final clothing rendering.

[0618] Optionally, the fusing of the intermediate clothing rendering and the color matching information to obtain the final clothing rendering includes: looping the following steps until a preset loop stop condition is met: fusing the current intermediate clothing rendering and the color matching information to obtain a new clothing rendering; determining the obtained new clothing rendering as the current intermediate clothing rendering; after the loop stops, determining the current intermediate clothing rendering as the final clothing rendering.

[0619] Optionally, the color matching information includes: a set of regional color matching correspondences; the regional color matching correspondence set includes at least one set of correspondences between colors and regions in the clothing style diagram; the fusing of the clothing style diagram and the color matching information to obtain an intermediate clothing rendering includes: according to the set of regional color matching correspondences, for the regions in the clothing style diagram, fusing corresponding colors to obtain an intermediate clothing rendering; the fusing of the intermediate clothing rendering and the color matching information to obtain a final clothing rendering includes: according to the set of regional color matching correspondences, for the regions in the intermediate clothing rendering, fusing corresponding colors to obtain a final clothing rendering.

[0620] Optionally, the above method process may also include: filling the corresponding colors for the areas in the clothing style diagram according to the regional color matching correspondence set to obtain a color reference diagram; according to the regional color matching correspondence set, fusing the corresponding colors for the areas in the intermediate clothing rendering diagram to obtain a final clothing rendering, including: fusing the intermediate clothing rendering and the color reference diagram to obtain a final clothing rendering.

[0621] Optionally, obtaining a clothing style diagram includes: obtaining an initial clothing image; performing image recognition on the initial clothing image to determine clothing image content in the initial clothing image; and determining a clothing style diagram based on the clothing image content recognized in the initial clothing image.

[0622] Optionally, a method for determining the regional color matching relationship set includes any one of the following:

[0623] 1) Obtaining colors specified for regions in the clothing style diagram; constructing at least one set of correspondences between colors and regions in the clothing style diagram based on the specified colors, and obtaining a set of regional color matching correspondences.

[0624] 2) Obtaining atmosphere cues specified for the regions in the clothing style diagram, and automatically generating corresponding colors based on the obtained atmosphere cues; constructing at least one set of correspondences between colors and regions in the clothing style diagram based on the generated colors, and obtaining a set of regional color matching correspondences.

[0625] 3) Obtaining a color scheme; for the region in the clothing style diagram, randomly selecting a color from the obtained color scheme as the corresponding color, constructing at least one set of correspondences between the color and the region in the clothing style diagram, and obtaining a set of regional color correspondences.

[0626] For a detailed explanation of the process of this method, please refer to the explanation of S1101-S1103.

[0627] The following provides an application embodiment taking clothing products as an example.

[0628] Color matching plays a key role in fashion and other design fields, influencing people's perception and emotional experience of objects. In style design, clever color matching can enhance the design theme, highlight details, and convey a specific emotion or atmosphere. Common color matching principles include contrasting colors, analogous colors, monotone colors, complementary colors, triadic colors, and quasi-monochromatic colors.

[0629] In actual design, designers will choose the appropriate color combination based on factors such as the brand, target audience, and design theme. In addition, considering the season, fashion trends, and cultural differences are also factors that need to be taken into account in the design to ensure that the color combination can keep pace with the times and attract the target audience.

[0630] In fashion design, a color palette refers to multiple color schemes for a single style. Designers often don't just create a single color palette for a particular style; instead, they offer a comprehensive suite of color combinations to meet the aesthetic needs and fashion trends of different consumers. This collection of multiple color schemes is called a color palette.

[0631] Each color palette may share certain commonalities, such as similar hues, brightness, or saturation, to maintain a cohesive design style. However, these color palettes may also differ in certain ways to meet the individual needs of different groups, adapt to seasonal changes, or follow the evolution of fashion trends.

[0632] Color-matching design emphasizes not only the coordination of individual colors but also the overall effect of different colors. This design strategy can enrich product lines, increase customer choice, and help brands better adapt to diverse consumer tastes in the market.

[0633] Currently, designers consider a variety of factors when creating color combinations and series color matching designs, including occasions, target audiences, brands, etc. These designs are carefully matched manually by designers, and in order to display the styles, these designs need to be made into actual products.

[0634] The advantages of manually completing color matching and series color matching by designers are as follows: more creativity and flexibility. Manual color matching is full of creativity. Designers can match colors according to their personal aesthetic feelings and intuition, and flexibly respond to different design challenges; more personalized. Manual matching can better reflect the designer's personality and style, inject uniqueness into the design, and enhance product differentiation; more thoughtful. Designers can use their professional knowledge to take into account factors such as brand image, seasonal changes, and target audiences to develop color schemes that better meet actual needs.

[0635] However, there are also many problems, such as subjectivity and individual differences. Manual color matching is subjective to a certain extent, and the aesthetics and tastes of different designers vary, which may lead to a certain degree of inconsistency; time and cost. Manual color matching is relatively time-consuming, especially when a large number of color schemes need to be produced, which may increase the time and cost of design; limitations. Designers have limited experience and vision, and may be constrained by individual experience and limitations, and cannot fully consider certain potential innovations and changes; it is difficult to display style effects. After the style is completed with color matching and the series is in harmony, the finished product needs to be produced for display, which requires a lot of cost investment.

[0636] The generative model-based intelligent color matching method proposed in this embodiment can reduce product variations caused by subjective aesthetic differences among designers, while also saving design time and costs while allowing designers to focus more on innovative design of the style itself. Furthermore, this method is not limited by the individual experience of different designers, providing greater possibilities for product design. Most importantly, the application of generative models can create more potentially innovative color schemes while enabling the generation of style display images, effectively reducing significant costs during the actual production phase.

[0637] This embodiment addresses the aforementioned issues by proposing a color matching system based on intelligent color matching and a generative model. This solution primarily comprises four modules: a color preprocessing module, a style parsing module, a fusion module, and a generation module. The color matching input module accepts two input formats: a specified color series and color matching keywords; the style input module similarly accepts two input formats: a style image and a model image.

[0638] As shown in FIG16 , FIG16 is a schematic diagram showing the principle of a color matching system provided by an exemplary embodiment.

[0639] This includes the color preprocessing module, style parsing module, fusion module, and generation module. The detailed process is explained below.

[0640] Step 1: Color preprocessing module.

[0641] As shown in FIG17 , FIG17 is a schematic diagram of the principle of a color preprocessing module provided by an exemplary embodiment.

[0642] Users can specify their desired color palette in two ways: directly by providing a color list, or by using emotionally expressive color keywords, such as "energized," "calm," and "vital." If the user chooses to enter a color keyword, the system will use a pre-trained color generation model to generate a specific color palette based on the user's provided keywords. This generated color palette, based on the user's input keywords, effectively conveys the desired emotion and message.

[0643] Users can directly provide a customized color palette to ensure the desired color scheme meets their specific requirements and tastes. This approach is suitable for users who have already clearly defined their desired colors. If users prefer to express emotions rather than specify colors, they can use color-related emotional keywords such as vitality, calmness, and vitality. This method is more flexible, allowing users to guide color generation through emotional keywords without having to deeply understand specific color combinations. The pre-trained color generation model can automatically generate corresponding color palettes based on the emotional keywords entered by the user. It has a certain degree of intelligence and creativity, and can better meet user expectations.

[0644] The user-specified color palette, or the one generated by entering color keywords, is then fed into the color encoding model. In this step, the color encoding model generates an encoding vector that characterizes the emotional atmosphere of the color palette. This encoding vector serves as an input control condition for the subsequent generation model, precisely controlling the desired emotional characteristics of the generated image.

[0645] The color encoding model is responsible for converting a series of color schemes specified or generated by the user into a specific encoding vector. This model not only considers the visual effects of colors, but also focuses on their potential emotional and atmospheric characteristics.

[0646] Step 2: Style parsing module.

[0647] The input for a style image can be a pure style image or a model wearing the style image. Considering that users may need to see how the style will look in a real-life scenario, the user can directly input a model wearing the style image, which effectively reduces the cost investment in the actual production stage.

[0648] The input style image or model image will be parsed by the parsing model. The function of the parsing model is to parse and partition the various areas of the style image. For example, the collar, sleeves, cuffs, pockets and other parts of the clothing will be accurately parsed into separate areas, providing detailed information for subsequent color matching for different areas.

[0649] In addition to identifying individual regions, the parsing model also helps the network model understand their composition, providing a better foundation for subsequent color matching. Understanding and parsing style images at a higher level of detail improves the system's understanding of style structure, allowing for more personalized and customized color matching.

[0650] The parsing model can be a model divided by plates, or a model divided by trained random reasonable blocks. The model can be an algorithm such as region segmentation and region detection.

[0651] Step 3: Fusion modules.

[0652] As shown in FIG18 , FIG18 is a schematic diagram of the principle of a fusion module provided by an exemplary embodiment.

[0653] The fusion module's input consists of two parts: the color palette specified or generated in the color preprocessing module, and the style analysis diagram and designated block color scheme (optional) in the style analysis module. When specifying block color schemes, users can clearly define the color composition of each area; when not specifying block color schemes, users have a wider range of color combinations, and the system will generate a color scheme based on the overall color scheme.

[0654] When necessary, users can choose to specify the color scheme of each area to clarify the color composition of each area. This provides precise control over local details and is suitable for users who want to emphasize specific design elements. Without specifying the color input for the corresponding block, the system will obtain a larger color matching space and can more flexibly generate color matching according to the overall design requirements. This provides a greater space for creativity, making the generated results more diverse and unique. This design aims to balance the user's need for fine control over color matching and the expectation of creative space, so that the system can meet the user's personalized needs while maintaining sufficient flexibility.

[0655] The designated area color matching is the area color matching correspondence relationship set in the above method embodiment.

[0656] The color assignment module is responsible for skillfully matching each masked area in the input style parsing diagram with a series of color schemes. If the user provides a specific block color scheme, the module will assign it according to the user's specified rules; if this input is not specified, the model will use a reasonable random method to match colors, ensuring that each area in the parsed style diagram presents a reasonable color space. The generated reference image will serve as the reference input for the final generative model, ensuring that the output of the generative model will not have significant color deviations in each area.

[0657] Specifically, a machine learning model can be used to automatically integrate the specified color scheme and style analysis chart. Specifically, the model can automatically determine the color scheme corresponding to the style area and the specific method of color matching.

[0658] The color assignment module sets matching rules based on the user-entered color palette and the designated area color scheme. This helps ensure that the user's specific requirements for certain areas are fully met. If the user does not provide a designated area color scheme, the module will be flexible and use the model to randomly assign colors. This ensures diversity in the generated results while ensuring that the colors of each area appear in a reasonable overall space. Through the assignment module, each area in the generated style analysis diagram will have a reasonable color space. This helps the generation model more accurately maintain the user's desired overall color tone and matching in the subsequent process.

[0659] The color fusion module also performs regular or random color fusion depending on whether the user inputs a specific block color scheme. Unlike the color assignment module, the color fusion module will maximize the preservation of key information such as the style image's folds, shape, and outline during the fusion process. Complementing the reference image generated by the color assignment module, the fusion module generates a fusion image as input to the final generation model to ensure that the output does not deviate from the original style's shape, outline, and other important information while matching colors. This helps ensure that the generated style is more consistent with the user's desired design style and does not deviate from the overall appearance of the original style.

[0660] Step 4: Generate module.

[0661] The reference image and fused image generated in the fusion module serve as two inputs to the generative model, guiding the model to produce the final result image. The fused image ensures that key information of the original style, such as shape, folds, and outline, is preserved during the generation process; while the reference image is used to maximize the reproduction of the color matching. By comprehensively considering these two inputs, as well as possible randomness, the generative model produces a final result that retains key information of the style while also achieving a reasonable color combination. This process aims to ensure that the generated style is both similar in appearance to the original style and has a reasonable and aesthetically pleasing color combination, so that the final result can meet the user's comprehensive design requirements.

[0662] In addition, the color coding vector encoded by the color coding model in the color preprocessing module also serves as one of the guiding conditions for the generative model. This vector, which carries the encoded color information, helps the generative model more accurately grasp the desired emotional atmosphere. There are profound connections between different colors and different emotions, such as vibrant red, peaceful and tranquil blue, and vibrant green. The overall color tone of the style also profoundly affects the desired emotional effect. Soft and warm tones can create a comfortable and warm atmosphere, while bright and vivid tones can create an atmosphere of vitality and passion.

[0663] The introduction of color-coded vectors enables the generative model to create diverse emotional effects in a more targeted manner, thereby meeting the personalized needs of different users for design works. The generative model can more intelligently convey the emotional atmosphere desired by users in the design.

[0664] The generation module can accept more control conditions, such as gradient color guidance information.

[0665] This solution can effectively achieve the following results:

[0666] 1. Through the above process, this embodiment enables intelligent color matching for styles and colors, consistent color matching for series, and ultimately, the resulting display images. For brand manufacturers, the generative model intelligent color matching method, driven by algorithms and data, can mitigate subjective aesthetic differences between designers, ensure consistency in the base color of products, and enhance the stability of the overall brand image. Furthermore, it is not subject to the individual experience and limitations of individual designers, making it more objective. This means that differences in experience levels no longer affect product color design, ensuring a wider range of design possibilities. The application of generative models expands the possibilities of color design, enabling more creative and potentially innovative color schemes. This helps brands innovate and attract a wider audience. In addition to creating more innovative color combinations, it also enables the automatic generation of style display images, reducing significant costs during the actual production phase and improving efficiency and controllability. For designers, automated color matching reduces the time and cost of manual matching, allowing them to complete tasks more efficiently, helping to accelerate product launches and reduce production costs during the design phase. It also allows them to focus more on the innovative design of the style itself, thereby enhancing the uniqueness and market appeal of the design.

[0667] Smart color matching and series color matching can help brand manufacturers reduce a lot of costs in the actual production stage. Without having to produce all the products as finished products, they can preview all color matching and the effect of the products with the same color matching in advance, which improves the efficiency and controllability of production. At the same time, it can help designers complete tasks more efficiently, speed up the time to market of products, reduce the production costs in the design stage, and focus more attention on the innovative design of the style itself.

[0668] 2. Through the color generation model and color encoding module, a given color keyword is generated into a specific color series. These colors are then encoded to form a corresponding encoding vector. This vector represents the emotional atmosphere of the color. For example, red is encoded as a vector representing passion and energy, while green is encoded as a vector representing vitality. In the generative model, these encoding vectors serve as conditional inputs, imbuing the resulting image with a specific emotional atmosphere, more accurately conveying and expressing the desired color sensation.

[0669] The vector encoded by the color coding model serves as one of the control conditions of the generative model and plays a certain guiding role in the generative model, which can more accurately enable the generated result image to convey the emotional atmosphere information that the user wants to convey.

[0670] 3. The Color Assignment Module and the Color Fusion Module take the generated color series and the parsed style map as input, generating and outputting a block-based color fusion reference map and a fusion map of the style and color of each block. These serve as conditional inputs for the subsequent generation model. The fusion map retains key information such as the original outline, texture, and lighting of the style map, providing a higher-quality initial condition for subsequent generation. The reference map maximizes the preservation of the original colors of the generated color series, ensuring that the final output style color matching result map is consistent with the expressed color keywords. This integrated process aims to improve the accuracy of the generated results and the degree to which they meet design expectations.

[0671] The reference image and fusion image output by the fusion module can maintain the restoration of color matching while retaining the shape, outline, light and shadow and other information of the original style to the greatest extent, so that the final result image meets the original expectations.

[0672] The embodiments of this specification also provide corresponding device embodiments.

[0673] As shown in FIG19 , FIG19 is a structural diagram of a product color matching device provided by an exemplary embodiment.

[0674] The apparatus may include the following units.

[0675] The acquisition unit 3301 is used to acquire a product style image and color matching information; the color matching information includes at least one color information;

[0676] A first fusion unit 3302 is configured to fuse the product style image and the color matching information to obtain an intermediate rendering;

[0677] The second fusion unit 3303 is used to fuse the intermediate rendering and the color matching information to obtain a final rendering.

[0678] Optionally, the second fusion unit 3303 is configured to:

[0679] The intermediate rendering and the color matching information are input into a preset model to obtain a final rendering output by the preset model; the preset model is used to reduce the loss of color information; the loss of color information in the final rendering is less than the loss of color information in the intermediate rendering.

[0680] Optionally, the acquisition unit 3301 is further configured to: acquire atmosphere information corresponding to the color matching information; and the second fusion unit 3303 is configured to: fuse the acquired atmosphere information, the intermediate rendering and the color matching information to obtain a final rendering.

[0681] Optionally, the second fusion unit 3303 is configured to:

[0682] The following steps are executed in a loop until a preset loop stop condition is met: fusing the current intermediate rendering and the color matching information to obtain a new rendering; determining the obtained new rendering as the current intermediate rendering;

[0683] After the loop stops, the current intermediate rendering is determined as the final rendering.

[0684] Optionally, the color information includes: a set of regional color correspondences; the set of regional color correspondences includes at least one set of correspondences between colors and regions in the product style diagram; the first fusion unit 3302 is used to: according to the set of regional color correspondences, for the regions in the product style diagram, fuse the corresponding colors to obtain an intermediate rendering; the second fusion unit 3303 is used to: according to the set of regional color correspondences, for the regions in the intermediate rendering, fuse the corresponding colors to obtain a final rendering.

[0685] Optionally, the device also includes a reference unit 304, which is used to: fill the corresponding colors for the areas in the product style image according to the regional color matching correspondence set to obtain a color reference image; the second fusion unit 3303 is used to: fuse the intermediate rendering and the color reference image to obtain a final rendering.

[0686] Optionally, the acquisition unit 3301 is used to: acquire an initial product image; perform image recognition on the initial product image to determine product image content in the initial product image; and determine a product style image based on the product image content recognized in the initial product image.

[0687] Optionally, the method for determining the regional color matching correspondence set includes any one of the following:

[0688] Obtaining a color specified for a region in the product style diagram; constructing at least one set of correspondences between colors and regions in the product style diagram based on the specified colors to obtain a set of regional color matching correspondences;

[0689] Obtaining an atmosphere cue specified for a region in the product style diagram, and automatically generating corresponding colors based on the obtained atmosphere cue; constructing at least one set of correspondences between the colors and the regions in the product style diagram based on the generated colors, to obtain a set of regional color matching correspondences;

[0690] Obtain a color scheme; for the area in the product style diagram, randomly select a color from the obtained color scheme as the corresponding color, construct at least one set of correspondences between the color and the area in the product style diagram, and obtain a set of regional color correspondences.

[0691] The detailed explanation of the above device embodiment can be found in the explanation of the method embodiment.

[0692] As shown in FIG. 20 , FIG. 20 is a schematic structural diagram of another product color matching device provided by an exemplary embodiment.

[0693] The apparatus may include the following units.

[0694] The clothing acquisition unit 4401 is used to acquire clothing style images and color matching information; the color matching information includes at least one color information;

[0695] The third fusion unit 4402 is configured to fuse the clothing style image and the color matching information to obtain an intermediate clothing effect image;

[0696] The fourth fusion unit 4403 is configured to fuse the intermediate clothing rendering and the color matching information to obtain a final clothing rendering.

[0697] Optionally, the fourth fusion unit 4403 is used to: input the intermediate clothing rendering and the color matching information into a preset model to obtain a final clothing rendering output by the preset model; the preset model is used to reduce the loss of color information; the loss of color information in the final clothing rendering is less than the loss of color information in the intermediate clothing rendering.

[0698] Optionally, the clothing acquisition unit 4401 is further used to: obtain atmosphere information corresponding to the color matching information; the fourth fusion unit 4403 is used to: fuse the acquired atmosphere information, the intermediate clothing rendering and the color matching information to obtain a final clothing rendering.

[0699] Optionally, the fourth fusion unit 4403 is used to: loop through the following steps until a preset loop stop condition is met: fuse the current intermediate clothing rendering and the color matching information to obtain a new clothing rendering; determine the obtained new clothing rendering as the current intermediate clothing rendering; after the loop stops, determine the current intermediate clothing rendering as the final clothing rendering.

[0700] Optionally, the color matching information includes: a set of regional color matching correspondences; the set of regional color matching correspondences includes at least one set of correspondences between colors and regions in the clothing style diagram; a third fusion unit 4402 is used to: according to the set of regional color matching correspondences, for the regions in the clothing style diagram, fuse corresponding colors to obtain an intermediate clothing rendering; a fourth fusion unit 4403 is used to: according to the set of regional color matching correspondences, for the regions in the intermediate clothing rendering, fuse corresponding colors to obtain a final clothing rendering.

[0701] Optionally, the above-mentioned device may also include a clothing reference unit 404 for: filling the corresponding colors for the areas in the clothing style diagram according to the regional color matching correspondence set to obtain a color reference diagram; a fourth fusion unit 4403 is used for: fusing the intermediate clothing effect diagram and the color reference diagram to obtain a final clothing effect diagram.

[0702] Optionally, the clothing acquisition unit 4401 is used to: acquire an initial clothing image; perform image recognition on the initial clothing image to determine clothing image content in the initial clothing image; and determine a clothing style diagram based on the clothing image content recognized in the initial clothing image.

[0703] Optionally, a method for determining the regional color matching relationship set includes any one of the following:

[0704] 1) Obtaining colors specified for regions in the clothing style diagram; constructing at least one set of correspondences between colors and regions in the clothing style diagram based on the specified colors, and obtaining a set of regional color matching correspondences.

[0705] 2) Obtaining atmosphere cues specified for the regions in the clothing style diagram, and automatically generating corresponding colors based on the obtained atmosphere cues; constructing at least one set of correspondences between colors and regions in the clothing style diagram based on the generated colors, and obtaining a set of regional color matching correspondences.

[0706] 3) Obtaining a color scheme; for the region in the clothing style diagram, randomly selecting a color from the obtained color scheme as the corresponding color, constructing at least one set of correspondences between the color and the region in the clothing style diagram, and obtaining a set of regional color correspondences.

[0707] The detailed explanation of the above device embodiment can be found in the explanation of the method embodiment.

[0708] The various technical features in the above embodiments can be arbitrarily combined as long as there is no conflict or contradiction between the combinations of features. However, due to space limitations, they are not described one by one. Therefore, the arbitrary combination of the various technical features in the above embodiments also falls within the scope of the present invention.

[0709] An embodiment of this specification also provides an electronic device, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor implements any of the above method embodiments by running the executable instructions.

[0710] FIG21 is a schematic diagram of the structure of a device provided by an exemplary embodiment. Referring to FIG21 , at the hardware level, the device includes a processor 702, an internal bus 704, a network interface 706, a memory 708, and a non-volatile memory 710, and may also include hardware required for other services. One or more embodiments of this specification may be implemented based on software, such as the processor 702 reading the corresponding computer program from the non-volatile memory 710 into the memory 708 and then running it. Of course, in addition to software implementation, one or more embodiments of this specification do not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is, the execution subject of the following processing flow is not limited to each logic unit, but may also be hardware or logic devices.

[0711] The systems, devices, modules, or units described in the above embodiments may be implemented by computer chips or entities, or by products having certain functions. A typical implementation device is a computer, which may be in the form of a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email transceiver, game console, tablet computer, wearable device, or any combination of these devices.

[0712] In a typical configuration, a computer includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0713] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0714] The embodiments of this specification also provide a computer-readable storage medium having computer instructions stored thereon, which implement the steps of any of the above method embodiments when executed by a processor.

[0715] Computer-readable media include permanent and non-permanent, removable and non-removable media that can be used to store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, disk storage, quantum memory, graphene-based storage media or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0716] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0717] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0718] The terms used in one or more embodiments of this specification are for the purpose of describing specific embodiments only and are not intended to limit one or more embodiments of this specification. The singular forms "a," "an," "the," and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0719] It should be understood that although the terms first, second, third, etc. may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when..." or "when..." or "in response to determining."

[0720] The above description is merely a preferred embodiment of one or more embodiments of this specification and is not intended to limit one or more embodiments of this specification. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of this specification shall be included in the scope of protection of one or more embodiments of this specification.

Claims

1. A fabric fusion method, characterized in that, Including: Obtain a standard fabric diagram and a product style diagram; Fuse the product style diagram and the standard fabric diagram to obtain an intermediate effect diagram; Determine the image content in the standard fabric diagram for fusing with the product style diagram as the fabric diagram to be fused; Fuse the intermediate effect diagram and the fabric diagram to be fused to obtain a final effect diagram.

2. The method according to claim 1, wherein The step of fusing the intermediate effect diagram and the fabric diagram to be fused to obtain a final effect diagram includes: Input the intermediate effect diagram and the fabric diagram to be fused into a preset model to obtain the final effect diagram output by the preset model; the preset model is used to reduce the loss of fabric information; the loss of fabric information in the final effect diagram is less than the loss of fabric information in the intermediate effect diagram.

3. The method according to claim 1, wherein It also includes: Obtain the target fabric information of the standard fabric diagram; The step of fusing the intermediate effect diagram and the fabric diagram to be fused to obtain a final effect diagram includes: Fuse the obtained target fabric information, the intermediate effect diagram and the fabric diagram to be fused to obtain a final effect diagram.

4. The method according to claim 1, characterized in that The step of fusing the intermediate effect diagram and the fabric diagram to be fused to obtain a final effect diagram includes: Loop and execute the following steps until a preset loop stop condition is met: fuse the current intermediate effect diagram and the fabric diagram to be fused to obtain a new effect diagram; determine the obtained new effect diagram as the current intermediate effect diagram; After the loop stops, determine the current intermediate effect diagram as the final effect diagram.

5. The method according to claim 1, characterized in that The step of determining the image content in the standard fabric diagram for fusing with the product style diagram as the fabric diagram to be fused includes: According to the intermediate effect diagram, determine the image content in the standard fabric diagram for fusing with the product style diagram as the fabric diagram to be fused; or Superimpose the product contour in the product style diagram on the standard fabric diagram in a preset manner, and determine the image content included in the standard fabric diagram by the product contour as the fabric diagram to be fused.

6. The method according to claim 1, wherein The step of obtaining the standard fabric diagram includes: Obtain an initial fabric diagram; Perform preprocessing on the initial fabric diagram to obtain a standard fabric diagram.

7. The method according to claim 1, characterized in that The obtaining method of the product style diagram includes: Obtain an initial product image; Perform image recognition on the initial product image to determine the product image content in the initial product image; Determine the product style diagram according to the product image content recognized in the initial product image.

8. The method according to claim 1, wherein It also includes: Obtain target parameters; The step of obtaining the standard fabric diagram includes: obtain an initial fabric diagram; determine the repeated fabric diagram content in the initial fabric diagram based on the target parameters; Stitch several copies of the repeated fabric diagram content to obtain a standard fabric diagram.

9. The method according to claim 1, characterized in that, It also includes: Obtain target parameters; The step of determining the image content in the standard fabric diagram for fusing with the product style diagram as the fabric diagram to be fused includes: Superimpose the product contour in the product style diagram on the standard fabric diagram according to the relative position in the target parameters, and determine the image content included in the standard fabric diagram by the product contour as the fabric diagram to be fused.

10. A fabric fusion method, characterized in that, Including: Obtain a standard fabric diagram and a clothing style diagram; Fuse the clothing style diagram and the standard fabric diagram to obtain an intermediate clothing production effect diagram; Determine the image content in the standard fabric diagram for fusion with the clothing style diagram as the fabric diagram to be fused; Fuse the intermediate clothing effect diagram and the fabric diagram to be fused to obtain the final clothing effect diagram.

11. A fabric fusion device, characterized in that, Including: An acquisition unit for acquiring a standard fabric diagram and a product style diagram; A first fusion unit for fusing the product style diagram and the standard fabric diagram to obtain an intermediate effect diagram; A determination unit for determining the image content in the standard fabric diagram for fusion with the product style diagram as the fabric diagram to be fused; A second fusion unit for fusing the intermediate effect diagram and the fabric diagram to be fused to obtain the final effect diagram.

12. A fabric fusion device, characterized in that, Including: A diagram acquisition unit for acquiring a standard fabric diagram and a clothing style diagram; A third fusion unit for fusing the clothing style diagram and the standard fabric diagram to obtain an intermediate clothing effect diagram; A fabric diagram to be fused unit for determining the image content in the standard fabric diagram for fusion with the clothing style diagram as the fabric diagram to be fused; A fourth fusion unit for fusing the intermediate clothing effect diagram and the fabric diagram to be fused to obtain the final clothing effect diagram.

13. A product color matching method, characterized in that, Including: Acquire a product style diagram and color matching information; The color matching information includes at least one color information; Fuse the product style diagram and the color matching information to obtain an intermediate effect diagram; Fuse the intermediate effect diagram and the color matching information to obtain the final effect diagram.

14. The method according to claim 13, wherein The fusing the intermediate effect diagram and the color matching information to obtain the final effect diagram includes: Input the intermediate effect diagram and the color matching information into a preset model to obtain the final effect diagram output by the preset model; the preset model is used to reduce the loss of color information; the loss of color information in the final effect diagram is less than the loss of color information in the intermediate effect diagram.

15. The method according to claim 13, wherein The method further includes: acquiring the atmosphere information corresponding to the color matching information; The fusing the intermediate effect diagram and the color matching information to obtain the final effect diagram includes: fusing the acquired atmosphere information, the intermediate effect diagram and the color matching information to obtain the final effect diagram.

16. The method according to claim 13, characterized in that, The fusing the intermediate effect diagram and the color matching information to obtain the final effect diagram includes: Loop and execute the following steps until a preset loop stop condition is met: fuse the current intermediate effect diagram and the color matching information to obtain a new effect diagram; determine the obtained new effect diagram as the current intermediate effect diagram; After the loop stops, determine the current intermediate effect diagram as the final effect diagram.

17. The method according to claim 13, characterized in that The color matching information includes: a set of regional color matching correspondence relationships; the set of regional color matching correspondence relationships contains at least one set of correspondence relationships between colors and regions in the product style diagram; The fusing the product style diagram and the color matching information to obtain an intermediate effect diagram includes: according to the set of regional color matching correspondence relationships, fuse the corresponding colors for the regions in the product style diagram to obtain an intermediate effect diagram; The fusing the intermediate effect diagram and the color matching information to obtain the final effect diagram includes: according to the set of regional color matching correspondence relationships, fuse the corresponding colors for the regions in the intermediate effect diagram to obtain the final effect diagram.

18. The method according to claim 17, characterized in that, The method further includes: filling corresponding colors for the regions in the product style drawing according to the set of region color matching correspondences to obtain a color reference drawing; The step of, according to the set of region color matching correspondences, fusing corresponding colors for the regions in the intermediate effect drawing to obtain a final effect drawing includes: Fusing the intermediate effect drawing and the color reference drawing to obtain a final effect drawing.

19. The method according to claim 13, characterized in that, The step of obtaining the product style drawing includes: Obtaining an initial product image; Performing image recognition on the initial product image to determine the product image content in the initial product image; Determining the product style drawing according to the product image content recognized in the initial product image.

20. The method according to claim 17, wherein The determining method of the set of region color matching correspondences includes any one of the following: Obtaining the colors specified for the regions in the product style drawing; based on the specified colors, constructing at least one set of correspondences between the colors and the regions in the product style drawing to obtain the set of region color matching correspondences; Obtaining the atmosphere prompts specified for the regions in the product style drawing, automatically generating corresponding colors according to the obtained atmosphere prompts; based on the generated colors, constructing at least one set of correspondences between the colors and the regions in the product style drawing to obtain the set of region color matching correspondences; Obtaining a color matching scheme; for the regions in the product style drawing, randomly selecting colors from the obtained color matching scheme as the corresponding colors, and constructing at least one set of correspondences between the colors and the regions in the product style drawing to obtain the set of region color matching correspondences.

21. A product color matching method, characterized in that, It includes: Obtaining a clothing style drawing and color matching information; The color matching information includes at least one color information; Fusing the clothing style drawing and the color matching information to obtain an intermediate clothing production effect drawing; Fusing the intermediate clothing production effect drawing and the color matching information to obtain a final clothing production effect drawing.

22. A product color matching device, characterized in that, It includes: An obtaining unit, configured to obtain a product style drawing and color matching information; The color matching information includes at least one color information; A first fusing unit, configured to fuse the product style drawing and the color matching information to obtain an intermediate effect drawing; A second fusing unit, configured to fuse the intermediate effect drawing and the color matching information to obtain a final effect drawing.

23. A product color matching device, characterized in that, It includes: A clothing obtaining unit, configured to obtain a clothing style drawing and color matching information; The color matching information includes at least one color information; A third fusing unit, configured to fuse the clothing style drawing and the color matching information to obtain an intermediate clothing production effect drawing; A fourth fusing unit, configured to fuse the intermediate clothing production effect drawing and the color matching information to obtain a final clothing production effect drawing.

24. An electronic device, including: A processor; A memory for storing processor-executable instructions; Wherein, the processor realizes the method according to any one of claims 1 to 10 and 13 to 21 by running the executable instructions.

25. A computer-readable storage medium, on which computer instructions are stored, and when the instructions are executed by a processor, the steps of the method according to any one of claims 1 to 10 and 13 to 21 are realized.

Images