Automatic generation of stylized icons

Abstract

In an embodiment of the present disclosure, a method for automatically generating stylized icons is proposed. After obtaining the query text input by the user, a trained generator is first used to generate a structured icon that can represent the structure of the object, and then the structured icon is stylized, such as filling in color or adding other styles, so as to generate a high-quality stylized icon for the user. In an embodiment of the present disclosure, structured icons and stylized icons are generated respectively in two stages, wherein the structured icon can clearly represent the structure of the object, and the stylized icon can have richer colors and styles. Therefore, the stylized icons generated according to the embodiments of the present disclosure are of higher quality and more realistic, thereby improving the user experience when the icon is generated.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of computers, and more particularly, to a method, a device, and a computer program product for generating a stylized icon. Background Art

[0002] An icon is a graphic symbol with referential meaning, which is highly condensed and easy to communicate. In the computer field, an icon can represent a picture or an object. For example, various icons in an operating system help users quickly find and locate target programs or files. Usually, each icon in a set of icons can have the same size and attribute format, and can usually have a relatively small size.

[0003] Icons are widely used. For example, users may need to use icons to display and express themselves when writing documents, preparing slides, or chatting online. Icons are widely used because they are more vivid than text and simpler than general images. Some common document editing tools have a certain number of icons preset, and users can select one or more to insert into the document when editing. In addition, users can also make their own icons through professional drawing software, or use keywords to search for related icons on the Internet. Summary of the invention

[0004] In an embodiment of the present disclosure, a method for automatically generating stylized icons is proposed. After obtaining the query text input by the user, a trained generator is first used to generate a structured icon representing the structure of the object, and then the structured icon is stylized, such as filling in color or adding other styles, so as to generate a high-quality stylized icon for the user. In an embodiment of the present disclosure, structured icons and stylized icons are generated respectively in two stages, wherein the structured icon can clearly represent the structure of the object, and the stylized icon can have richer colors and styles. Therefore, the stylized icons generated according to the embodiments of the present disclosure are of higher quality and more realistic, thereby improving the user experience when the icon is generated.

[0005] This Summary is provided to introduce a selection of concepts in a simplified form that are further described in the Detailed Description below. This Summary is not intended to identify key features or essential features of the disclosure, nor is it intended to limit the scope of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0007] Figure 1A block diagram illustrating a computing device / server in which one or more embodiments of the present disclosure may be implemented;

[0008] Figure 2 A flow chart showing a method for automatically generating stylized icons in which the present disclosure may be implemented;

[0009] Figure 3 A schematic diagram showing a process for automatically generating stylized icons in which the present disclosure may be implemented;

[0010] Figure 4 A schematic diagram showing an example set of structured icons according to an embodiment of the present disclosure;

[0011] Figure 5 A schematic diagram showing an example set of stylized icons according to an embodiment of the present disclosure;

[0012] Figure 6 A schematic diagram showing a process for generating a structured icon according to an embodiment of the present disclosure is shown;

[0013] Figure 7 A schematic diagram showing an architecture for training an icon generation model using a generative adversarial network (GAN) according to an embodiment of the present disclosure; and

[0014] Figure 8 A schematic diagram of a process for training an icon generation model using GAN according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0015] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0016] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0017] Traditionally, when a user needs to use an icon, such as when editing a document or a slideshow, he or she usually uses keywords to search for related icons on the Internet, or makes his or her own icons through professional mapping software. However, the icons retrieved from the Internet may have poor relevance and uneven quality, and making icons by oneself requires the user to have a certain design foundation, which is difficult. Therefore, traditional methods are difficult to obtain the icons required by users, and the quality of the icons obtained is not guaranteed. In addition, some current icon generation tools or icon creation tools can only generate simple icons, or can only generate some text art words, which results in poor icon quality and is not realistic enough.

[0018] To this end, an embodiment of the present disclosure proposes an automatic generation technology for stylized icons. The inventor of the present invention realizes that in order to generate a high-quality stylized icon, the structural features of the icon can be determined first, and then the icon can be stylized, which can make the generated icon higher in quality and more realistic, while reducing the workload of experienced designers. Therefore, in an embodiment of the present disclosure, structured icons and stylized icons are generated respectively in two stages, wherein the structured icon can clearly characterize the structure of the object, and the stylized icon can have richer colors and styles. The stylized icons generated according to the embodiments of the present disclosure are of high quality and more realistic, and even arbitrary icons can be generated, thereby improving the user experience when the icon is generated.

[0019] The following references Figures 1 to 8 To illustrate the basic principles and several example implementations of the present disclosure. Figure 1 1 shows a block diagram of a computing device / server 100 in which one or more embodiments of the present disclosure may be implemented. It should be understood that Figure 1 The computing device / server 100 shown is merely exemplary and should not constitute any limitation on the functionality and scope of the embodiments described herein.

[0020] like Figure 1 As shown, computing device / server 100 is in the form of a general-purpose computing device. Components of computing device / server 100 may include, but are not limited to, one or more processors or processing units 110, memory 120, storage device 130, one or more communication units 140, one or more input devices 150, and one or more output devices 160. Processing unit 110 may be an actual or virtual processor and is capable of performing various processes according to a program stored in memory 120. In a multi-processor system, multiple processing units execute computer executable instructions in parallel to increase the parallel processing capabilities of computing device / server 100.

[0021] The computing device / server 100 typically includes a plurality of computer storage media. Such media may be any available media accessible to the computing device / server 100, including but not limited to volatile and non-volatile media, removable and non-removable media. The memory 120 may be a volatile memory (e.g., registers, caches, random access memory (RAM)), a non-volatile memory (e.g., a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. The storage device 130 may be a removable or non-removable medium, and may include a machine-readable medium, such as a flash drive, a disk, or any other medium, which may be capable of being used to store information and / or data (e.g., training data for training) and may be accessed within the computing device / server 100.

[0022] The computing device / server 100 may further include additional removable / non-removable, volatile / non-volatile storage media. Figure 1 As shown in , a disk drive for reading or writing from a removable, non-volatile disk (e.g., a "floppy disk") and an optical drive for reading or writing from a removable, non-volatile optical disk may be provided. In these cases, each drive may be connected to a bus (not shown) by one or more data media interfaces. The memory 120 may include a computer program product 125 having one or more program modules, such as an icon generation model according to an embodiment of the present disclosure, which are configured to perform various methods or actions of various embodiments of the present disclosure.

[0023] The communication unit 140 enables communication with other computing devices via a communication medium. Additionally, the functionality of the components of the computing device / server 100 may be implemented in a single computing cluster or multiple computing machines that are capable of communicating via a communication connection. Thus, the computing device / server 100 may operate in a networked environment using a logical connection to one or more other servers, a network personal computer (PC), or another network node.

[0024] Input device 150 may be one or more input devices, such as a mouse, keyboard, trackball, etc. Output device 160 may be one or more output devices, such as a display, speaker, printer, etc. Computing device / server 100 may also communicate with one or more external devices (not shown) as needed, such as storage devices, display devices, etc., with one or more devices that enable a user to interact with computing device / server 100, or with any device that enables computing device / server 100 to communicate with one or more other computing devices (e.g., a network card, a modem, etc.) through communication unit 140. Such communication may be performed via an input / output (I / O) interface (not shown).

[0025] like Figure 1 As shown, the computing device / server 100 can receive text input 305 (which can be in the form of natural language, such as words, phrases, sentences, etc.) from the user through the input device 150, and then use the computer program product 125 to process the text input 305, and output the stylized icon 325 through the output device 160. For example, after the user enters the query of "cake", the stylized icon of the cake is output. In some embodiments, the user can also select one from the generated multiple cake icons and select the desired cake style by the user.

[0026] Those skilled in the art will appreciate that although Figure 1 1 shows that text input is received through the input unit 150 and the stylized icon is output through the output device 160, but the input may be received directly through the communication unit 140 and the output may be sent. Figures 2 to 8 An example embodiment of how the computer program product 125 generates a stylized icon based on a user's text input is described in detail.

[0027] Figure 2 1 shows a flow chart of a method 200 for automatically generating a stylized icon in which the present disclosure may be implemented. It should be understood that the method 200 may be implemented by reference to Figure 1 The computing device / server 100 described here performs. For the sake of clarity Figure 2 Method 200, herein also referred to Figure 3 The process 300 of automatically generating a stylized icon is shown and described together.

[0028] At 202, a text input for generating an icon of an object is obtained. Figure 3, the generator 310 obtains text input 305 from the user, such as "cake", which indicates that the user wants to generate an icon of a cake object. Different from the traditional method of retrieving icons from a network or a local resource library, the embodiment of the present disclosure can generate icons using an icon generation model, and the generated icons may not have existed before.

[0029] At 204, based on the text input, a first icon is generated that reflects the structure of the object, wherein the structure of the object is represented by lines. Figure 3 The generator 310 in (which has been trained, the following reference Figures 7 and 8 The example implementation of the training generator 310 is described) generating a simple cake structured icon 315 according to the text input 305 (e.g., "cake"). In the structured icon 315, the structural features of the object "cake" can be clearly reflected, and the structure of the cake is characterized by a set of lines, such as straight lines and / or curves. In some embodiments of the present disclosure, the structured icon can be a simple black and white icon.

[0030] At 206, the first icon is stylized into a second icon. In some embodiments, the first icon can be stylized into the second icon based on the reference image, so that the stylized icon has a style similar to the reference image. For example, Figure 3 The stylizer 320 in the embodiment converts the simple structured icon 315 into a stylized icon 325 with a mosaic style according to the style (e.g., mosaic style) of the reference image 318. Therefore, in the method 200 of the embodiment of the present disclosure, the structured icon 315 and the stylized icon 325 are generated respectively through two stages, wherein the structured icon 315 can clearly represent the structure of the object, and the stylized icon 325 has richer colors and styles.

[0031] In some embodiments, the first icon can also be stylized into a second icon based on a reference color or tone. For example, if the reference color is blue, the first icon can be stylized into a blue tone icon. Alternatively, the first icon can also be stylized into a second icon based on some parameters in the machine learning model. In the embodiments of the present disclosure, the stylization process may include but is not limited to color filling, effect filling, style rendering, other pixel conversions, and the like of the icon. Therefore, the stylized icons generated according to the embodiments of the present disclosure are of higher quality and more realistic, and can enhance the user experience when the icon is generated.

[0032] In some embodiments, after the structured icon 315 is generated, the structured icon 315 can be edited according to the user's intention. For example, the user can adjust the shape and properties of the lines in the structured icon. For example, the user can thicken the lines in the structured icon 315, or enlarge the candlestick lines in the structured icon 315, and so on. In this way, the user can be provided with the option of further adjusting the generated icon, so that the generated icon is more in line with the user's intention, further improving the user experience. After the user adjusts the object structure in the structured icon 315, the stylizer 320 will perform stylization conversion on the adjusted structured icon 315.

[0033] In some embodiments, the stylizer 320 may further consider the semantic information of the object when adding style to the structured icon 315. For example, assuming that the generated structured icon is a leaf icon, and according to the semantics and meaning of leaves, its common colors are green, red, yellow, etc., so the stylizer 320 colors the black and white structured icon 315 based on the semantics of leaves, such as adding green, red, yellow, etc., but does not add white, black, etc., which do not conform to the semantics of leaves at all.

[0034] refer to Figure 4 , which shows a schematic diagram of an example set 400 of structured icons according to an embodiment of the present disclosure. Figure 4 As shown, the example set 400 includes a structured icon set 410 of cakes, a structured icon set 420 of pizzas, a structured icon set 430 of ducks, and a structured icon set 440 of bananas. These icons can be simple black and white, and their respective structural features are characterized by lines, etc. These icon sets can be, for example, structured icon sets generated by the generator 310 of the present disclosure based on corresponding user queries. For example, when the user enters "cake", a structured icon set 410 of cakes can be generated for user selection. When the user enters "pizza", a structured icon set 420 of pizza can be generated for user selection. When the user enters "duck", a structured icon set 430 of ducks can be generated for user selection. When the user enters "banana", a structured icon set 440 of bananas can be generated for user selection.

[0035] In some embodiments, the generator 310 may generate a plurality of structured icons for a query input by a user, such as Figure 4 As shown in , the user can then select the one he wants most for subsequent stylization. Alternatively, the generator 310 can also generate only one structured icon (e.g. Figure 3 The structured icon 315 shown in FIG. 1 is then stylized by the stylizer 320 directly without user selection.

[0036] Figure 5 FIG. 5 is a schematic diagram showing an example set 500 of stylized icons according to an embodiment of the present disclosure. Figure 5 As shown, structured icons 510, 520, 530 and 540 can be simple structured icons shown by the generator 310 based on different user queries. Next, the stylizer 320 can stylize these structured icons 510, 520, 530 and 540 into icons of different styles. For example, for the structured icon 510: if the reference image is a pencil style image 551, a pencil style cake icon 511 is generated; if the reference image is a stone style image 552, a stone style cake icon 512 is generated; if the reference image is a mosaic style image 553, a mosaic style cake icon 513 is generated; if the reference image is a candy style image 554, a candy style cake icon 514 is generated; if the reference image is a starry sky style image 555, a starry sky style cake icon 515 is generated. In some embodiments, the reference image can be selected by the user. Alternatively, a default reference image can also be set for the user without the user's selection. If the user is not satisfied with the generated stylized icon, the reference image can be replaced and re-stylized.

[0037] Similarly, the stylizer 320 can also generate banana icons 521-525 of various styles, duck icons 531-535 of various styles, and pizza icons 541-545 of various styles. Therefore, the embodiments of the present disclosure can generate high-quality icons of various styles according to the user's intention, thereby improving the user experience. In addition, the scheme according to the embodiments of the present disclosure can be embedded in some office software as a functional module to provide users with high-quality stylized icons and improve the user's office efficiency.

[0038] In some embodiments, if multiple semantics of multiple objects, such as "grass" and "horse", can be obtained from a user query (for example, a sentence "a horse on the grass"), then icons of grass and horse can be generated respectively. Next, the icons of grass and horse can be combined to generate a combined icon of horse on grass. In this way, more complex icon generation requirements of users can be met, further improving user experience.

[0039] In some embodiments, an animated icon may be created based on a user query. For example, a series of multiple icons may be generated, and these icons may have a sequential relationship. By combining the series of icons, an animated icon may be created instead of a static icon. In this way, more icon resources may be provided to users, further improving the user experience.

[0040] Figure 6 FIG. 6 is a schematic diagram of a process 600 for generating a structured icon according to an embodiment of the present disclosure. Figure 6 As shown, the generator 310 first obtains the word vector 605 and random noise 608 of the user query, wherein the user query may include at least one of the category, keyword, text description and attribute associated with the object, and the random noise 608 is usually generated based on a random value between 0 and 1, which may be a random variable subject to Gaussian distribution. In addition, the resolution of the structured icon to be generated may also be specified, such as 64×64.

[0041] The generator 310 first generates vectorized structure data 620 based on the word vector 605 and the random noise 608, wherein the vectorized structure data 620 may include line and curve parameters, such as coordinate data of the line and curve. Figure 6 As shown, a straight line can usually be represented by the coordinates of 2 points, while a curve can usually be represented by the coordinates of 4 points. In some embodiments, the vectorized structure data 620 can be stored in the format of a scalable vector graphics (SVG) file. Compared with ordinary images, vector graphics have clearer lines and smoother edges. Therefore, some embodiments of the present disclosure can make the generated icons of higher quality through vectorization. In some embodiments, the generator 310 can be implemented by a fully connected layer.

[0042] Continue to refer Figure 6 , draw the model and then convert the vectorized structure data 620 into raster data 630 of the icon, such as Figure 6 As shown, the raster data 630 may include a set of straight lines and a set of curves. Next, the drawing model draws a structured icon 640 based on the set of straight lines and the set of curves in the raster data 630 through maximum pooling. The structured icon 640 has a higher line quality. It should be understood that although Figure 6 The drawing model of the drawing action in is set separately from the generator 310, but the drawing model can also be included in the generator 310. In some embodiments, the drawing model can be implemented by two-dimensional deconvolution of several layers (e.g., 5 layers, 6 layers, etc.).

[0043] In some embodiments, some icon rules may be set to adjust the generated structured icon. For example, after the generator 310 generates the structured icon, the vertices with similar distances in the structured icon are connected, thereby optimizing the structured icon. In addition, some other icon rules may be set for optimization, for example, irregular lines may be made smoother, irrelevant lines may be removed, the size and direction of the object structure may be adjusted, and so on.

[0044] Continue to refer Figure 6, the discriminator 650 can determine whether the generated structured icon 640 is true or false, as shown in 655, thereby helping to train the generator 310 to generate more realistic structured icons. In an embodiment of the present disclosure, the generator 310 and the discriminator 650 can form an icon generation model, and the trained icon generation model can generate high-quality stylized icons for user queries. Figure 7 and Figure 8 An example implementation for training an icon generation model including a generator 310 and a discriminator 650 is shown.

[0045] Figure 7 FIG. 7 is a schematic diagram of an architecture 700 for training an icon generation model using a GAN according to an embodiment of the present disclosure. Figure 7 As shown, the architecture 700 mainly includes a generator 310 and a discriminator 650. During the training process, the generator 310 generates a structured icon 410 based on the text input 305, and then transmits the generated structured icon 410 and the real structured icon 710 to the discriminator 650. The discriminator 650 trains the discriminator 650 and the generator 310 based on the judgment 720 of the authenticity of each input icon. Compared with training with ordinary images, since the embodiments of the present disclosure use simple structured icons for training, it is easier and faster to train an icon generation model.

[0046] like Figure 7 As shown, an example of training with a real icon of a cake for the object "cake" is shown. Similarly, real icons of other categories can also be used for training, including but not limited to animals, people, food, sports, etc. In some embodiments, the Wasserstein GAN (WGAN) algorithm can be used to train the icon generation model of the present disclosure. Compared with ordinary GAN, WGAN is easier to train and converges faster.

[0047] Figure 8 FIG. 8 is a schematic diagram of a process 800 for training an icon generation model using a GAN according to an embodiment of the present disclosure. Figure 8 As shown, the generator 310 generates corresponding generated data 815 based on the input 805, and then passes the generated data 815 and the real data 825 to the discriminator 650, and the discriminator 650 determines the authenticity 830 of each input data.

[0048] During the training process, the discriminator 650 and the generator 310 are iteratively trained. During this process, the discriminator 650 and the generator 310 are constantly optimizing their respective networks, thereby forming a competitive confrontation until the two sides reach a dynamic balance (i.e., Nash equilibrium). After the training is completed, the generator 310 can simulate the distribution of the real data 825 (the generated data 815 is the same as the real data 825), and the discriminator 650 cannot judge the authenticity of the generated data at this time, and the accuracy of the judgment is 50%.

[0049] In the specific iterative training process, the generator 310 is first fixed and the discriminator 650 is trained, as shown by arrow 840. The discriminator 650 can mark the output after processing the real data 825 as 1, and mark the output after processing the generated data 815 as 0. In this way, the discriminator 650 can be trained first. Next, the discriminator 650 is fixed and the generator 310 is trained, as shown by arrow 850. The generator 310 optimizes its own generation network so that the generated data 815 output by it is consistent with the real data 825, and the generated data can deceive the judgment of the discriminator 650. The discriminator 650 and the generator 310 can be trained iteratively multiple times until the icon generation model becomes converged.

[0050] The methods and functions described herein may be performed at least in part by one or more hardware logic components. For example, but not limited to, illustrative types of hardware logic components that may be used include field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chip (SOCs), complex programmable logic devices (CPLDs), etc.

[0051] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0052] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0053] In addition, although each operation is described in a specific order, this should be understood as requiring such operation to be performed in the specific order shown or in a sequential order, or requiring that all illustrated operations should be performed to obtain the desired result. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present disclosure. Certain features described in the context of a separate implementation can also be implemented in a single implementation in combination. On the contrary, the various features described in the context of a single implementation can also be implemented in multiple implementations individually or in any suitable sub-combination.

[0054] Some example implementations of the present disclosure are listed below.

[0055] In one aspect, a computer-implemented method is provided, including obtaining text input for generating an icon of an object, generating a first icon embodying a structure of the object based on the text input, wherein the structure of the object is represented by lines, and stylizing the first icon into a second icon.

[0056] In some embodiments, obtaining text input for generating an icon of an object includes: obtaining text input and random input, wherein the text input includes at least one of a category, a keyword, a text description, and an attribute associated with the object, and the random input is a random variable that obeys a Gaussian distribution.

[0057] In some embodiments, generating a first icon having a structure of an object includes: generating vectorized structural data based on text input and random input, wherein the vectorized structural data includes coordinate data of straight lines and curves; obtaining raster data of the first icon based on the vectorized structural data, wherein the raster data includes a set of straight lines and a set of curves; and drawing the first icon based on the set of straight lines and the set of curves in the raster data.

[0058] In some embodiments, drawing the first icon includes: drawing candidate icons based on a set of straight lines and a set of curves in the raster data; and optimizing the candidate icons into the first icon by connecting vertices with similar distances in the candidate icons.

[0059] In some embodiments, stylizing a first icon into a second icon includes: receiving user input for the first icon, wherein the user input indicates adjusting at least one of a shape and a property of a line in the first icon; and adjusting a structure of an object in the first icon based on the received user input.

[0060] In some embodiments, stylizing the first icon into the second icon further includes: determining semantic information of the object based on the text input; and filling the first icon with color based on the structure and semantic information of the object.

[0061] In some embodiments, wherein the generation of the first icon is performed by a generator, the method further comprises: using a generative adversarial network (GAN) to iteratively train a discriminator and a generator, wherein the discriminator is used to determine the authenticity of the icon.

[0062] In some embodiments, the method further includes: generating at least two icons based on at least two semantics obtained from the text input; and combining the at least two icons into a single new icon.

[0063] In some embodiments, the method further includes: obtaining a series of icons based on the text input; and creating an animated version of the icons based on the series of icons.

[0064] In another aspect, an electronic device is provided. The electronic device includes: a processing unit; a memory coupled to the processing unit and storing instructions, the instructions performing actions when executed by the processing unit, the actions including: obtaining text input for generating an icon of an object; based on the text input, generating a first icon embodying a structure of the object, wherein the structure of the object is represented by lines; and stylizing the first icon into a second icon.

[0065] In some embodiments, obtaining text input for generating an icon of an object includes: obtaining text input and random input, wherein the text input includes at least one of a category, a keyword, a text description, and an attribute associated with the object, and the random input is a random variable that obeys a Gaussian distribution.

[0066] In some embodiments, generating a first icon having a structure of an object includes: generating vectorized structural data based on text input and random input, wherein the vectorized structural data includes coordinate data of straight lines and curves; obtaining raster data of the first icon based on the vectorized structural data, wherein the raster data includes a set of straight lines and a set of curves; and drawing the first icon based on the set of straight lines and the set of curves in the raster data.

[0067] In some embodiments, drawing the first icon includes: drawing candidate icons based on a set of straight lines and a set of curves in the raster data; and optimizing the candidate icons into the first icon by connecting vertices with similar distances in the candidate icons.

[0068] In some embodiments, stylizing a first icon into a second icon includes: receiving user input for the first icon, wherein the user input indicates adjusting at least one of a shape and a property of a line in the first icon; and adjusting a structure of an object in the first icon based on the received user input.

[0069] In some embodiments, stylizing the first icon into the second icon further includes: determining semantic information of the object based on the text input; and filling the first icon with color based on the structure and semantic information of the object.

[0070] In some embodiments, wherein the generation of the first icon is performed by a generator, the action further includes: using a generative adversarial network (GAN) to iteratively train a discriminator and a generator, wherein the discriminator is used to determine the authenticity of the icon.

[0071] In some embodiments, the actions further include: generating at least two icons based on at least two semantics obtained from the text input; and combining the at least two icons into a single new icon.

[0072] In some embodiments, the actions further include: obtaining a series of icons based on the text input; and creating an animated version of the icons based on the series of icons.

[0073] In another aspect, a computer program product is provided. The computer program product is stored in a non-transitory computer storage medium and includes machine executable instructions, which when executed in a device, cause the device to: obtain text input for generating an icon of an object; based on the text input, generate a first icon that embodies the structure of the object, wherein the structure of the object is represented by lines; and stylize the first icon into a second icon.

[0074] In some embodiments, obtaining text input for generating an icon of an object includes: obtaining text input and random input, wherein the text input includes at least one of a category, a keyword, a text description, and an attribute associated with the object, and the random input is a random variable that obeys a Gaussian distribution.

[0075] In some embodiments, generating a first icon having a structure of an object includes: generating vectorized structural data based on text input and random input, wherein the vectorized structural data includes coordinate data of straight lines and curves; obtaining raster data of the first icon based on the vectorized structural data, wherein the raster data includes a set of straight lines and a set of curves; and drawing the first icon based on the set of straight lines and the set of curves in the raster data.

[0076] In some embodiments, drawing the first icon includes: drawing candidate icons based on a set of straight lines and a set of curves in the raster data; and optimizing the candidate icons into the first icon by connecting vertices with similar distances in the candidate icons.

[0077] In some embodiments, stylizing a first icon into a second icon includes: receiving user input for the first icon, wherein the user input indicates adjusting at least one of a shape and a property of a line in the first icon; and adjusting a structure of an object in the first icon based on the received user input.

[0078] In some embodiments, stylizing the first icon into the second icon further includes: determining semantic information of the object based on the text input; and filling the first icon with color based on the structure and semantic information of the object.

[0079] In some embodiments, where the generation of the first icon is performed by a generator, the machine executable instructions, when executed in the device, further cause the device to: use a generative adversarial network (GAN) to iteratively train a discriminator and a generator, where the discriminator is used to determine the authenticity of the icon.

[0080] In some embodiments, the machine executable instructions, when executed in the device, further cause the device to: generate at least two icons based on at least two semantics obtained from the text input; and combine the at least two icons into a single new icon.

[0081] In some embodiments, the machine executable instructions, when executed in the device, further cause the device to: obtain a series of icons based on the text input; and create an animated version of the icons based on the series of icons.

[0082] Although the present disclosure has been described in language specific to structure and / or methodological logic actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are merely example forms of implementing the claims.

Claims

1. A method for generating a stylized icon, comprising: Get the text input used to generate the icon of the object; Based on the text input, generating a first icon embodying a structure of the object, the structure of the object being represented by lines; Identify reference images with corresponding styles; Stylizing the first icon into a second icon according to the style of the reference image; generating at least two icons based on at least two objects identified in the text input; as well as The at least two icons are combined into a single new icon.

2. The method according to claim 1, further comprising: A noise input of random values ​​is generated, wherein the text input includes at least one of a category, a keyword, a text description, and an attribute associated with the object.

3. The method according to claim 2, wherein generating a first icon having a structure of the object comprises: Based on the text input and the noise input, generate vectorized structural data, the vectorized structural data including coordinate data of straight lines and curves; Based on the vectorized structural data, obtaining raster data of the first icon, the raster data including a straight line set and a curve set; as well as The first icon is drawn based on the set of straight lines and the set of curves in the raster data.

4. The method according to claim 3, wherein drawing the first icon comprises: Drawing candidate icons based on the set of straight lines and the set of curves in the raster data; as well as The candidate icons are optimized to be the first icons by connecting vertices with similar distances among the candidate icons.

5. The method of claim 1, wherein stylizing the first icon into a second icon comprises: receiving a user input for the first icon, the user input indicating adjusting at least one of a shape and an attribute of a line in the first icon; as well as Based on the received user input, the structure of the objects in the first icon is adjusted.

6. The method according to claim 5, wherein stylizing the first icon into a second icon further comprises: Based on the text input, determining semantic information of the object; as well as The first icon is filled with color based on the structure and the semantic information of the object.

7. The method according to claim 1, wherein the generating of the first icon is performed by a generator, the method further comprising: A generative adversarial network (GAN) is used to iteratively train a discriminator and the generator, wherein the discriminator is used to judge whether an icon is real or fake.

8. The method according to claim 1, further comprising: Based on the text input, obtain a series of icons; as well as Based on the series of icons, an icon in animated form is created.

9. The method of claim 1 , wherein identifying the reference image comprises: A selection of the reference image is received from a user of the device.

10. The method of claim 1, wherein identifying the reference image comprises: At least one default reference image is identified for a user of the device in the absence of receiving a user selection of the reference image.

11. The method according to claim 10, further comprising: causing display of the second icon on the device; In response to causing the display, a selection of another reference image is received from the user of the device.

12. The method of claim 1, wherein the second icon has a similar style to the reference image.

13. The method of claim 1, wherein stylizing the first icon into the second icon is based on at least one of: a reference color, a color fill, an effect fill, or a style rendering of the reference image.

14. An electronic device comprising: Processing unit; a memory coupled to the processing unit and storing instructions, the instructions performing actions when executed by the processing unit, the actions comprising: Get the text input used to generate the icon of the object; Based on the text input, generating a first icon embodying a structure of the object, the structure of the object being represented by lines; Identify reference images with corresponding styles; Stylizing the first icon into a second icon according to the style of the reference image; generating at least two icons based on at least two objects identified in the text input; and The at least two icons are combined into a single new icon.

15. The apparatus of claim 14, wherein the actions further comprise: A noise input of random values ​​is generated, wherein the text input includes at least one of a category, a keyword, a text description, and an attribute associated with the object.

16. The apparatus of claim 15, wherein generating a first icon having a structure of the object comprises: Based on the text input and the noise input, generate vectorized structural data, the vectorized structural data including coordinate data of straight lines and curves; Based on the vectorized structural data, obtaining raster data of the first icon, the raster data including a straight line set and a curve set; as well as The first icon is drawn based on the set of straight lines and the set of curves in the raster data.

17. The device of claim 16, wherein drawing the first icon comprises: Drawing candidate icons based on the set of straight lines and the set of curves in the raster data; as well as The candidate icons are optimized to be the first icons by connecting vertices with similar distances among the candidate icons.

18. The device of claim 14, wherein stylizing the first icon into a second icon comprises: receiving a user input for the first icon, the user input indicating adjusting at least one of a shape and an attribute of a line in the first icon; as well as Based on the received user input, the structure of the objects in the first icon is adjusted.

19. A computer program product stored in a computer storage medium and comprising machine executable instructions which, when executed in a device, cause the device to: Get the text input used to generate the icon of the object; Based on the text input, generating a first icon embodying a structure of the object, the structure of the object being represented by lines; Identify reference images with corresponding styles; Stylizing the first icon into a second icon according to the style of the reference image; generating at least two icons based on at least two objects identified in the text input; as well as The at least two icons are combined into a single new icon.

Images