Research On Color Visual Style Transfer Of Mobile Application Interface Based On Deep Learning

The visual appearance of the mobile application interface is an important part of the interaction between the user and the mobile application,and its color visual effect occupies an important position in the entire interaction process.In the color design stage of the traditional UI interface,UI designers need to choose a suitable color style based on product positioning and facing the crowd.But the problem that follows is that UI designers need to design multiple color styles for an interface and choose from them,which will increase the designer’s workload.Therefore,when multiple color styles of a mobile application interface are displayed in front of a UI designer,it will reduce the work of the UI designer and contribute to the color design of the application interface.In order to realize multiple color styles of an application interface,UI designers can select application interfaces of different color styles to change the target interface.Therefore,the existing research mainly uses deep learning methods to realize the color style transfer of the application interface,but the image of the application interface generated by these methods is low in definition and cannot identify the component elements in the interface;in addition,there is no clear indicator.It is used to measure the effect of color style migration of application interface.Therefore,starting from the color vision characteristics of mobile application interfaces,this article conducts research on the color style migration of mobile application interfaces.The main contents of the work are as follows:(1)Interface color visual style based on mobile applications.In order to realize a color visual style transfer based on mobile applications,the color style components extracted from the application interface should be determined first.To solve this problem,this paper proposes a color visual style based on mobile application interface.The model starts from the color point of mobile application interface,and determines its color style according to the color composition and proportion of the mobile application interface and combined with color psychology.Finally,the color moments are used to extract the color characteristics of each style of interface map,and the color characteristics of the application interface are quantified.Through the verification of the similarity of the color features of the color moments,the color moment quantification index of each color style is obtained,which provides migration conditions for subsequent mobile application interface color style migration and evaluation of the generated results.(2)Research on interface color style migration based on component element contour constraints.Aimed at ImagineNet,the problem of clarity and missing component element outlines in the color style transfer of the application interface is studied.This paper proposes a color style transfer method for mobile application interface.This method is improved on the image style conversion model CycleGAN framework.First,the ResNet-9 network structure is used as the converter in the CycleGAN network;secondly,the image style attention SRM module is added to In the converter;finally,LSGANS least square loss is used to replace the loss function of the original GAN adversarial network on the objective function of CycleGAN.The experimental results show that the method proposed in this paper achieves a 97%similarity on the structural similarity of the application interface image SSIM,which is higher than the 85% similarity of the ImagineNet research method.In the peak signal-to-noise ratio PSNR index,the research method in this chapter is 15.54% higher than that of ImagineNet.Therefore,the method proposed in this paper can ensure the clear outline of the component elements after the color migration of the application interface.(3)The color style transfer method based on the mobile application interface image proposed in the previous chapter is mainly for the color style transfer of a type of interface,which leads to the main learning of the color style feature distribution of a type of interface during model training.Determined by the color characteristics of an interface.Therefore,the model can only achieve the migration of the main body color of the interface during the color style migration process,and the color information of some component elements is not learned,resulting in a large difference between the color characteristics of the generated interface and the migration interface,and it is not fully realized.Interface color and visual style migration.In addition,because the component elements in the application interface have obvious color characteristics,this paper proposes a color style transfer algorithm based on the semantic matching of interface component elements.First,transfer the color style of the migration interface and the target interface through the migration model in the previous chapter.Afterwards,Mask RCNN(target recognition and instance segmentation)is used to identify the category and function description information of the component elements in the application interface and perform similarity matching to determine the correspondence between the migration style interface and the component elements of the target interface.Then,the corresponding component elements are replaced with RGB colors to generate a stylized target interface.Finally,the color moment feature of the color visual style model in Chapter 3 is used as the migration index,and the color moment similarity loss between the migration interface and the stylized target interface is calculated and fed back to the adversarial model.The experimental results show that the generated interface after migration is significantly improved than the method(2)in terms of the consistency of the color characteristics of the application interface,and has a better color visual effect.In addition,in terms of the consistency of the color style of the application interface,the interface generated by the research method in this chapter and the migration interface have the same color style.To sum up,in order to meet the user’s visual individual needs,this article implements a color visual style transfer model based on mobile application interfaces.The color migration of the target interface was carried out through the deep learning method and a good migration effect was achieved,laying the foundation for the visual development of the mobile application interface.