基于SASGAN的戏剧脸谱多样化生成  

Diversified generation of theatrical masks based on SASGAN

作　　者：古天骏 熊苏雅 林晓[1,3,4] GU Tianjun;XIONG Suya;LIN Xiao(The College of information,Mechanical and Electrical Engineering,Shanghai Normal University,Shanghai 200234,China;Institute of Marine Equipment,Shanghai Jiao tong University,Shanghai 200240,China;Shanghai Engineering Research Center of Intelligent Education and Big data,Shanghai Normal University,Shanghai 200234,China;The Research Base of Online Education for Shanghai Middle and Primary Schools,Shanghai 200234,China)

机构地区：[1]上海师范大学信息与机电工程学院,上海200234 [2]上海交通大学海洋装备研究院,上海200240 [3]上海师范大学上海智能教育大数据工程技术研究中心,上海200234 [4]上海市中小学在线教育研究基地,上海200234

出　　处：《图学学报》2024年第1期102-111,共10页Journal of Graphics

基　　金：国家自然科学基金项目(61502220,61572326,61775139,61872242)。

摘　　要：为解决现有自动生成的戏剧脸谱在分辨率和真实性上效果不佳的问题,提出了基于自注意力机制的风格化生成对抗网络(SASGAN)。首先在StyleGAN的基础上引入了自注意力机制以及矢量量化方法,增强了对脸谱图案几何结构特征的提取,接着通过多样化差异性增强(DDG)扩充数据,采用脸谱色调辅助算法对DDG方法进行补充,建立了包含12599张图像的戏剧脸谱数据集,最后在此数据集上进行训练,生成了兼顾多样性和真实性的脸谱图像。实验结果表明,对于戏剧脸谱图像,DDG方法较传统方法在数据增广方面有着较大提升,而SASGAN则提升了戏剧脸谱图像的分辨率和真实性,在主观视觉上得到了理想的效果。To address the problem of low resolution and lack of realism in existing automatically generated theatrical masks,a stylized generative adversarial network(SASGAN)based on a self-attentive mechanism was proposed.Firstly,SASGAN introduced the self-attentive mechanism and vector quantization method based on StyleGAN,thereby enhancing the extraction of geometric structure features of mask patterns.Subsequently,the diversified differentiation generation(DDG)method was supplemented with a mask hue-assisted algorithm by expanding the data with DDG to build a theatrical mask dataset containing 12,599 images.The final training was performed on this dataset to generate mask images with both diversity and realism.The experimental results demonstrated significant improvement in data augmentation for theatrical masks using the DDG method compared to the traditional methods,while SASGAN enhanced the resolution and realism of theatrical masks,achieving the desired effect in subjective visualization.

关 键 词：戏剧脸谱 生成对抗网络 图像生成 注意力机制 矢量量化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]