基于生成对抗网络的书法纺织图案设计开发
2021-03-15陈涵沈雷汪鸣明张希莹任祥放
陈涵 沈雷 汪鸣明 张希莹 任祥放



摘要: 随着区域民族纺织品全球化格局的形成,中国风格成为了全球纺织品流行趋势之一。中国传统纺织图案在世界范围的流行对其设计开发方法提出了更高要求。文章以书法纺织图案为研究对象,提出了基于深度学习生成对抗网络的传统纺织图案开发方法。该方法解决了传统开发方法技术壁垒高、效率低、可控性差、资源损耗大等问题,同时在深度学习层面解决了中国传统纺织图案样本少、规格杂、重意不重形等训练难点。经对比实验、主观评估与设计运用,该方法相较当前典型方法更适合应用于传统纺织图案的设计开发,具备先进性与实践价值。
关键词: 生成对抗网络;纺织品设计;深度学习;传统图案;中国书法
Abstract: With the formation of a global pattern of regional ethnic textiles, chinoiserie has become one of popular trends of global textiles. The popularity of traditional Chinese textile patterns in the global context puts forward higher requirements for its design and development methods. By taking calligraphy textile patterns as the research object, this paper comes up with a development method of traditional textile patterns based on the deep learning generative adversarial network. The proposed method solves the problems of traditional development methods, such as high technical barrier, low efficiency, poor controllability and large resource consumption. At the same time, at the level of deep learning, it solves the training difficulties of traditional textile patterns, such as few samples, miscellaneous specifications and emphasis on meaning but not shape. Through comparative experiment, subjective evaluation and design application, it is found that the proposed method is more suitable for the design and development of traditional textiles than existing typical method, and has advanced and practical value.
Key words: generative adversarial network; textile design; deep learning; traditional pattern; Chinese calligraphy
书法作为中国及中国文化辐射地区特有的文字艺术表现形式古已有之。水和墨通过不同比例的混合形成变化丰富的墨色,配合不同的留白布局表现出“气韵生动”的画面感[1]。书法元素图案打破了中国传统纹样以线为主的艺术框架,摆脱了固有的理性形态和羁绊,画面虚实相合,个性独特鲜明[2]。书法图案因其多变的艺术表现形式与深入人心的东方符号意向,在纺织品设计中应用广泛。在区域民族纺织品全球化的格局下,以书法纺织图案为代表的中国传统纺织图案在世界范围流行,对其设计开发方法提出了更高要求[3]。
随着纺织图案开发需求的不断提升,越来越多的学者投身于数字化智能纺织图案开发的研究中来[4]。……
