基于深度学习的产品概念草图生成设计研究  

Product Conceptual Sketch Generation Design Using Deep Learning

作　　者：李雄 苏建宁[1,3] 张志鹏 LI Xiong;SU Jianning;ZHANG Zhipeng(School of Mechanical&Electronical Engineering,Lanzhou University of Technology,Lanzhou 730050;School of Bailie Mechanical Engineering,Lanzhou City University,Lanzhou 730070;School of Design Art,Lanzhou University of Technology,Lanzhou 730050)

机构地区：[1]兰州理工大学机电工程学院,兰州730050 [2]兰州城市学院培黎机械工程学院,兰州730070 [3]兰州理工大学设计艺术学院,兰州730050

出　　处：《机械工程学报》2023年第11期16-30,共15页Journal of Mechanical Engineering

基　　金：国家自然科学基金资助项目(52165033)。

摘　　要：概念草图设计作为人类的高阶视觉认知活动,是辅助设计师记录、构思、创造和评估想法的重要手段,对生成创新概念具有积极影响。为模拟设计师的这种高阶视觉认知行为,实现智能化辅助创意草图设计,提出一种基于深度学习的产品概念草图智能生成设计集成方法框架,包括端到端的草图设计GAN(Sketch2Render-GAN)和草图神经风格迁移网络(Sketch-NST)两个核心模块。前者实现概念草图生成与渲染,后者执行草图风格特征变换。分别以手电钻和自行车头盔为实验对象进行了验证,结果表明该方法框架可快速获得大量具有创新概念的草图,并实现草图自动渲染及风格变换。有助于辅助设计师在视觉认知层面突破设计固化,提高设计效率。此外,为改善工业设计师与AI模型间的人机设计协作模式,还开发了智能草图设计生成器(S-SDG_v0.1),从而有效降低设计师应用智能算法辅助设计的门槛。Concept sketching,as a higher order human visual cognitive activity,is an important tool to assist designers in recording,ideating,creating,and evaluating ideas and has a positive impact on the generation of innovative concepts.In order to simulate this higher order visual cognitive behaviors of designers and to achieve intelligent assistance in creative sketching,a deep learning-based design integrated framework for intelligent generation of product concept sketches is proposed,which includes two core modules:an end-to-end sketch design GAN(Sketch2Render-GAN)and a sketch-neural style transfer network(Sketch-NST).The first module implements sketch generation and rendering,while the second performs sketch style features transformation.The hand drill and bicycle helmet were used as design objects respectively,and experimental results show that the proposed approach framework can quickly obtain many innovative concept sketches and implement automatic sketch rendering and style transformation.The findings also show that the approach framework helps designers to break through design solidification at the visual perception level and increase design efficiency.Furthermore,a smart-sketch design generator(S-SDG_v0.1)was developed to facilitate human-machine design collaboration between designers and AI models,which effectively reduces the threshold of designers to apply intelligent algorithms to assist design.

关 键 词：视觉认知 深度学习 生成设计 草图设计 生成对抗网络 

分 类 号：TG156[金属学及工艺—热处理]