Application of DCGAN to Design Model of Automobile Modeling
-
摘要: 为了进一步提升设计方案的效率,缓解人工劳动的强度和压力,通过计算机自动设计达到快速生成创新造型设计方案的目的,提出一种基于深度卷积生成对抗网络(Deep convolution generative adversarial networks,DCGAN)的汽车造型设计模型。该模型通过构建基础产品造型模块的设计方案数据集,利用深度学习的DCGAN对数据集进行训练以提高设计方案图像质量。最后,将生成的两种不同种类汽车造型方案与专家方案进行满意度对比,结果显示模型生成的方案能够得到与专家设计方案相近的评分,证明了所提模型的有效性和合理性。
-
关键词:
- 造型设计 /
- 产品设计 /
- 深度卷积生成对抗网络 /
- 汽车设计
Abstract: In order to further improve the efficiency of design schemes and reduce human labor, through the automatic design of computers to achieve the purpose of quickly generating innovative styling design schemes, design model for automobile modeling with deep convolution generative adversarial networks (DCGAN) was proposed. The model builds the design solution data set of the basic product modeling module, and the design plan image quality was improved by using the deep learning DCGAN to train the data set. Finally, comparing the generated two different types of car styling schemes with the expert schemes for satisfaction, the results show that the scheme generated via the present model can get a score similar to the expert design scheme, the validity and rationality of the present model were proved. -
表 1 One-hot编码结果
家用汽车车型 SUV车型 01 10 表 2 训练注意细节
步骤 细节 1 实验中进行的预处理只是将训练图像缩放到tanh激活函数的范围[-1, 1]内 2 最小批训练, 批大小是16 3 所有权值都是从以0为中心的正态分布初始化的, 标准差为0.02 4 在LeakyReLU中, 所有模型的斜率都设置为0.2 5 DCGAN使用Adam优化器来加速训练 6 Adam优化器的学习率设置为0.000 2 7 将动量参数β1从0.9降为0.5以防止震荡和不稳定 表 3 数据集组成数量
名称 家用汽车车型数量 SUV车型数量 原始数据集 8 565 2 771 网络爬虫方法 1 090 1 531 传统方法 4 360 11 084 总计 14 015 15 386 训练集 11 212 12 308 测试集 2 803 3 077 -
[1] JANSSEN P. A generative evolutionary design method[J]. Digital Creativity, 2006, 17(1): 49-63 doi: 10.1080/14626260600665736 [2] YOUMANS R J. The effects of physical prototyping and group work on the reduction of design fixation[J]. Design Studies, 2011, 32(2): 115-138 doi: 10.1016/j.destud.2010.08.001 [3] VISWANATHAN V, LINSEY J S. Physical models and design thinking: a study of functionality, novelty and variety of ideas[J]. Journal of Mechanical Design, 2012, 134(9): 091004 doi: 10.1115/1.4007148 [4] SU J N, ZHANG S T. Research on product shape innovation design method with Human-Computer interaction through genetic algorithm[C]//Proceedings of the 2010 IEEE 11th International Conference on Computer-Aided Industrial Design & Conceptual Design. Yiwu: IEEE, 2011: 301-305 [5] 苏建宁, 张秦玮, 吴江华, 等. 产品多意象造型进化设计[J]. 计算机集成制造系统, 2014, 20(11): 2675-2682 doi: 10.13196/j.cims.2014.11.004SU J N, ZHANG Q W, WU J H, et al. Evolutionary design of product multi-image styling[J]. Computer Integrated Manufacturing Systems, 2014, 20(11): 2675-2682 (in Chinese) doi: 10.13196/j.cims.2014.11.004 [6] 王亚辉, 余隋怀, 陈登凯, 等. 基于深度学习的人工智能设计决策模型[J]. 计算机集成制造系统, 2019, 25(10): 2467-2475 doi: 10.13196/j.cims.2019.10.006WANG Y H, YU S H, CHEN D K, et al. Artificial intelligence design decision making model based on deep learning[J]. Computer Integrated Manufacturing Systems, 2019, 25(10): 2467-2475 (in Chinese) doi: 10.13196/j.cims.2019.10.006 [7] 丁满, 李明惠, 赵芳华, 等. 基于生成对抗网络的产品色彩智能设计方法[J]. 机械设计, 2019, 36(11): 133-138 https://www.cnki.com.cn/Article/CJFDTOTAL-JXSJ201911022.htmDING M, LI M H, ZHAO F H, et al. Product color intelligent design method based on generative adversarial network[J]. Journal of Machine Design, 2019, 36(11): 133-138 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JXSJ201911022.htm [8] 唐贤伦, 杜一铭, 刘雨微, 等. 基于条件深度卷积生成对抗网络的图像识别方法[J]. 自动化学报, 2018, 44(5): 855-864 doi: 10.16383/j.aas.2018.c170470TANG X L, DU Y M, LIU Y W, et al. Image recognition with conditional deep convolutional generative adversarial networks[J]. Acta Automatica Sinica, 2018, 44(5): 855-864 (in Chinese) doi: 10.16383/j.aas.2018.c170470 [9] ZHU L, CHEN Y S, GHAMISI P, et al. Generative adversarial networks for hyperspectral image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(9): 5046-5063 doi: 10.1109/TGRS.2018.2805286 [10] LIN C H, YEN H R, CHUANG S C. The effects of emotion and need for cognition on consumer choice involving risk[J]. Marketing Letters, 2006, 17(1): 47-60 doi: 10.1007/s11002-006-4146-2 [11] 国务院发展研究中心产业经济研究部, 中国汽车工程学会, 大众汽车集团(中国). 中国汽车产业发展报告(2013)[M]. 北京: 社会科学文献出版社, 2013Industrial Economic Research Department, Development Research Center of the State Council, China Automotive Engineering Society, Volkswagen Group (China). Annual report on automotive industry in China (2013)[M]. Beijing: Social Sciences Academic Press, 2013 (in Chinese) [12] CAGAN J, VOGEL C M. Creating breakthrough products: innovation from product planning to program approval[M]. Upper Saddle River: Prentice Hall PTR, 2002 [13] 郭磊, 阿丽莎, 胡钢. 双三次Q-Bézier曲面拼接的车身设计[J]. 机械科学与技术, 2017, 36(1): 114-118 doi: 10.13433/j.cnki.1003-8728.2017.0117GUO L, A L S, HU G. Designing car body with blended cubic Q-Bézier surfaces[J]. Mechanical Science and Technology for Aerospace Engineering, 2017, 36(1): 114-118 (in Chinese) doi: 10.13433/j.cnki.1003-8728.2017.0117 [14] 卢兆麟, 程若丹, 石清吟, 等. 基于自然语言处理的汽车造型风格推导与评价[J]. 汽车工程, 2016, 38(5): 553-560 https://www.cnki.com.cn/Article/CJFDTOTAL-QCGC201605006.htmLU Z L, CHENG R D, SHI Q Y, et al. Vehicle styling feature derivation and evaluation based on natural language processing[J]. Automotive Engineering, 2016, 38(5): 553-560 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-QCGC201605006.htm [15] 王波, 罗际, 朱睿. 汽车造型设计的线-型分析方法[J]. 汽车工程, 2010, 32(6): 470-476 https://www.cnki.com.cn/Article/CJFDTOTAL-QCGC201006002.htmWANG B, LUO J, ZHU R. Line-style analysis technique for car styling[J]. Automotive Engineering, 2010, 32(6): 470-476 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-QCGC201006002.htm [16] 胡伟峰, 赵江洪. 用户期望意象驱动的汽车造型基因进化[J]. 机械工程学报, 2011, 47(16): 176-181 https://www.cnki.com.cn/Article/CJFDTOTAL-JXXB201116029.htmHU W F, ZHAO J H. Automobile styling gene evolution driven by users' expectation image[J]. Journal of Mechanical Engineering, 2011, 47(16): 176-181 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JXXB201116029.htm [17] MCCORMACK J P, CAGAN J, VOGEL C M. Speaking the Buick language: capturing, understanding, and exploring brand identity with shape grammars[J]. Design Studies, 2004, 25(1): 1-29 [18] 姚干勤, 薛澄岐, 王海燕, 等. 基于意象认知的客车造型设计方法[J]. 东南大学学报(自然科学版), 2016, 46(6): 1198-1203 https://www.cnki.com.cn/Article/CJFDTOTAL-DNDX201606015.htmYAO G Q, XUE C Q, WANG H Y, et al. Design method for coach styling design based on image cognition[J]. Journal of Southeast University (Natural Science Edition), 2016, 46(6): 1198-1203 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-DNDX201606015.htm [19] GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial nets[C]//Proceedings of the 27th International Conference on Neural Information Processing Systems. Montreal: ACM, 2014: 2672-2680 [20] 朱斌. 基于深度学习的产品情感化智能设计[D]. 杭州: 浙江大学, 2018ZHU B. Product emotional intelligent design based on deep learning[D]. Hangzhou: Zhejiang University, 2018 (in Chinese) [21] HOFER A P, HALMAN J I M. The potential of layout platforms for modular complex products and systems[J]. Journal of Engineering Design, 2005, 16(2): 237-255 [22] 郏维强, 刘振宇, 刘达新, 等. 基于模糊关联的复杂产品模块化设计方法及其应用[J]. 机械工程学报, 2015, 51(5): 130-142 https://www.cnki.com.cn/Article/CJFDTOTAL-JXXB201505017.htmJIA W Q, LIU Z Y, LIU D X, et al. Modular design method and application for complex product based on fuzzy correlation analysis[J]. Journal of Mechanical Engineering, 2015, 51(5): 130-142 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JXXB201505017.htm [23] 侯亮, 唐任仲, 徐燕申. 产品模块化设计理论、技术与应用研究进展[J]. 机械工程学报, 2004, 40(1): 56-61 https://www.cnki.com.cn/Article/CJFDTOTAL-JXXB200401011.htmHOU L, TANG R Z, XU Y S. Review of theory, key technologies and its application of modular product design[J]. Chinese Journal of Mechanical Engineering, 2004, 40(1): 56-61 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JXXB200401011.htm [24] 单春来, 李永成, 侯文彬. 基于模块化设计的车身装配结构优化[J]. 汽车工程, 2018, 40(5): 617-624 https://www.cnki.com.cn/Article/CJFDTOTAL-QCGC201805019.htmSHAN C L, LI Y C, HOU W B. Optimization of car body assembly structure based on modular design[J]. Automotive Engineering, 2018, 40(5): 617-624 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-QCGC201805019.htm [25] 宋守许, 刘志峰, 唐涛, 等. 模块化产品评价方法研究[J]. 农业机械学报, 2004, 35(5): 181-184+189 https://www.cnki.com.cn/Article/CJFDTOTAL-NYJX200405044.htmSONG S X, LIU Z F, TANG T, et al. Evaluation method of modular products[J]. Transactions of the Chinese Society of Agricultural Machinery, 2004, 35(5): 181-184+189 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-NYJX200405044.htm [26] 单泉, 陈砚, 雷毅, 等. 面向改型设计的模块化产品评价方法[J]. 计算机集成制造系统, 2008, 14(3): 438-443 https://www.cnki.com.cn/Article/CJFDTOTAL-JSJJ200803004.htmSHAN Q, CHEN Y, LEI Y, et al. Modular product evaluation method oriented to innovation design[J]. Computer Integrated Manufacturing Systems, 2008, 14(3): 438-443 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JSJJ200803004.htm [27] CHEN Y L, LIN H L, LIU D S. Optimal configuration design of open architecture products considering product adaptability and diversity of modules[J]. Procedia CIRP, 2018, 76: 73-78 [28] LI X Z, QIU S Q, MING H X G. An integrated module-based reasoning and axiomatic design approach for new product design under incomplete information environment[J]. Computers & Industrial Engineering, 2019, 127: 63-73 [29] TORSTENFELT B, KLARBRING A. Conceptual optimal design of modular car product families using simultaneous size, shape and topology optimization[J]. Finite Elements in Analysis and Design, 2007, 43(14): 1050-1061 [30] KRAUSE J, STARK M, DENG J, et al. 3D object representations for fine-grained categorization[C]//Proceedings of the IEEE International Conference on Computer Vision Workshops. Sydney: IEEE, 2003: 554-561 [31] VIJAYANARASIMHAN S, GRAUMAN K. Efficient region search for object detection[C]//Computer Vision and Pattern Recognition (CVPR). Colorado Springs: IEEE, 2011: 1401-1408 [32] ARBELÁEZ P, HARIHARAN B, GU C H, et al. Semantic segmentation using regions and parts[C]//2012 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Providence: IEEE, 2012: 3378-3385 [33] RADFORD A, METZ L, CHINTALA S. Unsupervised representation learning with deep convolutional generative adversarial networks[C]//4th International Conference on Learning Representations. San Juan: ICLR, 2016: 1-16 [34] KINGMA D P, BA J. Adam: a method for stochastic optimization[C]//3rd International Conference on Learning Representations. San Diego: ICLR, 2015: 1-15 [35] SINGH A, KUMAR S. Dual hesitant fuzzy set and intuitionistic fuzzy ideal based computational method for MCGDM problem[J]. International Journal of Natural Computing Research, 2018, 7(3): 17-41 [36] MIRZA M, OSINDERO S. Conditional generative adversarial nets[J]. arXiv preprint arXiv: 1411.1784, 2014 [37] BERTHELOT D, SCHUMM T, METZ L. BEGAN: boundary equilibrium generative adversarial networks[EB/OL]. (2017-12-06)[2020-03-25]. https://arxiv.org/abs/1703.10717 [38] ARJOVSKY M, CHINTALA S, BOTTOU L. Wasserstein GAN[EB/OL]. (2017-05-31)[2020-03-29]. https://arxiv.org/pdf/1701.07875.pdf