基于多尺度扩张注意力的Styleformer汽车造型设计

徐敏靖; 范永胜; 桑彬彬; 苏谦

doi:10.13568/j.cnki.651094.651316.2024.03.09.0002

您当前的位置：

首页 >

文章列表页 >

基于多尺度扩张注意力的Styleformer汽车造型设计

信息科学与技术 | 更新时间：2026-01-21

- 基于多尺度扩张注意力的Styleformer汽车造型设计
- 基于多尺度扩张注意力的Styleformer汽车造型设计
- 新疆大学学报（自然科学版中英文） 2024年41卷第5期页码：591-598
- 作者机构：
  
  重庆师范大学计算机与信息科学学院
- 作者简介：
- 基金信息：
  
  国家自然科学基金青年科学基金项目“开放环境下多粒度稳健知识获取及增量学习方法研究”（62306054）
- DOI：10.13568/j.cnki.651094.651316.2024.03.09.0002
  中图分类号： U462
- 纸质出版：2024
- 稿件说明：
移动端阅览
[1]徐敏靖,范永胜,桑彬彬,等.基于多尺度扩张注意力的Styleformer汽车造型设计[J].新疆大学学报(自然科学版中英文),2024,41(05):591-598+605.
[1]徐敏靖,范永胜,桑彬彬,等.基于多尺度扩张注意力的Styleformer汽车造型设计[J].新疆大学学报(自然科学版中英文),2024,41(05):591-598+605. DOI： 10.13568/j.cnki.651094.651316.2024.03.09.0002.

DOI：10.13568/j.cnki.651094.651316.2024.03.09.0002.

摘要

为了提升汽车设计图像的质量，开发了两个专门的数据集并通过多尺度扩张注意力机制优化现有的图像生成模型，以便捕获汽车的主体特征和结构细节．首先，创建了“WhiteCarSet”和“WhiteCarContour”数据集，分别聚焦于汽车的本体特征采集和汽车轮廓的精确收集．其次，针对图像生成模型，我们对Styleformer进行了改进，引入了多尺度扩张注意力操作，以增强模型捕获长距离依赖和理解对象全局结构的能力．在多个数据集上的实验结果显示，改进后的模型FID值相较于传统模型实现了15.57%的提升．在相同的改进模型Styleformer-M而数据集不同的情况下，使用我们开发的数据集，FID值比其它数据集提高了约22.29%.

Abstract

In order to improve the quality of automotive design images

two specialized datasets were developed

and the existing image generation models were optimized through a multi-scale dilated attention mechanism to better capture the main features and structural details of cars. Firstly

the “WhiteCarSet” and “WhiteCarContour”datasets were created

focusing respectively on the collection of the intrinsic features of the car body and the precise collection of the car contours. Subsequently

the Styleformer image generation model was refined by introducing a multi-scale dilated attention mechanism

which bolstered the model's capability to capture long-range dependencies and comprehend the global structure of objects. Ultimately

experimental results across multiple datasets demonstrated that the improved model achieved a significant enhancement in the Fréchet Inception Distance(FID)score

with an increase of up to 15.57% compared to traditional models. Moreover

utilizing our custom-developed datasets under the same enhanced model

Styleformer-M

the FID scores improved by approximately 22.29% over other datasets.

关键词

Keywords

references

杨文忠，张志豪，柴亚闯，等.基于GBRT模型的交通事故预测[J].新疆大学学报(自然科学版)(中英文)，2020,37(1):36-43.YANG W Z,ZHANG Z H,CHAI Y C,et al.Traffic accident prediction based on GBRT model[J].Journal of Xinjiang University(Natural Science Edition in Chinese and English),2020,37(1):36-43.(in Chinese)

杨文忠，杨蒙蒙，温杰彬，等.基于One Class-SVM+Autoencoder模型的车辆碰撞检测[J].新疆大学学报(自然科学版)(中英文)，2020,37(3):271-276+281.YANG W Z,YANG M M,WEN J B,et al.Vehicle collision detection based on One Class-SVM+Autoencoder model[J].Journal of Xinjiang University(Natural Science Edition in Chinese and English),2020,37(3):271-276+281.(in Chinese)

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.HU 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)

OH S,JUNG Y,LEE I,et al.Design automation by integrating generative adversarial networks and topology optimization[C]//International Design Engineering Technical Conferences and Computers and Information in Engineering Conference.American Society of Mechanical Engineers,2018,51753:V02AT03A008.

夏进军，周方舟，樊真成，等.基于深度学习的汽车造型设计工具研究[J].包装工程，2021,42(18):42-49+6.XIA J J,ZHOU F Z,FAN Z C,et al.Automotive modeling design tool based on deep learning[J].Packaging Engineering,2021,42(18):42-49+6.(in Chinese)

YU J H,LIN Z,YANG J M,et al.Generative image inpainting with contextual attention[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Salt Lake City,UT,USA.IEEE,2018:5505-5514.

HO J,JAIN A,ABBEEL P.Denoising diffusion probabilistic models[C]//Proceedings of the 34th International Conference on Neural Information Processing Systems.December 6-12,2020,Vancouver,BC,Canada.ACM,2020:6840-6851.

PARK J,KIM Y.Styleformer:Transformer based generative adversarial networks with style vector[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR).New Orleans,LA,USA.IEEE,2022:8973-8982.

VASWANI A,SHAZEER N,PARMAR N,et al.Attention is all you need[C]//31st Conference on Neural Information Processing Systems(NIPS2017).Long Beach,CA,USA.2017:5998-6008.

TOVEY M,PORTER S,NEWMAN R.Sketching,concept development and automotive design[J].Design Studies,2003,24(2):135-153.

YANNOU B,CLUZEL F,DIHLMANN M.Evolutionary and interactive sketching tool for innovative car shape design[J].Mechanics&Industry,2013,14(1):1-22.

JIAO J Y,TANG Y M,LIN K Y,et al.Dilate Former:Multi-scale dilated transformer for visual recognition[J].IEEE Transactions on Multimedia,2023,25:8906-8919.

YANG L J,LUO P,LOY C C,et al.A large-scale car dataset for fine-grained categorization and verification[C]//2015 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Boston,MA,USA.IEEE,2015:3973-3981.

KRAUSE J,STARK M,JIA D,et al.3D object representations for fine-grained categorization[C]//2013 IEEE International Conference on Computer Vision Workshops.Sydney,NSW,Australia.IEEE,2013:554-561.

HARTIGAN J A,WONG M A.Algorithm AS 136:A K-means clustering algorithm[J].Journal of the Royal Statistical Society.Series C(Applied Statistics),1979,28(1):100-108.

CANNY J.A computational approach to edge detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1986,PAMI-8(6):679-698.

GONG X Y,CHANG S Y,JIANG Y F,et al.AutoGAN:Neural architecture search for generative adversarial networks[C]//2019IEEE/CVF International Conference on Computer Vision(ICCV).Seoul,Korea(South).IEEE,2019:3223-3233.

浏览量

105

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

SWLS：沙湾市遥感图像土地覆盖语义分割数据集