基于深度学习的汽车造型设计工具研究

Automotive Modeling Design Tool Based on Deep Learning

中文关键词: 汽车造型  深度学习  设计工具  StyleGAN2  标签体系

英文关键词: automotive modeling  deep learning  design tools  StyleGAN2  label system

中文摘要:

      目的 研究基于深度学习的汽车造型设计工具。方法 使用交叉分类法对数据集进行整理并标记，形成全标签数据库，结合基于StyleGAN2的汽车图像生成模型，形成基于深度学习的汽车造型设计工具，并通过对比实验检验AI设计工具的生成速率和质量优劣。结果 通过AI生成汽车造型的速率是一般设计师的数千倍，在质量比分中，AI生成的质量高于一般设计师。结论 基于深度学习的汽车造型设计工具可作为汽车造型设计的新工具。

英文摘要:

      This paper studies the automotive modeling design tools based on deep learning. The cross-classification method is used to arrange and mark the data set to form a full-label database. In combination with styleGAN2-based automobile image generation model, a vehicle modeling design tool based on deep learning is formed, and the generation rate and quality of AI design tools are tested through comparative experiments. Through the AI design, the automobile modeling generation rate is thousands of times higher than that of ordinary designers, and in the quality score, the quality generated by AI is higher than that of ordinary designers.By partially‍ and controllable design target modeling, it is proved that the automobile modeling design tool based on deep learning is a new tool, and the rate of generating images by the new tool is faster than that of the designer’s hand drawing.