Research on Generative Design of Analogical Reasoning Car Styling Based on 3D Generative Adversarial Networks
Analogical reasoning refers to extracting elements from familiar fields or previous experience and applying them to new concepts.It is one of the most commonly used design methods by designers in the stage of product shape design.Taking automobile exterior design as the carrier,starting from the logical thinking process of design and the characteristics of 3D generative adversarial model,this paper proposes the idea and method of automatic model generation based on analogical inference design,combined with the IM-GAN obtained by embedding the implicit field decoder(IM-NET)into the generative adversarial network(GAN),to realize the analogous fast modeling.Firstly,the car and aircraft models in the public data set ShapeNet are used to build a design object and inspiration source data set.Then,after the model training is completed,by interpolating in the latent space of the generator,an intermediate interpolation style of any degree between any two existing products can be obtained,and an innovative scheme for analogy design of product 3D shape can be explored.Finally,from the designer's point of view,this paper compares the implicit function representation with point cloud and voxel representation,and discusses the advantages and disadvantages of different geometric representation methods in assisting designers in shape design.The experimental results show that the model based on implicit function representation can obtain better auxiliary design effects and inspire designers to have more product modeling ideas than the other ones.
analogical reasoninggenerative adversarial networkscar shape designgenerative design system
NETL
NSTL
万方数据
