Product Emotional Intelligent Design Based On Deep Learning

With the advancement of technology and the development of society,the users'emotional needs are paid more and more attention by academia and industry circles.At the same time,new requirements are put forward for the rapid iteration of products by market.Emotional design of products can give the emotional characteristics of products,bringing users a pleasant and joyful experience.Lots of research has been done at home and abroad.Product image tries to quantitatively describe the relationship between users' perceptual emotion and product,but the traditional recognition methods based on manual feature extraction can not establish accurate models.With the development of deep learning technology in recent years,significant research progress has been made in many fields by powerful and automatic learning ability from data.In this paper,an emotional intelligent design method based on deep learning was proposed.The specific work consists of two parts.The first part is about the product image recognition model based on convolutional neural network.The model can automatically identify the product image from product images.The second part is about product emotional intelligent design model based on generative adversarial network,which can automatically generate a large number of product design schemes.As a kind of typical product,shoes were taken as research sample in this paper.According to the shoes samples,a four dimensional semantic space of product image was established which contained "male-female","modern-retro","casual-formal","gorgeous-plain".Semantic differential method was applied to evaluate the product image of shoes samples to set up product image dataset as the basis of subsequent product image recognition experiment.Then the convolution neural network VGGNet 16 and AlexNet were respectively used to fine-tune the product image dataset to establish product image recognition model,which achieved high recognition accuracy in all four dimensions.The feasibility in product image recognition by convolution neural network is proved.Based on the product image recognition model,an emotional intelligent product design model was proposed based on the conditional generative adversarial network.Through the game iterative training of generator and discriminator,the model can automatically generate a large number of emotional product design schemes under the guidance of product image,which proved the feasibility of the model.The effectiveness of the model is also verified in subsequent research experiment.By using powerful self-learning ability of deep learning,this paper provides new ideas for product image recognition research and broadens the method of product emotional design at the theoretical level.At the application level,the users' emotional needs can be integrated into the actual product design,meanwhile the model can generate a large number of product design schemes which meet the demand of product image.The proposed methods provides not only reference for designers to grasp the product image but also alternative design schemes,which can shorten the product design cycle and improve the product iteration speed,having broad application prospects.