Neural Network Models in Fashion Design Recommendation with Interactive Visualization Methods

Abstract

As people demand for personalized clothing continues to grow, the application prospects of visual interactive design software in the clothing industry are very promising. But the traditional design process depends a lot on the work of designers, and only experienced designers can think about how style and pattern affect the overall look of clothing. As an artificial intelligence technology that has been around for a while, neural networks can be used to help with design by suggesting materials for clothes. This paper suggests a way to recommend clothing materials to designers using clothing design software by combining interactive visualization and neural network models. The method utilizes the progress and content information of the designer's current project and extracts features through a convolutional neural network (CNN) to recommend the design materials that may be needed next. The experimental results show that this method can not only provide better material recommendations for designers but also improve design efficiency and shorten the design time. Based on the Polyvore dataset, the test results show that when using this method, designers chose materials at the top of the recommendation 67% of the time, reduced the time between material selections by 43%, and reduced the average number of completed steps in the project by 56%. This method performs well in many aspects and can be integrated into clothing design software as an effective tool for clothing designers.