DeepWear refers to research on proposing new designs by identifying the features of one or more fashion brands, using deep learning (DCGANs). Although an outstanding designer produces many fans, there is a limit to the amount of work that the person can undertake. However, a computer incorporating a deep-learning neural network can learn by observing the products, and much like an outstanding assistant, is able to output many ideas for new products. Moreover, by learning the features of many designers, it is also able to propose new brands by blending the features of multiple brands. The intelligence of computers based on learning of fashion design, can be used for the generation of ideas by designers, and can also provide hints for new products to manufacturers. Moreover, it also enables end users to place order for garments that match both the features of a favorite brand and the users’ personal preferences.
■ Case 1.
You have commenced a service in which the product is made in your factory after obtaining the image of the original design from the customer, but it is unable to optimize the designs. You think that it will be difficult to sustain the brand as a service if quality variations occur due to the customer’s design experience. You are looking for ways to optimize the design based on the ideal brand image of the customer and the closer brand image that the provider wants to realize before cutting the pattern and sewing the garment.
Output of several images as proposals for new design after learning brand features
In the general production method, the patterner generates the production drawing based on the 2D images created by the fashion designer, and then a worker stiches together the cloth pieces cut for each part to complete the garment. In our system, for a specific brand for which deep learning has been undertaken to learn the features of the designer’s creations, we have mechanized the creation of sketches that constitute the basic data necessary in generating patterns. We are using deep convolutional generative adversarial networks (DCGANs), as the images need to be generated at a resolution appropriate for the patterners to use.
計算機・デザイナー・顧客の知能を融合させ、新たな服飾⽂化を構築
Establishing a new dress culture by fusing intelligence of computers, designers, and customers
We are intending the social implementation of a program to generate new garments from design images by concurrently conducting research on the process of user evaluation of pattern making, sewing, and the finished product. Moreover, it is possible to apply this to a service for directly ordering tailored brand products matching personal preferences, by developing software to enable latent variable adjustments by the customers themselves.