Application for Generating Shoe Design from Shoe Sketches

Abstract

Shoe design plays a vital role in the shoe industry in designing colorful and attractive shoes. It has important applications in the shoe industry since there is a frequent disconnection between the consumer and the shoe designer while designing a straightforward simple shoe and the high authorities in the shoe industry typically reject the proposed design of the shoe made by the shoe designer. Therefore, the shoe sketch is useful in creating the colorized design of the shoe. Designing the shoe of a simple shoe requires plenty of expertise in the shoe designing domain in the shoe industry but experts in this domain face hindrances while designing the shoe from scratch. Professional shoe designers are required to create such sketches, but it is always possible that it would not always be accessible to complete the acceptable shoe design work provided to them. With the help of this project, it will be possible for even a novice user to sketch a specific pair of shoes. To do this, a database of simple or rough sketches is created. Realistic shoe pictures are created using these drawings. This sketch-to-photo translation is performed using Generative Adversarial Network (GANs) models. In the context of this specified domain, we have developed a second variation of conditional generative adversarial networks (cGANs). The model in question is a Cycle-GAN model that training has been done on the edged2shoe dataset for this project