Text to Image Generation of Fashion Clothing

The recent exponential advancements in Generative Adversarial Networks have a major contribution in development of dynamic, highly compelling and fascinating image generation. This paper proposes an approach a framework that will accept text input from the user about the fashion pattern and the model will generate images of fashion clothing based on the text input. The model proposed can assist people be their own designers for creating a range of fashion clothing for themselves using the power of Deep Learning and Generative Adversarial Networks (GAN). Paper aims to engender realistic images using StackGAN model consisting of 2 stages, using highly-annotated dataset Fashion-Gen [4]. The initial stage generates low-resolution contours of the crude shape around a real-world human model based on the provided text description by the user. The second stage inputs stage one output to yield high resolution images with fascinating and detailed style-mapping to the text input. As compared to the precedent datasets, the dataset suggested to use is Fashion-Gen which is simply in resemblance with real-world and consists of High Definition quality images which would further lead to accurate explications. It is an exemplar till date as it provides a conglomeration of features with regards to parameters like angles, multitudinal categorization and sub-categorization of clothing. Such richly annotated dataset enables us to deploy powerful StackGAN model in plausible image generation. The model is then deployed on cloud for training and further use in the next stage of building an interactive web interface for the user to use the model efficiently.