Deep Learning: A New Way of Looking into Fashion

Currently, Fashion-Fit using AI focuses on finding the perfect fit for users using GAN’s (General Adversarial Net¬works) and various methods proposed in the Artificial Intelligence Industry. However, due to changes in fashion trends and various preferences of the "type of fit" users want for their clothing the prior methodology is not meant for the perfect fit. These days a lot of people wear tighter or even get an oversized fit based on the user’s preference. In this paper we use Deep Learning techniques like CNNs and GANs to implement neural body fit by looking at various human body poses from large datasets and concatenating this data to make an accurate "desired" fit based on different human poses. This will set a new benchmark for virtual try-on. In the model being proposed, we are building Convolutional Neural Networks (CNNs) for body segmentation to make it easier to map a human pose. We will then map the human body using CNNs but not necessarily to find the "perfect fit". The CNNs will be used to make measurements that are similar to sizes available within the fashion industry allowing users to see how they would fit within those garments virtually. We will then use 2D renders, using uploaded pictures of the user from 4-5 different angles in order to allow the user to see how he/she would look in the outfit of their choice. These methods can give an accurate prediction helping every fashion e-commerce business. Personalized size and fit recommendations are critical for a fashion e-commerce business. Predicting the optimal fit increases consumer happiness while lowering expenses associated with size-related returns.