Building a fashion search engine with deep learning

More specifically, we will design a model that takes a fashion image as input (the image on the left below), and outputs a few most similar pictures of clothes in a given dataset of fashion images (the images on the right side). We will train such neural networks to classify the clothing images into 6 categorical labels and use the feature layer as the deep features of the images. To classify the images, we use a model based on deep residual networks (ResNet).

Source: blog.floydhub.com