Fashion Recommender
Introduction
Nowadays AI plays a predominant role in any industry to increase productivity. The fashion industry is also adapted AI to solve varieties of problems that cannot be solved through human intervention. Another US entrepreneur wanted to enchase the user experience of his clothing business by integrating AI. The customer’s requirement was to build a cloth recommendation system when a user uploads an image of an outfit. The inventory is already managed by their legacy system. We had to develop the AI integration part separately as an additional service.
Context
After conducting deep research on fashion recommendation systems, our development team decided to use CNN (Convolutional Neural Network) model along with five layers as the backend solution. This performs the image feature extraction and classification part and has used image embedding from each network to compare. We have used a library called, ‘Fastai’ to perform all the model training-related tasks. The client asked us to focus mainly on attributes such as fabric, texture, shape, and part. After the model training, we had again train the model with the owner’s inventory images and map them with the specific label. Then whenever a customer uploads an image the model identifies the attributes and recommends the most similar outfits from the store.