Visually-Aware Fashion Recommendation

Last updated on Mar 7, 2023
PDF Code Slides
Visual Depiction of a simple use-case present in the dataset

The work is divided into three main components:

  1. Carry out Exploratory Data Analysis (EDA) on the Amazon Fashion dataset by plotting several graphs and plots that showcase different aspects of the datasets.
  2. Use a simple Matrix Factorization-based Latent Factor Model to generate baseline results.
  3. Build a Visually-Aware Fashion Recommender System by employing Siamese Convolutional Network to jointly train the image representation of the fashion apparel and the recommender system. Bayesian Personalized Ranking is used to obtain a ranked list of items for each user. The visually-aware model achieves an AUC score of 0.7112, increasing the baseline score by 0.18.
Recommender Systems Machine Learning Deep Learning
Niraj Yagnik
Niraj Yagnik
Computer Science Grad Student and Researcher

My research interests include but are not limited to Recommendation Systems, Natural Language Processing, Machine Learning, and Social Computing.