SizeNet: Weakly Supervised Learning of Visual Size and Fit in Fashion Images
Finding clothes that fit has increasingly become a hot topic in the fashion e-commerce industry. This is mainly due to customer frustrations and the large ecological and economic footprint that size-related returns has on the industry. Most approaches addressing this problem are based on statistical methods which rely on historical data and suffer from the cold start problem where thousands of new articles appear on shopping platforms every day. In this talk, we discuss SizeNet - a weakly-supervised teacher-student approach that leverages the power of statistical models combined with the rich visual information from article images to learn visual cues for size and fit characteristics, and is capable of tackling the challenging cold start problem. The strong advantage of this approach is demonstrated through experiments performed on thousands of textile garments from hundreds of different brands.
Nour works as Senior Applied Scientist at Zalando on machine learning-driven solutions to tackle the complex challenge of delivering size and fit advice to millions of customers. Her focus is in utilizing state-of-the-art computer vision and AI techniques to leverage visual cues present in fashion images. Nour co-organized the 7th Women in Computer Vision workshop at the Computer Vision and Pattern Recognition 2020 conference. The workshop goal is to raise visibility of female computer vision researchers and share experience and career advice between students and professionals. She completed a Master degree in Computer Science from Saarland University - Max Planck Institute for Informatics.