Strange Loop

Computer Vision and NLP for Multi-Task Fashion Modeling

Shoprunner aggregates millions of products from 140 retailers which represent thousands of brands. In order to make these products findable and searchable by users it is important for Shoprunner to be able to standardize the attributes (style, color, pattern etc) of these millions of products. Even after defining what attributes to model, choosing the best way to predict attributes is difficult because every product can be represented in a variety of forms such as images, product description, title,and brand name. These different data representations each have their strengths and weaknesses. Images encode information such as color and pattern well while other attributes related to length and cut may be well captured in text descriptions. This session will go through the multi-task learning ensemble that the Data Science team at Shoprunner has built using both custom multi-task CNNs for images and fine-tuned Bert model for text classification in Pytorch for attribute modeling.

Michael Sugimura

Michael Sugimura


I am a Senior Data Scientist who focuses on computer vision applications at Shoprunner. In particular I work on building attribute models using multi-task ensembles and building detection and segmentation models applied to the e-commerce space. When not working on data science at work, Michael is usually working on side data science projects and blogging about them, doing martial arts (Wing Chun Kung Fu, Muso Jikiden Eishin Ryu Iaijutsu, Kendo), or cooking.