A Stitch Fix Data Scientist introduced Deep Style that can one day boost the combination of their recommender system and stylists to provide even better personalization than they’re able to today.
Stitch Fix was founded in 2011 with a goal to make shopping fun, effortless and empowering for busy women on the go. Their trained personal stylists are armed with the best tools and technology to help them hand-select clothing and accessories that fit their customers lifestyle and body shape.
In looking to improve their client’s experience, they are developing algorithms to help their stylists make better fixes through a robust recommendation system. A recent post on their engineering blog takes a look specifically at how to build an automated process using photographs of clothing to quantify the style of some of items in their clothing collection. Then they will use this style model to make new computer generated clothing.
Their deep learning network learns how to represent the important features in an image (in their case attributes of style) without ever being explicitly told what those representations should look like. The auto-encoder models will be used to effectively query a computer to design new clothing.
The team published their open source code on GitHub which can be used for using a variational auto-encoder for latent image encoding and generation. Don’t forget about the tip they included, “If you have an CUDA capable NVIDIA GPU, use it. The model can train over 10 times faster by taking advantage of GPU processing.”