67 — FashionBrain Project: A Vision for Understanding Europe’s Fashion Data Universe
Read on 27 October 2017FashionBrain is a project that aims to bring big-data analysis to the generally lower-tech fashion industry by collating buyer preferences via social media, predict upcoming trends, and adjust internal product design to match that.
The envisioned project will include a variety of data acquisition systems — for instance, photo-based product search, matching known product catalogs for those seen in an image — as well as better search tools to navigate existing retailer websites and digital warehouses.
One of the main tasks of this project is to design a taxonomy for the categorization and organization of fashion-related data and databases. This taxonomy presents the novel contribution of eliminating the usual top-level gender-separation of objects, which reduces redundancy and is also a dumb idea anyhow.
The trend of data-access from this database over time can be used as a modelable time-series, which FashionBrain aims to predict to anticipate fashion trends.
To be completely transparent, I had a really hard time following the technical aspects of this paper; there’s a lot of buzzword salad (“Hadoop, MapReduce, key-value stores and distributed (R)DBMSs are also often used for some data analytics…“) and not too much technical detail on implementation. But I’m looking forward to seeing how this group uses MonetDB to address some of their scaling challenges.