Several weeks ago Google Arts & Culture and the Bussiness of Fashion (BoF) released the interactive visualization ‘Runway Palette‘ by Cyril Diagne. The tool was released at the BoF VOICES 2019 conference and boasts over 144,000 looks from more than 3,800 fashion shows and 992 designers.

[Experiments with Google] came together with The Business of Fashion to view their collection of 140,000 photos of runway looks from almost 4,000 fashion shows around the world. If you could attend one fashion show per day, it would take you more than 10 years to see them all.

The interactive tool is both visually stunning and highly informative — It is also free to use and quite easy to get lost in! If you’re curious about which designers influenced your closet you can upload your own photo and find looks that match.

A number of processing steps went into creating RunwayPalette including: removal of background from the images, UMAP generation, training of performant models and parallel processing. If you’d like to learn more about the project, check out Diagne’s seven-part summary on Twitter. You can also check out this Colab using GeomLoss to map the UMAP to a grid (3D visualization to 2D).

Written by Rebecca Minich, Product Analyst, Data Science at Google. Opinions expressed are solely my own and do not express the views or opinions of my employer.