Custom Scarves Designed for Paperspace

Tomorrow we’ll be shipping out custom scarves created for Paperspace, a cloud based service solution for Machine Learning. We worked directly with CEO Dillon Erb to transform some of the machine learning models near and dear to Paperspace into four futuristic scarf designs.

The Process

Trained machine learning models for Paperspace’s Gradient product were provided as source material in .h5 and .bin formats. DataViz, a tool created by GlitchTextiles founder, Phillip Stearns, written in Processing was used to translate the raw binary data of the models into graphics. We generated several dozens of candidates in the first round.

While interesting on their own, Dillon felt that the moments of interest came from the breaks and disruption in regular patterns, and wanted a little more variation than the data in the files alone was providing, and suggested that we collage different portions of the visualizations together. Photoshop would normally be used for this, but we went with InDesign for working with multiple designs as spreads, since all we needed to do was crop and reposition different parts of our source elements.

After a final round of revisions, we had 4 designs ready to weave and send over as a proof for selection of the final design.

Though we originally set out to create one scarf design, Dillon couldn’t decide which of the four he liked best, so decided to order batches of all four and he’s keeping the proof as a blanket for around the office. Nice!

All scarves are cotton woven in the same format as our throws, cut and sewn in our studio. We even added a custom label for the project. These are going to make excellent company holiday gifts!