Introduction

Shirley Wu is a self-described award winning creative focused on data-driven art. While perusing her work, one particular visual stood out to me.

Description

This is the build in Shirley's words:

Ever since graduating college, I've made it a goal to travel abroad at least once a year. Over the years, I've accumulated almost 4000 images over the course of 13 trips. I've taken those pictures and extracted the five primary colors from each, for a total of roughly 20,000 colors.

I find the idea both creative and playful. Motivated, I decided to create my very own color wheel.

Process

Here, I sketch the procedure. I'll also include possible customizations.

Procedure

There are roughly four steps involved:

1. Select Photos
2. Identify Dominant Colors
3. Plot on a Make-Shift Color Space
4. Add Labels

We start by collecting photos from a trip. There is no need to be picky here — anything and everything should be included! From here, we need to identify the dominant colors present in each photo. It turns out that ''dominant colors'' are not well defined, so there is some flexibility here. Finally, we plot on a color space of our choosing. Optionally, we can add labels representing key demarcations (e.g. days, changes in scenery, etc) present in the photos.

Customizations

As mentioned above, there is a lot of opportunity to experiment. Here, I make use of the convert method implemented in the Python Imaging Library. Discussion online suggests that it internally uses a variant of K-Means Clustering to produce a color pallete.

Examples

A year ago, my family celebrated a cousin's wedding. From preparation to execution, the event took up 5 days. Here, I cobbled together all the photos I could and made the following color wheel.

As noted above, this is great for album covers, posters, and even Instagram stories!