Data Art x ML4W: Data Selfie with BodyPix


For my data portrait, I decided to try to visualize all of the places I’ve called home these last 26 years.

I was born in Massachusetts, where I lived for roughly half of my life. Then, in short order, I moved to Toronto, Montreal, New York, San Francisco, and finally back to New York. Some places I lived for years, some for less than two months.

The house where I spent the first 12 years of my life.

The house where I spent the first 12 years of my life.

Me, aged 3, in my first house.

Me, aged 3, in my first house.

First, I created a JSON file of all my permanent addresses and calculated the days I lived there. Then I used p5.js to make a preliminary outline of the time spent in each location.

Screen Shot 2019-09-23 at 8.02.25 PM.png
Screen Shot 2019-09-23 at 8.01.25 PM.png

I replaced each colored rectangle with a photo of the house it represented. I wanted to represent the transience of these places by adding an interactive element — in order to see photos of the places I lived, the viewer has to walk across the canvas. I achieved this using the BodyPix model that uses tensorflow.js.

fast data selfie.gif

The area of each photo is proportional to the days I spent living there. In later years, the facades all start to morph and blur together.

You can see my full code here.

Caroline NeelComment