The following images document the first steps in the creation of this visualization.
I was exploring different styles in order to understand which one could be the most informative and easy to read.
I was pretty convinced that a dark scheme would work.
Its contrast made it more simple to view the bigger circles, but was lacking in showing the amount of people involved in the crashes. Also, it was still hard to see all the data about the crashes without injuries, which are a huge part of the total crashes that happen every month.
So I started experimenting more with more 3d elements.
Finally, after many revisions, I left behind the certainties of a dark theme and I switched to a light one:
It was better, but I was still not happy.. it was too nice to the eye, but the topic was harsh. It was not something to be presented in a "nice" way.
Plus, the DOF made the image more pleasing but less clear, so at the end I decided for a more neutral style: less aesthetically pleasing, but more precise.
The full Houdini graph responsible for the generation of the Viz is the following:
It all begins with loading the csv data of the crashes in SOP:
These are the tweakable parameters (it's all procedural, baby) that I exposed on a null node:
Here are some screen grabs of the Houdini viewport:
The streets data were downloaded from a GeoFabrik dump of central Italy, then converted to GeoJSON format and finally imported inside Houdini through a custom python node (GeoJSON LineString):
Here's the final timelapse rendered with Mantra:
