One of the biggest challenges with large datasets is trying to understand and comprehend them.Visualisation can offer insight into large datasets but it can be very challenging. It requires an understanding of two different fields, the computational and the artistic. I have only started to learn it through continuous iteration and revision. My first attempts at data visualisation were time lapses, as shown on my other page, but this progressed with my knowledge of animation and rendering improved.
I will highlight some of the more interesting approaches below:
Graphing the development of software through a version control system
Any folder structure can be thought of as a tree. As file as are added removed we can visualise those changes. This approach used gource to visualise 4 years of software development on a project.
Merging different datasets with different resolutions
I soon found that drawing everything at once killed the machine. The ability to render different datasets at different times and in different resolutions is necessary for detailed results.
Visualising full wave form data
When a laser scanner records what the light bounced off (a return) it sometimes gets multiple values. Normally these are averaged out and a single point is recorded. This throws away a lot of useful data that could be used for further analysis. The problem is that when we record that additional data it starts adding up to terabytes of points that cannot be visualised through traditional vertex animation techniques. Instead we converted the point cloud into a format normally used with MRI and medical data for recording scan results. This allowed us to examine the data in near realtime:
Visualising to give an idea of scale and resolution
I love visualising survey data but never more so when it emphasises of quality and scale of the results. Some additional tips and tricks are to control the lighting and depth of field to give that impression you are looking at a model.
My Christchurch visualisation demonstrates several of these techniques, including depth of field and enabling and disabling textures
And then there are the examples where you move through the survey to impress the sense of scale.
This one just keeps going and going: