How do you visualise big data?

"There is no denying that the age of big data is upon us. […] We all collect, produce and share more information than it is possible for us to ever use. We are almost drowning in this ocean of digital data, which we use to try to give meaning to our lives, and also to attempt to understand the meaning of life in general." - NRC Handelsblad, 22 August 2018

How can you calculate realistic shadows for a computer game in a fraction of a second? And how can you visualise major flooding simulations? Elmar Eisemann designs these types of applications, while also ensuring that they do not overburden your computer’s processing power.

Images are very powerful: people are naturally drawn to them.

“Images are very powerful. People are naturally drawn to them, much more than with text. That is why it is good to use images to attract or hold attention, particularly when conveying information. That is exactly what my research group does: we develop algorithms to create images for various applications.

Compare it to the drawings for assembling an IKEA wardrobe. Those drawings need to be schematic. These types of unrealistic images are often also important when you work with big data. Imagine that you want to track down harmful cells in a dataset of millions of other cells, then you will want to group the cells with deviant values into handy clusters. And ideally, to mark them with a colour to make them stand out. Making these images too detailed would only make things confusing. With ‘visual analytics’, we discover trends in data. This is an enormous technological challenge, as the data sets are massive. That is why we create algorithms that are easily scalable.

Incidentally, you can also use data to make realistic images. In the context of water management, for example. If you let a group of people discuss abstract flooding maps, it will primarily be the well-informed experts who dominate the discussion. After all, they know the best way of interpreting the maps and are able to think of the consequences of the various scenarios by themselves. This is a lot more difficult for laypeople. Our 3D simulations transport the viewer into the situation in question, allowing them to actually see the water level during a potential flood. Realistic images like this mean that everyone can visualise the impact of flooding. You could say that it democratises the discussion.

You could say that our 3D models democratise a discussion.

I also try to do this with other data sources: to visualise data in such a way that the desired information becomes accessible to all. Our techniques can be used in a range of applications, such as in architecture, health care and in video games. For example, we use algorithms to create a realistic simulation of how light is distributed. That is useful when you are designing a building, but also when you want to create shadows in a video game.

But how do you avoid such exquisite animations making excessive demands on your computer’s processing power? It is primarily immersive virtual reality environments that consist of an enormous number of images, up to 120 a second. Of course, that looks fantastic, but your computer needs to be able to handle it. With my research group, I look for effective methods of making these programs less demanding. One approach is to use an eye tracker, and only show the part of the screen that you are actually looking at in focus. Another option is to show a series of low-resolution images that together look like high-resolution images. You sometimes have to trick people a little, for their own good”.

You sometimes have to trick people a little, for their own good.

Text: Merijn van Nuland | Photography: Mark Prins