The following content has been provided by the University of Erlangen-Nürnberg.

Thanks for the invitation. As already said, I want to give you a little peek into what

are we doing in what we call visual computing. In fact, I think it's really within the computer

science in one of the disciplines that is pretty far away from software engineering.

In fact, I'm afraid the code that my students write is not really well designed with that

respect. In fact, it's other sorts of questions that interest us. I would like to give you

a little bit insight into what we are working on and what is interesting for us. Good. So,

visual computing, what is it about? First of all, the first thing you usually see when

my students present a paper or so, a very, very important part is always to have a nice

demo in the end, a video that shows something that explains why this is good, what has been

done and so on. So, we always have wonderful demos. I hope so. And it always looks nice,

but students usually don't know in the very beginning that it requires a lot of math.

And it's a typical thing if people start or students start, for instance, working for

a computer games company, they are very often shocked how much math they need. And just

to compute the trajectory of a bullet or whatever in a computer game already requires lots of

physics, math. It's a differential equation that you have to solve. And there are different

methods to do that efficiently with small step sizes and large step sizes. And all that

stuff you hate up to now in your studies suddenly come up again and you have a real world problem

that requires that. That is something I would like to show you also today. But it's not

all. Another important aspect is hardware architectures. We do most of our stuff. We

try to do it in real time. Graphical systems are usually systems where a user interacts

with. So, we have very hard constraints on computation times. Things should be reactive.

And so, it's very important to use, for instance, GPUs for these tasks. And GPUs are highly

parallelized. So, you have to rethink most of the algorithms. How can they implement

it on a GPU so that they run efficiently? And a general thing, good programming skills.

It's a lot of implementation work using libraries from other smart libraries and all that stuff.

Libraries to load 3D models and so on in all different languages. Usually we program in

C++, but lots of the stuff is also with Python and with all languages you can imagine.

Okay, so having said that as a short intro, maybe what is visual computing? Visual computing

has been the term that came up about 10 years ago. And to make things simple, one could

say visual computing is mainly about computer graphics and computer vision. And computer

graphics means generating images from virtual objects. So, we have 3D worlds, models of

persons, cars, whatever. And we want to generate images from them. Usually in a way such that

they look as realistic as possible. So, computer games are a good example. There's lots of

computer graphics in there to make the scenes really look great and have all these lighting

effects and whatever. That's computer graphics. But there's also the exactly opposite direction.

That's computer vision. We have an image given and we want to know how did the scene look

like? The real world scene, how does that look like? So, you can say these are opposite

problems or inverse problems. Okay, and visual computing is now about both of these. There

are many other sub-disciplines, but these are really the core of visual computing. So,

in computer graphics, we have the task generate an image from a 3D model. Here, again, an

example. This is how such a model could look like in a very simple form. That's from the

70s, this model here. And then we have that 3D model in some representation and then we

want to generate nice images from that. So, that's a screenshot from a typical computer

game where you can see computer graphics images of such models. And computer vision is just

the other way around. Here, it's given a real world image and now we want to find out, okay,

here is a car, here is a car, this is the road, these are the road marks and so forth.

So, we want to extract information from images. And now you can see this information can be

on different levels, it can be real 3D points in space, but also just information like this

is a car. Okay, and yeah, this is what we think about in visual computing. And in fact,