06.05.2009

fmx09: Keynote by Thomas Ertl (Uni Stuttgart): Innovation in Visualization

The program for this room/track on innivation of visualization seems in English, so I'll stick to that language for blogging.
***

Ertl is part of the VUS/US Group: Visualisation and Interactivity at the University of Stuttgart.

Computing has moved from cumputing with numbers and text to computing with complex data: images and moving images.

Computers can generate visual data, both abstract and concrete - and they can alayze visual data.

By computer vision you can analyse catured data and compute with it.

Visualization and rendering is used to generate images out of data.

We now have > 10 Megapixel cameras, large storage capacities and Ghz CPUs etc. (og: in 1993 a OCR Unit cist 15.000 DMarks !)

The economic potential is still in its infancy. Lots will be possible.

It is open to entrepreneurs cause you don't have to invest a lot.
The business idea is the key, not company size.
Reserach in Visual Computing is very innovative.

Industrial research groups are abundant, there are lots of academic posotions for visual computing and there are more and more finding opportunities (e.g. Visual Analytics).

[Teaching presentation/side creation skills to academics would also be a great thing ;) -- Oliver]

Locations for reserach in BW are: Ulm, Fellbach, Stuttgart, Tübingen, Konstanz, Freiburg and others.

This is the peta-area (10^15 Bytes). This boundary is reached in petaflops, petabytes of storage, satellite image output etc.

Exa and Zetabyte area will foillow (10^18 ând 10^21)

2010 will see data volumes increasing by 1 zetabyte / year-

1 human "stotres" 1 Petabyte in a lifetime.

So the data will have to be analyzed by machines.

We need interaction (to explore) and visualization to deal with the data.

***
Science uses graphic depiction of abstract data.

We had line drawings for a long time...
Now we map abstract data to (3d) graphical objects in statical images.

Next step is to mal colors to 3d objects in real time.

'Adaptive visualization techniques': chose between speed and resulution.

Then you map 3d data to a 3 d scene and then rendered to a 2d screen.
Problem is: Occlusion: Information hidden bethind other information.

- 'move around the object'
- visualize different 'layers' by transparancy (in a model of the body e-.g.)
- feature extraction

Raw data is filtered ->visualized data is mapped ->renderable represeantaions are visualized.

You need GUPs to do that.

Areas:
- Scientific Visualization (medical, engineering etc.)
- Information Visualization (databases, networks, finance)
- Perception, Interaction, Usability (cognitive science, user studies)
- Technical Infrastructure (GPUs, augmented reality)

GPUs
- they turn into programmable stream processors
- The number crunching performance of graphics cards grows faster than the normal CPU power
- -> GPGPU
- 20 graphics cards drive a high resulution wall (15-20 times higher resuliution from a normal beamer)

Examples
- multi volume visualization (without sharder programming)
- differnet objects in one graph renderen by different methods
- e.g. detect malformations of internal blodd vessels

- head, skull and vessels are renderen by different methods in ONE image

- SimTech excellence Cluster: "hp-refinement in discontinuous Galerkin"
- traditional methods fail here, 30 min rendering time
- either refine the grid (h-refinement)
- or use higher order polinominals (p). 7th order!
-> hp-refinement

- protein Visualization
- 4068 atoms in 263 amino acids
- classical trinalge scaling does not scale here
- bounding box calulation in a vertex shader

- abstract notation of proteines (cartoon representation)
- generated by shader

- solvent excluded surfaces
- GPU gets the abstract structure and calculates the surfaces

- depict fuctional areas of molecules

- systems biology
- signal structures in cells
- what happens when a cell is hit by a signal?(oliver: why does ertl not use video?)

- compare visualizations with real microscopic images

****
Visual Analytics

- Visualize Data Mining Processes
- example : visualize patent data creation in different countries
- provide a cockpit view of interrelated patents on a 'desktop'
-

***
Visualization 2.0:
- UGC is one of the main factors of web 2.0
- Wikipedia is still in te 'plotting age'
- user geerated visualization (for the masses)
- create visual literacy
- e.g. "Many Eyes"
- Google Visaulization API
- Swivel

-> bring social computing into visualization

?: can reserach stand up to Microsoft, IBM and Google?

Sumamry:
- Visualization is a dynamic field
- combines different aspects of reasearch
- and a wide range of disciplines
- ?: What is a good visualization
- Visual Analytics
- user geerated visualization

** END

Facebook Kommentare