IN4086P Freestyle visualization
In this exercise, you will be designing an original visualisation based on a dataset of your choice. You can also use any visualisation software packages that you have available. We will supply a list of recommended datasets and publically available visualisation software systems that you can choose from as well. This exercise will be done in the same groups of two students as for the other exercises.
You have to motivate your solution by writing a two page report satisfying the following requirements:
- Analyse and describe the problem you have chosen, focusing especially on the questions that you aim to answer with your visualisation.
- Describe and motivate your technical choices.
- Demonstrate how your visualisation helps to answer the questions you described as part of the problem analysis.
- Motivate why you think that you have made an effective visualisation.
Think of this as the documentation that you will deliver to a client when performing an industrial/commercial visualisation contract. Mail this report to cpbotha at-symbol medvis dot org.
Your report will be judged based on the following assessment criteria:
- Presentation: Quality of documentation.
- Ingenuity: Technical challenge involved in the solution.
- Insight: Solid understanding of the issues involved.
- Completeness: The visualisation shows as much as possible of the valuable information contained in the data.
- Effectiveness: The solution adds significant value above a straight-forward visualisation.
Please see the IEEE Visualization 2004 Contest winner reports for examples and ideas on how to write such a short report. Click on the names of the contestants to see their reports.
In short:
|
1. Choose dataset |
The various visualisations will be judged and a winner will be selected. We are still considering what to do about prizes. 
List of datasets (feel free to bring your own!):
IEEE Visualization 2008 contest dataset - Very large multifield scalar time-dependent dataset.
IEEE Vis2004 contest dataset - this one is large, multi-variate and time-varying, but very interesting. There are 48 timesteps of 95 MBytes of data each. You could for example choose to do fewer timesteps, but this will definitely be quite a challenge.
Gordon Kindlmann's synthetic and brain datasets - the diffusion tensor imaging dataset of Gordon's brain. Choose this one if you like challenges and want to play with the Teem software.
Volvis.org has a number of simple volume rendering datasets. The data is simpler, so you'll need to use your imagination to make a good visualization.
The SigGraph Volume Visualization Datasets are also relatively simple (mostly medical) volume datasets, the same applies as in the case of the Volvis.org datasets.
You will find computational fluid dynamics (CFD) datasets here and here. You can also find many CFD example visualizations by googling.
The datasets section of the medvis.org website has links to repositories of medical datasets.
Information Visualisation contest and benchmark datasets from various conferences can be found here. A summary of these contests are given in the paper C. Plaisant et al., "Promoting Insight-Based Evaluation of Visualizations: From Contest To Benchmark Repository", see http://dx.doi.org/10.1109/TVCG.2007.70412
New data from Atmospheric simulations from Prof. Dr. Harm Jonker, in which we need insight in cloud formation (preferably in Paraview). Contact Gerwin de Haan for the data and more information.
Below is a list of visualisation software packages / systems that we have played with. Feel free to use anything else that you can lay your hands on!
The Visualization ToolKit - VTK is very popular in the Visualization world. You can drive this programming library from Python, Tcl, Java or C++.
ParaView is an open source front-end for VTK that can do stand-alone visualization but can also do parallel processing. Binaries for Windows and Linux are freely available.
VolView specialises in volume rendering and is also from Kitware, the makers of VTK. You can download and work with the eval version.
Open DX - an older tool, but still useful.
Mayavi is a Python-based front-end to VTK.
VisIt is yet another VTK-frontend with parallel processing capabilities. VisIt was designed with very large datasets in mind.
Simian is an advanced volume rendering tool; check the gallery for some beautiful examples. Use this if you have the necessary GPU hardware available at home or in your laptop. Other good (and slightly more modern) volume renderers you should have a look at are ImageVis3d and Voreen.
MeVisLab is a hardcore medical visualisation package, also based on blocks-and-lines visual programming, that includes advanced direct volume rendering capabilities.
Google Earth Engine combines the visual power of Google Earth with Geo-processing in large datasets available.