Software packages

The scripts below need the natural image statistics toobox natter, which can be cloned from http://kyb.mpg.de/bethge/git/natter.git. To obtain the correct commit use:
git clone  http://kyb.mpg.de/bethge/git/natter.git
cd natter
git checkout 12cdba0f1dcbed6a98089daf493c1469dac7aab0
Due to their size we currently cannot provide the data and precomputed files (largedata.tar.gz and data.tar.gz). Please write me an email if you want to obtain it. Apart from the data, you need python scripts (described in a bit more detail below) and additional scripts. Put them all in the directory where you cloned the natter to and issue:
tar -xzf data.tar.gz
tar -xzf python.tar.gz
tar -xzf largedata.tar.gz
tar -xzf fabee.tar.gz

paths

In your working directory, please create the following paths
mkdir figs
mkdir figs/MI
mkdir figs/temporalSigma
For sampling simulated eye movements, you will also need the van Hateren database. You need to download the .iml images, put it into a directory of your choice and adapt the path in sampleEyeMovementData.py accordingly. We used a cleaned up version of the images in which ignored all images that contained zeros.

python scripts

The dependencies of the python scripts are described below. Some of the files which are computed by the scripts are already contained in data so you should have no problem running any of the scripts below. Note that training models takes a lot of time and memory.

trainDynamicModels.py

needs:

The files in data/. The simulated eye movement data in largedata/.

yields:


evaluateDynamicModels.py

needs:

in data/. The simulated eye movement data in largedata/.

yields:


CVPlot.py

needs:

in data/. The data sets in largedata/.

yields:


trainStaticModels.py

needs:

The files in data/. The datasets in largedata/.

yields:


trainStaticModels2.py

needs:

The files in data/. The datasets in largedata/.

yields:


multiinformation.py

needs:

The files in data/. The datasets in largedata/.

yields:


sampleEyeMovementData.py

needs:

The van Hateren dataset (need to adapt the path in the script).

yields:


radialDistribution.py

needs:

The files in data/. The datasets in largedata/.

yields:


temporalSigmaAdaptation2.py

needs:

The files in data/. The datasets in largedata/.

yields:


sigmaDynamic.py

needs:

The files in data/. The datasets in largedata/.

yields:


dynamicPlot.py

needs:

The files in data/.

yields:

University of Tuebingen BCCN CIN MPI